A Global Analysis of Tortoise and Freshwater Turtle Distributions with Identification of Priority Conservation Areas
Abstract
There are currently ca. 317 recognized species of turtles and tortoises in the world. Of those that have been assessed on the IUCN Red List, 63% are considered threatened, and 10% are critically endangered, with ca. 42% of all known turtle species threatened. Without directed strategic conservation planning, a significant portion of turtle diversity could be lost over the next century. Toward that conservation effort, we compiled museum and literature occurrence records for all of the world's tortoises and freshwater turtle species to determine their distributions and identify priority regions for conservation. We constructed projected range maps for each species by selecting geographic information system–defined hydrologic unit compartments (HUCs) with verified locality points, and then added HUCs that connected known point localities in the same watershed or physiographic region and that had similar habitats and elevations as the verified HUCs. We analyzed a total of 305 turtle species and assigned each to 1 of 7 geographic regions of the world. Patterns of global turtle species distributions were determined and regional areas of turtle species richness identified. In only 2 areas of the world did as many as 18 or 19 species occur together in individual HUCs. We then compared species distributions with existing global conservation strategies (GCSs) and established biodiversity priority areas. Presence of a species in a GCS was defined as ≥ 5% its range. Of the 34 biodiversity hotspots, 28 collectively contain the projected ranges of 192 turtle species, with 74 endemic; the 5 high-biodiversity wilderness areas contain 72 species, with 17 endemic; and 16 other wilderness areas contain 52 species, with 1 endemic. However, 116 turtle species have either < 50% of their ranges in existing GCSs (57 species) or do not occur in them at all (59 species, 19.3%), thus potentially leaving many tortoises and freshwater turtles without any regional GCS. For each of these 116 species we identify a priority Ecoregion for further conservation consideration, and we identify 3 new global Turtle Priority Areas for conservation based on aggregated Ecoregions. These are the Southeastern United States, Lower Gangetic Plain, and Coastal Australia Turtle Priority Areas.
Turtles have existed on Earth since the rise of the dinosaurs. The first fossil with clear turtle affinities is Odontochelys semitestacea from the Triassic of China (Li et al. 2008; Reisz and Head 2008), estimated to be 220 million years old, somewhat older than the earliest fossil turtle with a complete shell, Proganochelys, from the late Triassic of Germany (Gaffney and Meeker 1983; Gaffney 1990; Zug 1993). The turtle shell is a unique and successful body plan that has enabled turtles to persist over 200 million years of changing climates and despite the evolution of a diverse array of vertebrate predators. Today, tortoises and freshwater turtles are represented by as many as 460 taxa (species and subspecies) found throughout the tropical and temperate regions of the world (Iverson 1992b; Iverson et al. 2003; Fritz and Havas 2007; TTWG 2007; Rhodin et al. 2008).
Turtles represent one of the most threatened groups of vertebrates, with 10% of the ca. 317 currently recognized species considered critically endangered on the IUCN Red List of Threatened Species (Turtle Conservation Fund 2002; IUCN 2008) and approximately 63% of the assessed species and ca. 42% of all known species considered threatened (IUCN 2008). Exploitation and unregulated trade are the primary causes for sharp declines in many turtle species, especially those from Asia, with habitat loss and degradation also being major factors in widespread declines (van Dijk et al. 2000, Gibbons et al. 2000, Turtle Conservation Fund 2002). The persistence of such an ancient and iconic group is under concerted assault, and turtles have become prominent casualties of the looming global biodiversity crisis (van Dijk et al. 2000). Without directed strategic conservation planning, a significant portion of turtle diversity could be lost over the next century.
In view of their plight, knowledge of current turtle diversity and global distribution patterns could not be more important. Identification of areas of richness, endemism, and threat enable conservation assessments and prioritization of conservation options (Iverson 1992a; Stuart and Thorbjarnarson 2003; Rhodin 2006). Evaluating species by the size of their geographic range on the landscape provides a first estimation of possible threat to the individual species, and analyses of species' range overlaps reveal patterns of richness and endemism.
Diverse approaches to setting priorities in biodiversity conservation have been used by various conservation organizations. Most of these established templates prioritize areas of high irreplaceability, but differ in their emphasis on high or low vulnerability (Brooks et al. 2006). For example, the biodiversity hotspots (BHs) approach (Mittermeier et al. 1998, 2004; Myers et al. 2000) prioritizes areas of concomitant high irreplaceability and high vulnerability, and the high-biodiversity wilderness areas (HBWAs) approach prioritizes areas of high irreplaceability and low vulnerability (Mittermeier et al. 2003); whereas, the megadiversity countries (Mittermeier et al. 1997) and Global 200 (Olson and Dinerstein 1998) templates prioritize only regions of high irreplaceability. Turtles are disproportionately represented among threatened vertebrate species that require conservation action at the landscape scale (Boyd et al. 2008), but often fall outside traditional conservation priority regions.
Although patterns of species richness for freshwater and terrestrial turtles have been evaluated before (Iverson 1992a), these patterns have not been evaluated in the context of global conservation strategies (GCS). Therefore, our primary objectives were to 1) determine global and regional patterns of species richness and endemism in tortoises and freshwater turtles, 2) evaluate the effectiveness of existing biodiversity conservation strategies at incorporating areas of highest conservation importance for turtles, and 3) identify where significant additional conservation effort is needed by defining areas of high turtle richness and endemism that fall outside the currently recognized global biodiversity conservation strategies.
METHODS
We used a taxonomic list of extant tortoises and freshwater turtle species, totaling 305 species, that we compiled from recent reviews (Iverson 1992b; Ernst et al. 1994; van Dijk et al. 2000; Iverson et al. 2003; Thomson et al. 2006) and primary literature (Starkey et al. 2003; Spinks et al. 2004; Stuart and Parham 2004; Spinks and Shaffer 2007). The final list was a consensus among the authors, noting that the number of recognized turtle taxa is a subject of some contention (Lenk et al. 1999; Fritz and Havas 2007; Stuart and Parham 2007; Turtle Taxonomy Working Group [TTWG] 2007), even among the authors. Subsequent taxonomic changes and controversies, which are accumulating rapidly, are identified in TTWG (2007), Fritz and Havas (2007), and Rhodin et al. (2008). A complete analysis of turtle distributions and conservation areas would include the evaluation of all species, subspecies, evolutionarily significant units, and important management units, but for this initial analysis, we address only species.
Point locality data for all freshwater turtles and tortoises, but not marine turtles, were obtained from museum-verified records, published accounts, and databases (Iverson 1992b; Iverson et al. 2003; Kiester and Bock 2007); from the literature published since 1992; and from unpublished records provided by the authors. We did not attempt to reduce ranges to reflect recent extirpations, nor did we enlarge ranges to account for nonnative introductions.
The continents were subdivided into hydrologic unit compartments (HUCs) that delineate watershed boundaries. HUCs were derived from geographic information system (GIS) layers obtained from the Hydro 1K (1∶1,000,000 scale; USGS EROS Data Center, Sioux Falls SD, http://edc.usgs.gov), Australian River Basins (Geoscience Australia 2002; and World Wildlife Fund (R. Abell and C. Revenga, pers. comm.). We chose these GIS layers as mapping units because delineation methods were fairly uniform across the world; watershed basin HUCs averaged 4000 km2. Because they reflect topography and drainage patterns, HUCs delineate potential ecological boundaries of species distribution around point localities; although, knowing the exact habitats would require actual distributional survey data. HUCs were imported into ArcView 3.3TM and each HUC that included a turtle point locality for a species was included in the overall distribution for that species.
The integration of turtle point locality data and HUCs provided an initial approach to mapping distributions of turtle species. We then constructed total “projected range” maps (hereafter referred to as “range”) for each species by selecting additional HUCs that connected known point localities in the same larger watershed or physiographic region, and contained similar habitats and elevations as the adjacent verified HUCs. Ranges were used to calculate each species distribution in square kilometers in ArcView™. A few HUCs in some species' projected ranges were edited to eliminate areas where no turtles occurred (i.e., HUCs that partially included the high-altitude Himalayas and the central mountains of New Guinea, and nonoccupied portions of unusually large HUCs within the ranges of Testudo horsfieldii in Central Asia and Chelodina steindachneri in Western Australia).
We used the ArcView Spatial Analyst to overlay these species distributions on the following major biogeographic and continental regions: BHs (Mittermeier et al. 2004), wilderness areas (Mittermeier et al. 2003), and terrestrial Ecoregions (Olson et al. 2001). We considered a species to be endemic to any defined region if ≥ 95% of its range was included within that region. A species was considered present if < 95%–5% of its range was included, and in some instances we identify critical range (≥ 50%) and important range (< 50%–5%). To minimize commission errors in our analyses, we discounted (i.e., considered absent) species with < 5% of their range in a given region. However, the exact percentage of occurrence as calculated by our methodology, even if < 5%, is presented in Appendix 1.
RESULTS
Global Distribution of Turtles
Tortoises and freshwater turtles analyzed in our sample set total 305 species in 12 families and are found in 7 major biogeographic regions of the globe (Table 1): 1) North America (United States and Canada; Nearctic); 2) Central America (Mexico to Panama, including the Caribbean; northern Neotropical); 3) South America (southern Neotropical); 4) Mediterranean (Europe and east to the Caspian Sea, the Middle East, and northern coastal Africa; western Palearctic); 5) Sub-Saharan Africa (African continent south of the Saharan Desert, Madagascar and associated oceanic islands; Afrotropical); 6) Asia (Pakistan to Japan, including Indonesian and Philippine archipelagos; Oriental and eastern Palearctic); and 7) Australasia (Australia, New Guinea, and islands east of Weber's line; Australasian).
The use of these biogeographic regions allows for the most parsimonious aggregations of closely related species (i.e., all species in the genus Testudo are found in the Mediterranean; most members of Kinosternidae are found in Central America; Australasia and Asia do not share species). Only 21 of 305 species occur in more than 1 of the 7 regions (Table 1). In terms of phylogenetic depth, South America is the most diverse region with 7 families represented, and Australasia is the least (3 families, Table 2). Land tortoises (Testudinidae) comprise only 45 species (14.8% of 305 total) but are represented across 6 of the 7 global regions (Table 2).
In the northern latitudes, turtles reach lat 56°N in Europe (Emys orbicularis); whereas, Testudo horsfieldii reaches lat 51°N in central Asia. In eastern Asia, Pelodiscus sinensis reaches 52°N. In North America, 2 species (Chrysemys picta and Chelydra serpentina) reach latitudes of 52°N and 53°N, respectively. In the southern latitudes, chelonians reach lat 42°S in South America, represented by Geochelone chilensis. The southernmost reaches of the African continent (lat 35°S) harbor turtles, including 6 sympatric tortoises. The snakeneck, Chelodina longicollis, is found in southernmost mainland Australia (lat 40°S); however, New Zealand (lat 34°S–47°S) lacks native turtles. Large continental areas devoid of turtles include much of Canada, the Rocky Mountains, southern South America, Russia, Mongolia, the Tibetan Plateau, the Sahara, the Arabian Peninsula, and interior Australia (Fig. 1).
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Turtle Richness
Regions of relatively low turtle richness (1–7 species) occur in western North America and Mexico, eastern South America, the Mediterranean, large regions of Sub-Saharan Africa, eastern Asia, and most of Australia (Fig. 1). However, land tortoises (Testudinidae) have their greatest species richness in the southern portions of Sub-Saharan Africa, which includes Madagascar (Table 2). In the tropical regions, much of Central America, most of the Amazon drainage, western coastal Africa, and the northern coasts of Australia have HUCs containing 8 to 10 species. Greater richness (14–16 species) is found in the Amazon's Rio Negro drainage of Brazil (lat 4°S–5°S), the Malaysian Peninsula (lat 1°N–11°N), northern Vietnam (lat 16°N–22°N), the Ganges–Brahmaputra Basin from Bangladesh to the base of the Himalayas (lat 23°N–29°N), and the North American drainages that enter the Gulf of Mexico (lat 29°N–36°N; Fig. 1). Exceptional richness (18–19 species) occurs in very few individual HUCs: in Asia, part of the lower Ganges–Brahmaputra Basin (lat 26°N–28°N, 19 species; Shrestha 1997; Fig. 2a), and a smaller region in North America in the Mobile Basin, Alabama (lat 30°N–31°N, 18 species; Fig. 2b).
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Size of Turtle Ranges
Ranges for 17 species (5.5% of the total) were calculated at less than 10,000 km2 each (Appendix 1). These include Pelusios seychellensis, which has the smallest range (154 km2), endemic to the Seychelles Islands, and most likely extinct (Bour and Gerlach 2008; Gerlach 2008), and Chelodina mccordi, endemic to Roti Island, Indonesia, 1223 km2. Thirty-six species (11.8%) occupy an area of less than 25,000 km2 each (e.g., Graptemys oculifera, endemic to the Pearl River basin, southeastern United States, 22,348 km2; Fig. 3a). Eighty-nine species (29.2%) have an area of between 1 million and 10 million km2 (e.g., Chrysemys picta occupies 5.1 million km2 in North America). The largest range belongs to Pelomedusa subrufa at 16.2 million km2 in the Sub-Saharan African region (Fig. 3b). The largest percentage of turtle species (44%) have ranges falling between 100,000 and 1 million km2 (Fig. 4). Mean range size was 1,076,798 km2, and median range size was 331,919 km2, represented in rank order by Gopherus polyphemus.
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Turtles in GCSs
Twenty-eight of 34 BHs (Mittermeier et al. 2004) collectively contain the ranges for 192 species (Appendix 1; Table 3). Five BHs are known not to contain turtles (Chilean Winter Rainfall–Valdivian Forests, New Caledonia, New Zealand, Polynesia–Micronesia, and East Melanesian Islands) and 1 was discounted in our analyses (mountains of Southwest China) because it contained < 5% of the ranges of 3 species. Individually, BHs contain as few as 1 and as many as 51 species (e.g., Indo-Burma; Table 3). BHs contain ≥ 50% of the ranges of 120 species, with 74 of those endemic to BHs collectively (53 species are each endemic to a single BH and a further 21 to a combination of more than one BH; see Appendix 1). Only 1 species, Cuora amboinensis, occurs in 4 BHs. The Indo-Burma BH contains 15 endemics, Mesoamerica contains 10 endemics, Madagascar and the Indian Ocean Islands contains 7 endemics, and the Caribbean Islands BH contains 4 endemics (all Trachemys sp.).
All 5 HBWAs (Mittermeier et al. 2003) contain turtles (Table 3). Individually, HBWAs contain as few as 11 species and as many as 20 (Amazonia; Table 3). Collectively, 72 species are present. Of those, HBWAs contain ≥ 50% of the ranges of 40 species, and 17 are endemic. Only 1 species, Pelomedusa subrufa, is present in 2 HBWAs, Miombe–Mopane Woodlands and Congo Forests.
Of the 24 Other Wilderness Areas (OWAs; Mittermeier et al. 2003; Table 3), 5 do not contain turtles (Antarctica, Arctic Tundra, Greenland, Magellanic Forests, and Tasmania), and 3 more were discounted (Patagonia, Pacific Northwest, and Boreal Forests) because they contained < 5% of the ranges of up to 5 species. The remaining 16 OWAs contain 52 species. Of those, OWAs collectively contain ≥ 50% of the ranges of 12 species. One species is endemic to a single OWA (Elseya “South Alligator”; Arnhem Land Tropical Savanna) and 1 nearly so (93%, Acanthochelys pallidipectoris; Chaco). Twenty species have ranges in multiple (up to 3) OWAs (Appendix 1).
In combination, BH, HBWA, and OWA GCSs capture 106 turtle species as endemic (≥ 95%), 34.8% of the world total (305). An additional 140 species are present (< 95%–5%), and these are subdivided into those < 95%–50% present (83 species) and those < 50% present (57 species). However, 59 species (19.3%) are absent (< 5%) from these GCSs (Table 4), and the number of species either < 50% present or absent from GCSs is 116 (38.0%).
Turtles and Ecoregion-Focused Conservation
Of 867 Ecoregions worldwide (Olson and Dinerstein 1998; Olson et al. 2001), 680 include the ranges of turtle species. However, we excluded 330 of these 680, as well as the “Lake” Ecoregion (which is not unique to any continent) because of minimal overlap (< 5%) with turtle species ranges. Hence, 349 Ecoregions each contain ≥ 5% of varied numbers of species, ranging from 1 to 29 (Appendix 2). The top 5 Ecoregions of the world for turtles include the Southeastern Mixed Forest (United States, 29 species), the Southeastern Conifer Forest (United States, 25 species), the Northern Indochina Subtropical Forest (Vietnam, Laos, Myanmar, and China; 21 species), the Lower Gangetic Plain Moist Deciduous Forest (India, 18 species), and the Central Forest/Grasslands Transition Zone (United States, 18 species). The Northern Indochina Subtropical Forest is located in the Indo-Burma BH, but it is striking that the remaining top 4 world Ecoregions for turtles are found outside of the GCS schemes that we considered.
For each of the 116 turtle species that have either < 50% or no GCS coverage we identified a “first-priority” Ecoregion based on the species' greatest percent range of occurrence (Appendix 1), and exclusive of GCS. The 45 Ecoregions in this category are counted by global geographic region in Table 4 and listed in Appendix 1. Twelve of the 116 species are endemic to single Ecoregions (Australia, 5 species; China, 4; Congo, 1; United States, 2).
New Global Turtle Priority Areas (TPAs)
Toward our goal of identifying areas of global turtle species richness that were outside of existing GCSs, we identified 3 new global Turtle Priority Areas (TPAs) that collectively include 72 (62%) of the 116 species without adequate GCS coverage. As previously noted, some of these species have partial ranges in existing GCSs, but the following areas of high species richness lie outside of GCS regions.
Southeastern United States TPA
North America is the primary region for 48 species (Table 2). Three of these are included in GCSs (Gopherus agassizii, Actinemys marmorata, and Kinosternon sonoriense), and the remaining 45 each have ranges of < 50% (4 species) in GCSs or are absent (41 species; Appendix 1). Five Ecoregions collectively comprise a Southeastern United States TPA that includes portions or all of the range for 43 of the 45 species; only Glyptemys insculpta and Emydoidea blandingii are not included (Fig. 5): 1) Southeastern Mixed Forests, 29 species, 2) Southeastern Conifer Forests, 25 species, 3) Mississippi Lowland Forests, 10 species, 4) Piney Woods Forests, 13 species, and 5) Edwards Plateau Savanna, 3 species.
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Lower Gangetic Plain TPA
Asia is the primary region for 77 species (Table 2). Most Asian species are included in GCSs, but 18 have < 50% of their ranges in GCSs, and 6 are absent (24 species; Table 4, Appendix 1). However, 10 of these 24 are Chinese endemics or nearly so and 14 are found on the Indian subcontinent. The Lower Gangetic Plain Moist Deciduous Forests Ecoregion lies adjacent to the Indo-Burma and Himalaya BHs (Fig. 6) and contains 18 species, which includes 12 of the 14 Indian species in need of conservation coverage; only 2 Indian species, Geochelone elegans and Aspideretes leithii, occur in other areas.
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
Coastal Australia TPA
Australasia is the primary region for 35 species (Table 2). Twenty-five are included in GCSs, but 6 have < 50% of their ranges in GCSs, and 10 are absent from GCSs. Thus, 16 species in Australia have > 50% of their range outside of GCSs (Appendix 1). The Coastal Australia TPA complements the existing Kimberly, Arnhem Land, and Cape York Tropical Savanna OWAs and includes the following Ecoregions: 1) Carpentaria Tropical Savanna, 7 species, 2) Brigalow Tropical Savanna, 9 species, 3) Queensland Tropical Rain Forest, 1 endemic species, Elseya “Johnstone” (A. Georges, unpubl. data), and 4) Eastern Australia Temperate Forests, 9 species (Fig. 7). The Coastal Australia TPA collectively captures 15%–100% of the ranges of all 16 species and contains some portion of the ranges of 22 species (Appendix 1).
Citation: Chelonian Conservation and Biology 8, 2; 10.2744/CCB-0774.1
DISCUSSION
HUC Methodology
Knowledge of species distributions in certain global regions is thorough, down to local watershed levels (e.g., North America). For species in these areas, ranges and actual known distributions are effectively the same. In other regions of the world (e.g., Sub-Saharan Africa, Asia), museum and other known distributional data are limited, and therefore, ranges are extrapolations based on expert opinion, similar physiography and habitats, elevation, and drainage connections. At the broadest, most applied scale, the use of HUCs (watersheds) to map turtle distributions is an ecologically valid approach. However, all ranges overestimate the actual habitat available to each species, such as erroneously extending the ranges of turtles to high elevations (i.e., “Sky Islands” of the Madrean Pine-Oak Woodlands BH) when the species are restricted to the lowlands of the watershed, or extending turtle ranges into adjacent, but unoccupied, Ecoregions. Only 1 turtle species, Kinosternon oaxacae, has a significant portion (> 50%) of its range in the Madrean Pine–Oak Woodlands, but at lower elevations the HUCs clearly delineate the primary watershed conservation boundary for the turtles. Therefore, HUC-derived ranges are representative of landscape-level distributions, especially for many freshwater turtles where identification of drainage basin catchment boundaries, such as to control pollution inputs, is the first step in drawing accurate conservation boundaries.
The HUC approach to mapping distributions might arguably be less accurate for tortoises (Testudinidae) because these species are less likely to be restricted by drainages and are more capable of crossing drainage divides than some aquatic turtles. Future refinement of individual HUC ranges should be made by removing portions along ecoregional boundaries and elevational contours, when those aspects of a species distribution are well known.
Global Richness Patterns
Our primary objectives were to determine the global and continental patterns of species richness and endemism in tortoises and freshwater turtles, evaluate how well existing biodiversity conservation strategies overlap with the distributions of turtles, and identify areas of high turtle richness and endemism that fall outside the currently recognized GCSs for biodiversity, and on which future efforts should be focused.
The compilation of all ranges for the 305 turtle species using HUCs (Fig. 1) resulted in the identification of many areas of high species richness within the 7 global regions. Areas with up to 10 co-occurring species are found on the northern and northeastern coasts of Australia, the western coastal belt of Sub-Saharan Africa from Liberia southward to Gabon, and much of southeast Asia including southern China, Vietnam, and Cambodia, as well as Borneo. Areas where more than 11 species are likely to co-occur include the Mississippi and Atlantic Coast drainages of southeastern United States, the upper Amazon River drainage of Brazil and Colombia, the Ganges and Brahmaputra drainages of India and Bangladesh, the Irrawaddy drainage of Myanmar, the Salween drainage of southern Thailand south throughout Malaysia, and the coastal drainages of Sumatra. The areas of the world where the greatest species density is found include the upper portions of the Ganges and Brahmaputra Basin in India and the lower Mobile Bay drainage along the Gulf Coast of North America (Figs. 1–2).
Our 7 global turtle regions correspond loosely with biogeographic realms (Olson et al. 2001), with the exception that we categorized Central America, including Mexico, as a separate “northern Neotropical” realm, as similarly recognized by Bour (2008). Central America represents a wide region of overlap of Nearctic and Neotropical turtles, but also contains a sizeable endemic turtle fauna exemplified by the genera Dermatemys, Claudius, and Staurotypus, and large radiations within the genera Trachemys, Kinosternon, and Rhinoclemmys. Our 7 regions also minimized overlap of species between regions. For example, the turtle faunas on the African continent were assigned to 1 of 2 groups: 1) those species existing south of the Saharan Desert, and 2) those found on the north coast along the Mediterranean Sea, which are biogeographically allied with European and Middle East species (e.g., the genera Mauremys and Testudo). Only 2 of 48 Sub-Saharan species (Pelomedusa subrufa and Trionyx triunguis) enter our defined Mediterranean region. Turtles are largely absent from the vast majority of the Palearctic realm, and those that are present in the western Palearctic are derived from the Mediterranean region above; whereas, in the eastern Palearctic (i.e., northeast China and Japan, Korea, and Siberia) the turtle fauna is Indo-Malayan in origin. Also, as with many other faunas (Lomolino et al. 2006), Weber's Line effectively separates Asian from Australasian turtles.
Although turtles have had a successful 200+ million year history, their living diversity is among the lowest of major vertebrate clades (both older and younger). In addition, unlike most groups of organisms (Lomolino et al. 2006), they do not exhibit a pattern of highest diversity in the tropics. Indeed, their greatest diversities are reached at ca. lat 23°N–24°N (Ganges River basin in Asia) and lat 31°N–32° N (Mobile River basin in North America). The combination of their generally low richness, their unusual distribution patterns, and their unusual life history strategies (Heppell 1998) make them especially difficult to conserve.
Existing GCSs
Our analyses identified turtle and tortoise species that would, in theory, be afforded conservation attention, and possibly protection, under existing GCSs; although, occurrence in these areas certainly carries no guarantee of protection. By identifying these species and also the proportions of their ranges that coincide with GCS regions, we can evaluate the likely importance and effectiveness of particular conservation strategies, as well as the seriousness of landscape-level threats.
BHs are human constructs, ecologically characterized, but defined by disparate levels of human activity (i.e., BHs have lost 70% or more of their original native vegetation; Myers et al. 2000). The 34 BHs hold especially high numbers of endemic species, but their combined area of remaining habitat covers only 2.3% of the earth's land surface; over 50% of the world's plant species and 42% of all terrestrial vertebrate species are endemic to the hotspots (Mittermeier et al. 2004). Clearly, the most important hotspot for turtle conservation is Indo-Burma with 15 endemics and another 15 species with > 50% of their ranges encompassed therein. Similarly, the Mesoamerica BH has 10 species in each of those same 2 categories, and although they contain smaller numbers of species, it must be noted that the entire turtle faunas of Southwest Australia, Madagascar and the Indian Ocean islands, and the Caribbean islands are fully encompassed within their respective BHs.
Our analysis indicated that 74 turtle species (24%) are endemic to BHs, a figure that is significantly lower than the overall endemic terrestrial vertebrate percentage (42%; Mittermeier et al. 2004). However, conservation resources are flowing to BHs (Myers 2003), and the species that are endemic to BHs often receive considerable conservation attention because of the focus on hotspot conservation. Nevertheless, the persistence of these species in the wild depends on conservation success (sensu Brooks et al. 2002) in their respective hotspots of occurrence. Thus, although the identification of BHs has focused attention on their conservation, even if their protection was successful across the globe, only 120 turtle species (39% of the total) have ≥ 50% of their ranges within these hotspots. Hence, global turtle conservation must look well beyond the possible protection offered by the conservation of BHs.
Five HBWAs (Amazonia, Congo, New Guinea, North American Deserts, and Miombo–Mopane Woodlands) are large areas of exceptional diversity that remain mostly intact, with greater than 70% of their natural land cover remaining and with relatively low human density and threats (Mittermeier et al. 1998, 2003). Indeed, HBWAs in New Guinea and Amazonia rank below only the top 2 BHs (Indo-Burma and Mesoamerica) in terms of endemic turtle fauna. In theory, the 17 turtle species endemic to the HBWAs should be some of the most secure because by definition more than 70% of their habitat remains. An assumption, however, is that other threats, including human encroachment, disease, pollution, climate change, and exploitation (sensu Gibbons et al. 2000) are not impacting those turtles at high levels. Unfortunately, they are, as is the case with the exploitation of Amazonian river turtles for food (Ojasti 1996; Moll and Moll 2004; Conway-Gómez 2007) and the extraction for bush meat in the Congo (Luiselli 2003; Maran and Pauwels 2005). These 17 endemic species represent 6% of turtles, which is also lower than the overall endemism in other groups (18% of the world's plants and 10% of all terrestrial vertebrates; Mittermeier et al. 2003).
Of the 24 OWAs analyzed, some are large (i.e., Antarctica, Arctic Tundra), but do not contain turtles. Most of the others are smaller than 1 million km2. However, several OWAs individually include > 50% of the ranges of 9 turtle species (e.g., Banados del Este [Trachemys dorbigni]; Chaco [Acanthochelys pallidipectoris and Geochelone petersi]; Llanos [Podocnemis vogli]; Sahel [Geochelone sulcata]; Central Asian Deserts [Testudo horsfieldii]; Kimberly Tropical Savanna [Emydura victoriae and Elseya dentata]); and Arnhem Land Tropical Savanna [Elseya “South Alligator”]). Collectively, OWAs encompass ≥ 50% of the ranges of 12 species (the 9 above, plus Phrynops williamsi, Chelodina burrungandjii, and Emydura tanybaraga; Appendix 1).
Ecoregion Approach
When all the above GCS strategies are combined, 246 of 305 (80.7%) turtle species are addressed at some level. However, additional conservation strategies must be invoked if we are to meet the goal of “no turtle left behind” for the other 19.3%. Thus, we prioritized individual or groups of Ecoregions for additional turtle conservation strategies, designating 3 new TPAs. For this paper we did not include priority Ecoregions for every turtle species, but instead presented Ecoregions only for those species that must rely solely on them for conservation (i.e., they fall outside of BHs, HBWAs, and OWAs; Appendix 1). Ecoregions are important in conservation planning at regional levels (e.g., The Nature Conservancy's Ecoregion planning process, Partners for Amphibian and Reptile Conservation (PARC) habitat management guidelines–North America; Bailey et al. 2006) and identifying species at the Ecoregion level allows for maintenance of ecological processes on the local landscape. Ecoregion conservation is clearly important for the 12 turtle species that are endemic to single Ecoregions, including some of the rarest species in the genus Cuora in Asia (Appendix 1).
We assigned species to a priority Ecoregion based on the species' greatest percentage of range. This does not mean that we consider that Ecoregion to necessarily be the most important for the conservation of that species; it simply means that because the species has a significant portion of its range there, it would probably be a reasonable area to consider initially for protection. It is also necessary to look at subsequently ranked Ecoregions because, by selecting a first-priority Ecoregion based on greatest range, we may not address the needs for species conservation at the periphery of a species' range. We recognize that field inventory of status and threats, assessment of populations and their sizes, determination of quality of habitat, and the ability to achieve conservation (i.e., political will, local capacity, funding, etc.) will all combine to determine the most effective priority Ecoregion for each species. We also note that future analyses may integrate turtle ranges with recent freshwater Ecoregion delineations (Abell et al. 2008).
In North America, 14 Ecoregions were identified for 45 species. Many species clustered in several adjacent Ecoregions in the southeastern United States (Southeastern United States TPA). The two most turtle-species-rich Ecoregions in the world, Southeastern Mixed Forests and Southeastern Conifer Forests occur here (Fig. 5). However, the Blanding's turtle (Emydoidea blandingii) was identified as a focal species in conservation planning for the Great Lakes Ecoregion; whereas, the wood turtle (Glyptemys insculpta) was assigned to the New England Acadian Forests because no broader-level strategy previously identified will properly address these species' needs. For 3 wide-ranging North American species (Apalone spinifera, Chelydra serpentina, and Chrysemys picta), each of their ranges spanned 23–29 Ecoregions with < 10% of their range in any one; thus, no single Ecoregion was selected for their conservation.
In Central America, the occurrence of BHs and HBWAs resulted in the inclusion of many turtle species, but some Ecoregions were outside the GCS, including the Tamaulipan Mezquital, important for Gopherus berlandieri and Pseudemys gorzugi. For future analyses, some species (i.e., Trachemys yaquia and Kinosternon alamosae) marginally missed our cutoff for needing Ecoregion-focused attention, but would likely benefit from protection in the Sonoran–Sinaloan transition Ecoregion. Likewise, the Caatinga and the Argentine Monte Ecoregions lie outside of GCSs in South America but are important for Batrachemys tuberculata and Geochelone chilensis, respectively. Overall, 5 South American species would benefit from conservation focused on these Ecoregions.
Most Mediterranean species had substantial overlap with the Mediterranean Basin BH, but Emys orbicularis would likely benefit from additional conservation attention in the Central European Mixed Forests Ecoregion.
The distributions of turtles in Sub-Saharan Africa are poorly known (Bour 1983), and ranges were difficult to construct because museum data were sparser than for other regions (Iverson 1992b). The percentages of some species' ranges falling in or out of GCSs were difficult to determine with certainty because HUC layers for Africa do not always align well with the boundaries of Ecoregions and GCSs. Nevertheless, at least 23 species have > 50% of their ranges outside of GCSs, and several Ecoregions emerged as priorities for turtle conservation, including the West Sudanian Savanna as a priority area for at least 5 species and the Nama Karoo for 3 species (see Appendix 1). Pelusios broadleyi would benefit from focus on the Masai Xeric Grasslands and Shrublands Ecoregion. Further detailed analysis of the distributions of individual African species is needed for conservation planning (Burgess et al. 2005, Luiselli 2008).
Asia could be treated as 2 separate regions because it contains the fauna of the Indian subcontinent and that of Southeast Asia. The Indo-Burma BH contains species from both India and Southeast Asia but contains the greatest percentage range for the latter. The Ganges–Brahmaputra region of India and Bangladesh exhibits the greatest range overlap for turtles in a single area in the world, with 19 species known from each of 4 connected HUCs (Figs. 2a, 6; Lower Gangetic Plain TPA; see also Iverson 1992a, 1992c). The richness of turtles in the Ganges–Brahmaputra Basin is in part because of the overlap in the 2 faunas. The Lower Gangetic Plains Moist Deciduous Forest Ecoregion, representing the third-most species-rich turtle Ecoregion in the world (18 species), accounted for 12 species that were not previously covered by GCSs (Appendix 1). The Upper Gangetic Plains Moist Deciduous Forests Ecoregion, although clearly needed for conservation of the Ganges River system and its fauna, does not gain any more turtle species, just more range of the same species. Other Ecoregions that include the same high turtle species richness (but are within the Indo-Burma and Himalaya BHs) include Meghalaya Subtropical Forests, Brahmaputra Valley Evergreen Forests, Mizoram–Manipur–Kachin Rainforests, and Terai–Duar Savannas and Grasslands. Both the Deccan Thorn Scrub and the Eastern Highlands Moist Deciduous Forests are important Ecoregions for 3 species (Lissemys punctata, Geochelone elegans, and Aspideretes leithii) because these species are 84%–92% outside of GCSs. India has one of the most diverse turtle faunas with 28 species, and ranks among the top 5 countries in terms of importance for turtle conservation in Asia (Stuart and Thorjarnarson 2003) and the world (Rhodin 2006). In addition, climate change studies suggest that glacial melt in the Himalaya will affect water flow in the Ganges Basin (Xu et al. 2009), thus adding another conservation concern to this priority area.
There are 12 Southeast Asian species that are not accounted for in the Indo-Burma BH, and the Ecoregions needed for their conservation include the Jian Nan Subtropical Evergreen Forest (5 species), Chiangjiang Plain Evergreen Forest (3 species), the Yunnan Plateau Subtropical Evergreen Forest (3 species, all presumably endemic: Cuora mccordi, Cuora yunnanensis, Cuora zhoui), and the Qin Lin Mountains or Daba Mountains (1 species, Cuora pani).
For Australasia, those species that occur in Ecoregions in New Guinea (Southern New Guinea Freshwater Swamps and Lowland Rain Forests, and Trans-Fly Savanna Grasslands) are accounted for in the New Guinea HBWA. However, there are 16 species in Australia that were not adequately addressed by GCSs, for which we identified the Coastal Australian TPA. Three OWAs in northern coastal Australia correspond with the boundaries of Ecoregions of the same name (Kimberly, Arnhem Land, and Cape York Tropical Savannas). Additional Ecoregions for Australian turtle conservation include the Carpentaria Tropical Savanna, Brigalow Tropical Savanna, Queensland Tropical Rain Forest, and Eastern Australia Temperate Forests. Other Ecoregions that are important for some of the same turtles include the Victoria Plains Tropical Savanna, the Western Australia Mugla Shrublands, and the Einasleigh Upland Savanna.
Our analysis demonstrates that even if we assume that conservation actions will be effective under current GCSs and cover all turtle species within these geographic areas, at least 59 species fall completely outside that potential protection. We hope our classification of 3 new TPAs will create awareness of landscape-level protection opportunities for conservation of additional species in North America, Asia, and Australia. By our calculations, with the addition of our 3 TPAs to the existing GCSs discussed in this paper, only 10 species would not have been addressed. These include Emydoidea blandingii and Glyptemys insculpta (North America), Geochelone chilensis (South America), Pelusios broadleyi (Sub-Saharan Africa), and 6 Cuora species (Cuora aurocapitata, Cuora flavomarginata, C. mccordi, C. pani, C. yunnanensis, C. zhoui; Asia); however, we list Ecoregions for these species for future focus in Appendix 1.
Clearly, conservation strategies that target species-specific action (Buhlmann and Gibbons 1997; Rodrigues 2006), improve coverage of protected areas worldwide (Rodrigues et al. 2004), and address threats to turtles specifically (Turtle Conservation Fund 2002) and in the larger landscape (Boyd et al. 2008) must be employed collectively and synergistically in order to include and hopefully protect every species effectively under a conservation umbrella strategy. Our goal with this analysis was to develop an improved strategy to “leave no turtle behind” in the global race to conserve this unique and imperiled group of vertebrates.
Global patterns of species richness based on projected ranges in hydrologic unit compartments of all 305 species of tortoises (45) and freshwater turtles (260) included in this analysis. Scale of color codes indicates number of species for each area.
a) The world's greatest turtle species richness area, based on the co-occurrence of species in hydrologic unit compartments in the Ganges–Brahmaputra river basin drainages of India and Bangladesh in South Asia. b) The world's second-greatest turtle richness area, centered on the Mobile River basin in the southeastern United States. Color codes as in Fig. 1.
Examples of relative numbers of hydrologic unit compartments that comprise turtle ranges: a) Graptemys oculifera, 22,348 km2, southeastern United States; b) Pelomedusa subrufa, 16.2 million km2, Sub-Saharan Africa. Figs. 3a, b are not at equivalent scales.
Range distributions (in km2 of projected hydrologic unit compartments ranges) for 305 turtle species in this analysis.
Southeastern United States turtle priority area (TPA) for 43 turtle species not included in existing global conservation strategy areas. Ecoregions comprising the TPA include 1) Southeastern Mixed Forests, 2) Southeastern Conifer Forests, 3) Mississippi Lowland Forests, 4) Piney Woods Forests, and 5) Edwards Plateau Savanna. Vertical green striping indicates an adjacent high-biodiversity wilderness area (North American Deserts); diagonal red striping indicates nearby biodiversity hotspots (Mesoamerica and Caribbean). Color codes as in Fig. 1.
Lower Gangetic Plain turtle priority area (TPA) for 18 species, including 12 species not covered in existing global conservation strategies. A single Ecoregion encompasses the following TPA: 1) Lower Gangetic Plains Moist Deciduous Forests. Diagonal red striping indicates adjacent biodiversity hotspots: 3) Indo-Burma; 4) Himalayas; horizontal purple striping indicates an adjacent OWA: 2) Sundarbans. Color codes as in Fig. 1.
Coastal Australia turtle priority area (TPA) includes for 16 species not covered in existing global conservation strategies. Ecoregions comprising the TPA include the following: 1) Carpentaria Tropical Savanna, 2) Queenland Rainforest, 3) Brigalow Tropical Savanna, and 4) Eastern Australia Temperate Forest. Vertical green striping indicates a nearby high-biodiversity wilderness area (New Guinea), diagonal red striping indicates nearby biodiversity hotspots (Southwest Australia, Wallacea; horizontal purple striping indicates adjacent other wilderness areas: 5) Kimberly Tropical Savanna, 6) Arnhem Land Tropical Savanna, 7) Cape York Tropical Savanna, and 8) Australian Deserts. Color codes as in Fig. 1.
