A global assessment of endemism and species richness across island and mainland regions

Summing up the range equivalents of vascular plants for all 90 regions yields a total of 315,903. Given that the sum of range equivalents in the entire study area should be equal to the total number of species and assuming the data underlying our study accurately capture regional species numbers and degree of endemism, this suggests that recent estimates of global vascular plant richness ranging up to 422,000 species (29–31) may be too high. We found endemism richness of vascular plants to be geographically highly unevenly distributed (Fig. 1A) and it varied by 3 orders of magnitude (see Table S1). From a global perspective, the island of New Caledonia had by far the highest value with 1,350 range equivalents per 10,000 km²—a similar result has emerged from a coarse-scaled, grid-based study at the family level (32). Values were also high for other island regions, such as Polynesia–Micronesia, the Eastern Pacific, and the Atlantic Islands. Although accounting for only 3.6% of the terrestrial surface of the world, 26.1% of all range equivalents of plants are allotted to the 14 island regions. When standardized at 10,000 km², we found endemism richness to be 9.5 times higher on islands than in mainland regions (172.3 and 18.2 range equivalents per 10,000 km², respectively).

The island regions spanned a smaller latitudinal extent and were closer to the equator, possibly distorting the overall picture as species richness increases and average range size decreases toward the equator. We therefore repeated the calculation, restricting the mainland regions to the same latitudinal extent by excluding the 5 northernmost regions. As to be expected, this led to an increase in endemism richness on mainlands, but mainland regions still showed much lower average endemism richness (23.8 range equivalents per 10,000 km²).

Half of the top 20 regions in terms of endemism richness per standard area were island regions (Fig. 1A, Fig. S1). In fact, all but one (Japan) ranked among the top third of the 90 regions and all island regions had above average ranks (F_1,88 = 31.5, P < 0.001; Fig. 1A). The top 30 regions contained 51.6% of the total range equivalents of vascular plants on only 7.4% of the Earth's terrestrial surface (Fig. 3A). Qualitatively similar spatial patterns of endemism richness across the 90 regions were also found for terrestrial vertebrates (Fig. 1 B–F). Island regions had significantly higher endemism richness of vertebrates (8.1 times higher than for mainland regions) and islands contained 23.2% of the range equivalents of this group (Table S2). To test whether the plant-based delineation of biogeographic regions or the relatively coarse resolution of our mapping scheme had an effect on these findings, we divided up the 867 ecoregions (33) into mainlands and islands and compared their endemism richness. Endemism richness of island ecoregions for all 4 individual vertebrate classes and terrestrial vertebrates combined was significantly higher than for mainland regions (Fig. S2), which demonstrated that these results were robust toward the geographic delineation and across spatial scales.

Endemism richness of plants showed a relatively strong correlation with the entire group of vertebrates (r_s = 0.83), indicating a generally strong congruence of global patterns of endemism richness. The overall trend was similar, but stronger than recently found for patterns of richness (34). However, the degree of correlation within the 4 classes differed markedly (individual vertebrate classes: r_s = 0.73−0.82; Fig. 2 and Fig. S3). Furthermore, there were also striking exceptions, especially in the case of amphibians. The 4 vertebrate classes also performed differently in the degree of their range equivalents being captured by a ranking of areas on the basis of plant endemism richness per standard area. Among all vertebrate classes, amphibians were best represented by the plant-based ranking. On the other hand, relatively fewer mammalian range equivalents were covered (Fig. 3B). For instance, 48.5% of all amphibian but only 31.9% of all mammalian range equivalents were covered by the top 30 regions in terms of plant endemism richness. Together, the results highlight a strong role of taxon-specific characteristics [e.g., the relatively poor ability of amphibians to cross saltwater barriers (35)] on patterns of cross-taxon congruence limiting the use of surrogate taxa in conservation.

Our results clearly identified islands as global centers of endemism richness, a pattern that was consistent across plants and vertebrates (although least pronounced for amphibians) (Figs. 1, 2, and Fig. S3). The very high values of endemism richness on islands are especially noteworthy because species richness on islands is generally lower than on mainland sites with comparable climate (20, 36). Island floras and faunas are usually recognized to maintain a high degree of endemism because of their geographic isolation and the limited interchange with neighboring mainland or island biota. Moreover, volcanic archipelagos like the Canary Islands or Hawaii are good examples of relatively recent and rapid adaptive radiations that have resulted in many neoendemic taxa (20, 37, 38). On the other hand, some ancient continental fragments like New Caledonia, Madagascar, the Seychelles, or New Zealand harbor very distinct paleoendemic lineages (20). Similarly, also some oceanic islands and archipelagos also have a high proportion of relict island species and clades that show a remarkable resistance toward massive historical climatic changes that have caused their mainland relatives to become extinct (39). An intrinsic feature of islands may indeed be that they are climatically buffered by the thermally relatively unresponsive ocean masses (39). Both mechanisms, persistence of old lineages and adaptive diversifications, contribute to the high levels of insular endemism richness, but their relative importance remains to be quantified. At the same time, paleoendemic island lineages might deserve particular conservation attention, as these species are taxonomically isolated and represent a larger amount of unique evolutionary history than closely related neoendemic lineages (40).

Among mainland areas, regions with Mediterranean-type climate emerge as global centers of plant endemism richness (Fig. 1A). The South African Cape region, a prime example of exceptionally high extratropical plant richness and endemism (41–45), ranked second among all 90 biogeographic regions (Fig. S1). Interestingly, the high degree of endemism and the contribution of only a limited number of “Cape floral clades” to the astonishing richness of the region have been attributed to island-like biogeographic and evolutionary processes (42). The large peninsula of the Cape is surrounded by oceans on 3 sides and by desert regions to the north and it is thus geographically very isolated (42). A similar interplay between a unique climatic setting, geographic isolation, and high rates of in situ speciation might possibly explain the high endemism richness of plants in other “island-like” continental areas such as, e.g., the SW Australian floristic province or the Queensland tropical rainforests. Regions with wet tropical climate and complex topography emerged as mainland centers of plant endemism richness and shared centers across taxa. For instance, the northern Andes ranked ninth in terms of endemism richness of plants and seventh for terrestrial vertebrates, respectively (Fig. S1, Tables S1 and S3). This region has been previously shown to be a congruency center of different aspects of diversity such as species richness, threat, and endemism for vertebrates (15). Our results confirm this globally exceptional regional concentration of biodiversity from a plant perspective. The high values of tropical montane areas might be partly because of steep elevational and climatic gradients resulting in high species turnover (41, 46, 47). These regions are also regarded as both museums and cradles of biodiversity, where old taxa have survived, among others, because of relatively stable climates or the mitigation of climate change impacts by altitudinal range shifts. Additionally, new taxa are being rapidly generated because of new ecological opportunities caused by recent uplifts. In combination with limited gene flow and effective dispersal barriers, this has led to considerable radiations in many clades resulting in high numbers of range-restricted species (48–50).

The high values of endemism richness on islands emphasize their outstanding importance for global conservation of genetic resources. But their limited area (Fig. 4A) may make them especially vulnerable to anthropogenic impacts. We thus quantified the spatial relationship between endemism richness and various measures of past and projected future human impact. A significant relationship existed between plant endemism richness and past habitat loss (r_s = 0.24, P = 0.02, Fig. S4), but no significant difference was found between island and mainland (F_1,88 = 1.3; P = 0.26; Fig. 4B). While this indicates that island and mainland regions were sim-ilarly affected by habitat loss in the past, there are pronounced differences in the level of current threat and projected land cover change. The “human impact index,” an index quantifying the worldwide presence of humans per 1 km² grid cell (51), was significantly higher for islands (15.7) than for mainlands (9.8) (F_1,88 = 7.7; P < 0.01; Fig. 4C). “Human impact” showed a strong positive correlation with endemism richness (r_s = 0.45, P < 0.001). When we examined projected future land cover changes driven by either direct human activity (agriculture, deforestation, urbanization) or climate change as predicted for the year 2100 by the Millennium Ecosystem Assessment (52), an apparent dichotomy surfaced between island and mainland regions. Whereas future habitat loss driven by land-use change is projected to accelerate for island regions (F_1,88 = 2.8; P = 0.097; Fig. 4D), mainland regions are predicted to lose more of their original land cover because of climate change (F_1,88 = 3.0; P = 0.088; Fig. 4E). Although the 2 latter observations are not statistically significant, they suggest differences which might be explained by 2 island-specific characteristics. First, their oceanic setting might buffer islands from more severe impacts caused by climate change except for sea-level rise. Second, their small area and already high degree of infrastructure (compare high human impact index for islands, Fig. 4C) make islands more vulnerable to habitat destruction as access to remote parts with remaining primary vegetation is comparatively easy. The discrepancy between past and potential future threats of biodiversity may have strong implications for conservation and represents one of the most pressing questions for conservation planners (53). Importantly, the greater risk of islands toward habitat loss and their considerably higher levels of endemism richness is currently not reflected in the protected area network where islands (8.3% of area protected) are less protected than mainlands (10.6%) (Fig. 4F). On the basis of these findings, we suggest that conservation of island biodiversity requires the expansion of existing protected area networks to counteract the threats that island diversity faces as a result of the extraordinarily high levels of human impact. Furthermore, effective measures need to be taken to address threats that cannot be mitigated by protected areas alone, including threats arising from invasive species (54, 55). Island populations have been shown to be particularly susceptible to biotic perturbations (56), which are likely to increase with climate-change induced distributional shifts and increasing global trade. Such increased threats from invasive species require thoughtful measures in addition to the designation and effective management of protected areas.

There is high overlap between regions of highest endemism richness and “biodiversity hotspots” (5). This is not surprising given their two main criteria: ≥ 1,500 endemic plant species and the loss of ≥70% of its original vegetation because of human activities (1). Among the 20 regions with highest endemism richness per standard area, sixteen overlap with recognized hotspots (Fig. S1). When supplementing endemism richness with the criterion of habitat loss, a high overlap can also be noted (Fig. S4). However, the comparison with vertebrate endemism richness sheds critical light on the value of previously recognized plant hotspots for the conservation of animal groups (Figs. 1, 2, and Fig. S3). This is best illustrated by New Caledonia, which exhibited the world's highest plant endemism richness, but has no native extant amphibians. On the other hand, regions like the South American Pacific coastal deserts, the Guinea-Congolia/Sudania Regional Transition Zone, or the South American Chaco, are not outstanding in terms of plant endemism richness, but show consistently much higher values for vertebrates.

By calculating endemism richness for all terrestrial regions, the present study puts the hotspots into a global context and also reaffirms the procedure of prioritizing hotspots by means of a linear conversion of endemic species to a standard area (1). Of course, the resulting figures do not reflect the number of endemics present on the standard area (57, 58), but rather the minimum values of endemism richness (21, 25). A further review of hotspots should include much more detailed, taxon-based data, as performed for Africa (59) or for the Indo-Pacific (60).

A lot of data collecting and processing still needs to be accomplished to improve the accuracy of our knowledge on endemism richness on the global scale, especially for vascular plants. Only for 43 regions (48%), data quality was rated good or very good while being poor or very poor for 47 regions (Table S1). Overall, the data quality for the purpose of this study was better for island than for mainland regions: Ten of the 14 island regions (71%) had good or very good data quality. Partly, this might be attributed to the fact that islands are discrete spatial entities clearly defining the geographic extent of a study. In combination with their comparatively low species numbers, this may make islands more attractive to detailed floristic and faunistic inventorying and taxonomic work than less clearly defined, large mainland regions. It remains a great challenge that in most parts of the world, distribution data are only available for parts of the flora (45). Even the data set on Australian plant distributions, which is one of the best available continental data sets, covering about half of all known Australian plant species, has a pronounced geographic bias because of the absence of data from the Western Australian Herbarium (22). However, our approach of combining such “taxon-based” data sets with inventory data enabled us to partially address this problem by giving higher weight to inventory data whenever biases were known to exist in the taxon-based data and vice versa. Another potential source of error in the data might arise from the tendency to describe taxa as separate species on islands, artificially inflating endemism richness on islands in comparison to the mainland. However, quantitative studies testing this hypothesis are lacking and doubts remain as to whether this effect is important enough to explain the large differences in endemism richness between islands and mainlands.

Whenever priority setting and environmental impact assessments use endemism richness as a basis (e.g., when data are lacking or not of adequate quality for algorithm-based calculations), they should not rely on this criterion alone. Maximal representation and conservation of the global species pool will additionally benefit from taking into account further criteria such as biogeographic characteristics (2). Our study gives equal existence value to each species. However, for many purposes species' importance may vary and be affected by attributes such as their phylogenetic and functional uniqueness (61, 62). Further, (nonglobal) local extinctions, which our study does not quantify, may have serious repercussions for ecosystem functions and services (63, 64). Attention should also be paid to the influence of spatial scale: We here identified regions with high values of endemism richness at a rather broad spatial resolution, but it remains a future challenge to optimize conservation strategies by systematic assessments of conservation priorities within these regions (17). Finally, in an era characterized by climate change and the loss of important biological and socioeconomic functions of ecosystems (65, 66), halting or even reversing these developments should be an equally important target of a global conservation strategy.

>We are grateful to the following persons and institutions for granting us access to partly unpublished plant distribution data: Claudia Raedig, Gary Krupnick, Shawn Laffan, and all institutions which have built up the biogeographical information system data set on African plant distributions (BISAP). Frank La Sorte and Hamish Wilman provided valuable comments on the manuscript. We thank Henning Sommer, Wolfgang Küper, and Daud Rafiqpoor for stimulating discussions and anonymous reviewers for helpful, constructive, and critical comments. This project has been funded by the Academy of Sciences and Literature Mainz (“Biodiversity in Change” Program), the Wilhelm Lauer Foundation, and the German Federal Ministry of Education and Research (BIOLOG BIOTA Program). W. J. was supported by a National Science Foundation Grant BCS-0648733. H.K. was supported by a Feodor-Lynen Fellowship from the Alexander von Humboldt Foundation.