Effective global conservation strategies

Abstract

Arising from: R. Grenyer et al. Nature 444, 93–96 (2006)10.1038/nature05237; Grenyer et al. reply

Using data on the global distribution of mammal¹, bird² and amphibian³ species, Grenyer et al.⁴ conclude that planning based on individual taxa does not provide efficient solutions for the conservation of other taxa. They also report that the performance of existing global conservation strategies—endemic bird areas⁵, biodiversity hotspots⁶ and global 200 ecoregions⁷—in representing those taxa is often no better (and in some cases worse) than random. I argue here that the methodology used by Grenyer et al. was not appropriate for purported globally comprehensive analyses. Focusing on analyses of rare species as an example, I demonstrate how the data actually reveal substantial cross-taxon surrogacy and good performance of existing global conservation strategies.

Main

Grenyer et al. describe the scope of their findings as ‘global’, but their analyses were only performed on subsets of their data. For example, when investigating the congruence between rare birds and rare mammals, they did not obtain a global correlation (across ∼19,500 land-grid cells) but a correlation across cells holding rare species of birds or mammals (∼5,800 cells). Hence, they excluded ∼13,700 cells in which there was perfect agreement between rare birds and rare mammals, in that both were absent. Consequently, the correlation coefficient reported (r = 0.48) underestimates the true global correlation (r = 0.58). The same applies to all other r values reported, including those in their Fig. 2a–c (ref. 4).

The analyses of species numbers expected by chance in areas of variable size (random curves in their Fig. 4; ref. 4) suffer from the same methodological bias: for instance, the curve in their Fig. 4b is not for a true random selection, but for a random selection across just the ∼6,000 cells containing rare species. The true global random curve is much shallower (Fig. 1). Globally, endemic bird areas, hotspots and global 200 ecoregions perform substantially better in representing rare vertebrate species than would be expected by chance (Fig. 1). Given that hotspots were selected on the basis of plant endemism, this provides evidence for surrogacy of restricted-range plants in representing rare vertebrates.

Figure 1: Relative performance of different types of priority network in capturing rare species (see Fig. 4b of Grenyer et al.⁴).

Minimum complementary set representing each species at least once of rare mammals (M), rare birds (B), rare amphibians (A), and rare mammals, rare birds and rare amphibians combined (C). Position is also shown for biodiversity hotspots (H), endemic bird areas (E) and global 200 ecoregions (G). Performance is evaluated by comparing the position of these points for equivalent areas with the maximum number of rare species that can be represented (red line) and the number of rare species expected by chance (blue: 95% confidence range for randomly selected sets of cells; 100 replicates). The dashed line is the random line from Grenyer et al.⁴, which corresponds to a random selection across just the sites holding rare species, plotted on the dashed grey x axis, top. In Fig. 4 of Grenyer et al.⁴, the position of the random lines therefore cannot be compared with the positions of points E, H and G as they are plotted on different axes (the positions of points M, B, A and C are the same on both axes).

Grenyer et al. investigated cross-taxon surrogacy by counting how many target species are represented in minimum complementary sets selected for a particular surrogate taxon (their Table 1; ref. 4). For rare species, they found values ranging from 22.5% to 77.9% and concluded that surrogacy is low. However, these values alone are not informative: they need to be compared with what would be expected by chance, and what the maximum possible representation is, in an area of the same size⁸. Figure 1 provides this information for when the target is the representation of rare species across the three groups. It shows that minimum sets representing rare mammals, rare birds or rare amphibians individually represent substantially more overall rare species than would be expected by chance. Furthermore, these minimum sets are noticeably close to the maximum representation possible, which is indicative of a high degree of surrogacy.

Analyses of rare species are the most disrupted by the methodological problems described here, but the other two groups analysed by Grenyer et al. (all species and threatened species) are also affected.

In conclusion, the analyses in Grenyer et al. suffer from a systematic methodological bias that does not allow the results to be compared with the maximal possible representation. The prospects for global conservation planning are, in fact, positive, not dismal as portrayed⁴. It is true that better results will be obtained when high-resolution data become available for all taxa we aim to conserve. Nonetheless, at least for the terrestrial realm, good progress can be, and has already been, achieved by conservation planning based on existing data.

Methods

The following databases were used: ADHoC database of geographic ranges of birds², owned and developed by the NERC Avian Diversity Hotspots Consortium; global mammal database^1,4, owned and developed by J. Gittleman; Global Amphibian Assessment^3,9, developed by SSC-IUCN, CABS-CI and NatureServe. Rare species are those in the lower quartile of the range distribution of each taxonomic group⁴. Optimizations were achieved with the GNU Linear Programming Kit package.