Nonrandom Extinction and the Loss of Evolutionary History

See allHide authors and affiliations

Science  14 Apr 2000:
Vol. 288, Issue 5464, pp. 328-330
DOI: 10.1126/science.288.5464.328


The hierarchical nature of phylogenies means that random extinction of species affects a smaller fraction of higher taxa, and so the total amount of evolutionary history lost may be comparatively slight. However, current extinction risk is not phylogenetically random. We show the potentially severe implications of the clumped nature of threat for the loss of biodiversity. An additional 120 avian and mammalian genera are at risk compared with the number predicted under random extinction. We estimate that the prospective extra loss of mammalian evolutionary history alone would be equivalent to losing a monotypic phylum.

Current and projected species extinction rates exceed geologically normal background rates by several orders of magnitude (1–3), indicating that we face an extinction episode equivalent to mass extinctions of the paleontological past. When biodiversity is measured by evolutionary history, expressed as the total length of all the branches in the tree of life, a surprisingly high proportion is likely to survive even a massive extinction episode. This is because most species have close relatives and thus contribute little to the total branch length: Whole clades are lost only when all their species go extinct, which is unlikely under an assumption of phylogenetically random extinction. However, historical extinctions and current extinction risk are often not randomly distributed among species. For example, the 85 mammalian species extinctions since 1600 include at least five members of the extinct family Nesophontidae (4, 5), and the prevalence of current threat varies significantly among orders of mammals (6) and birds (7). This nonrandomness will result in the loss of more branch length and more higher taxa than predicted by random extinction (8). Here, we quantify how the clumping of extinction risk affects the amount of evolutionary history under threat in mammals and birds, using two measures of biodiversity: the number of higher taxa (genera) and the total phylogenetic branch length [commonly referred to as “phylogenetic diversity” (PD)] (9).

Nee and May (10) showed that surprisingly little PD is lost under even catastrophic extinction scenarios. In one of their simulations, 81% of the phylogenetic branch length remained even when only 5% of the species survived an extinction episode. Their simulations assumed that extinction was random—the “field of bullets” scenario—or could be optimized through management (so as to minimize loss of branch length) and indicated that the amount preserved would be influenced by the topology of the phylogenetic tree.

In principle, we can envisage two natural scenarios that would result in nonrandom distribution of extinction risk. First, any phylogenetic clumping of factors that promote risk would increase the chance of all species in polytypic taxa—and hence those taxa as a whole—being lost. Second, if such phylogenetically distributed traits have already mediated considerable extinction, then many monotypic genera or families might be the last survivors of once-larger clades. This could lead to a higher proportion than expected of monotypic genera, or species on long phylogenetic branches, being threatened. Nonrandom extinction risk has been documented in many groups (6–8, 11), but its impact on biodiversity loss has not hitherto been assessed.

We estimated the loss of biodiversity expected according to current assessments of species extinction risk and compared it with the loss that would result from a random extinction episode of equal severity. We used assessments of extinction risk from the 1996 World Conservation Union (IUCN) Red List (4), which is comprehensive in its coverage of mammals and birds. Three levels of extinction risk were selected for the analysis: endangered and higher (EN), vulnerable and higher (VU), and near threatened and higher (nt) (12). For each of these levels in turn, we imposed extinction of all species of at least that level, with all species at lower levels surviving (13). Species for which no threat classification can be made because of a lack of information are classified as data deficient (DD) by IUCN (14). We dealt with DD species in two ways. First, we assumed that they had no risk of extinction: Here a genus containing a DD species is never lost, a treatment that is therefore conservative. Second, we classified DD species as EN, so that they are all lost in all extinction regimes. This errs in the opposite direction but is probably nearer the truth: It is likely that disproportionately many DD species are at high risk of extinction (4).

For each threshold and each clade, we calculated the numbers of species, the numbers of genera (overall, monotypic, and polytypic), and (for primates and carnivores, the only two clades whose complete species-level phylogenies are available) the total phylogenetic branch length that stand to be lost (15). For comparison, we conducted simulations (1000 trials) in which the same numbers of species were removed at random (16).

Figure 1 shows the results for mammals and birds; Table 1 shows the taxonomic and phylogenetic results for Primates and Carnivora. The same general trends are apparent whether DD species are treated as highly threatened or secure. Three of the four data sets show far more genera to be at risk than would be predicted by the random extinction model. The fourth data set (Carnivora) shows a weaker tendency in the same direction (17). Within each data set, the difference between observed and expected loss (the “extra” loss) tends to increase with the proportion of species culled, at least until a large proportion (around 50%) of the species are lost. For mammals, birds, and primates, loss of all threatened (threshold at VU) species would lead to the loss of about 50% more genera than expected under the null model. About half of these additional genera are monotypic, indicating that members of monotypic genera tend to be more threatened than average species. The pattern is in fact more general: Across both mammals and birds, the probability of a species being threatened declines with the number of species in its genus, family, or order (18). Like clustering of threatened species within clades, this distribution will tend to counteract the ability of hierarchically structured phylogenies to retain diversity in the face of impending extinctions.

Figure 1

Numbers of monotypic and polytypic genera lost under different extinction regimes for (A) mammals and (B) birds. Dark bars: extinction of all species listed at or above the indicated threshold level of threat [see note (12); numbers in parentheses are percentages of species culled]. Light bars: random extinction of same intensity (mean of 1000 trials). Error bars: 2 standard deviations of the simulation distribution. ***P ≤ 0.001 (P values obtained directly from distribution of simulation results). For overall genus loss (monotypic + polytypic), all P ≤ 0.001. DD species were treated as being at no risk of extinction; treating them as EN led to qualitatively very similar results (33).

Table 1

Numbers of genera and phylogenetic diversity (PD) lost under different extinction regimes in primates and carnivores, for two treatments of data-deficient (DD) species. Level, threshold threat level [see (12)]; %spp, percentage of species culled; Obs., loss incurred with extinction of all species listed at or above threshold level; Mean, mean loss from 1000 random extinctions of same severity; #SDs, difference between Obs. and Mean, expressed in standard deviations of simulation results.

View this table:

There are also many mammal and bird genera—far more than expected under the null model (Fig. 1)—in which all two to six species are threatened (19). The extinction of all threatened species would lead to the loss of whole genera of unusual and highly valued groups, such as chimpanzees, golden-lion tamarins, chinchillas, manatees, and kiwis. The loss is not limited to the genus level: Several species-poor families (either monotypic, such as the aye-aye and kagu, or polytypic, such as rhinos and kiwis) and even orders (Microbiotheriidae, Proboscidea, Sirenia, and Apterygiformes) would also be lost, along with their unique biological characters. Although it is true that smaller proportions are lost of PD and genera than of species—an almost inevitable consequence of the hierarchical nature of phylogenies—the extra loss of biodiversity (relative to random extinction) is considerable. Loss of all threatened species of mammals and birds would lead to the loss of at least 85 and 38 extra genera, respectively (from the conservative simulations). Only three mammalian orders (Chiroptera, Carnivora, and Rodentia) have more than 85 genera, and there are only around 1150 mammalian and 2100 avian genera altogether.

The results for primates make it possible to estimate very roughly the extra PD that stands to be lost in mammals as a whole (20). The three thresholds of extinction risk and two treatments of DD species give six estimates of the extra PD lost per genus, averaging around 10 million years (My) per genus. Mammals as a whole stand to lose 85 extra genera, corresponding to an estimated 850 My of extra PD. The added loss of PD incurred through nonrandom extinction in mammals alone would therefore roughly equate to the loss of a monotypic phylum.

  • * To whom correspondence should be addressed. E-mail: a.purvis{at}


View Abstract

Stay Connected to Science

Navigate This Article