Report

Avian Extinction and Mammalian Introductions on Oceanic Islands

See allHide authors and affiliations

Science  24 Sep 2004:
Vol. 305, Issue 5692, pp. 1955-1958
DOI: 10.1126/science.1101617

Abstract

The arrival of humans on oceanic islands has precipitated a wave of extinctions among the islands' native birds. Nevertheless, the magnitude of this extinction event varies markedly between avifaunas. We show that the probability that a bird species has been extirpated from each of 220 oceanic islands is positively correlated with the number of exotic predatory mammal species established on those islands after European colonization and that the effect of these predators is greater on island endemic species. In contrast, the proportions of currently threatened species are independent of the numbers of exotic mammalian predator species, suggesting that the principal threat to island birds has changed through time as species susceptible to exotic predators have been driven extinct.

The colonization of each land mass by humans has broadly coincided with an increase in the extinction rate of the native biota (16). The oceanic island bird species lost to extinction after human colonization are estimated to number in the hundreds to thousands (3, 710). Although the exact causes of these extinctions are debated (1115), human colonization is typically associated with habitat destruction and fragmentation and with other processes that can eliminate species, including overexploitation of populations (5, 16). These processes may also be driven by the nonnative predator and herbivore species that humans introduce. In particular, the introduction of mammalian predators has caused the extinction of many populations and species of oceanic island birds (3, 1720) that, having evolved in their absence, lack appropriate anti-predator responses.

The probability that a bird species has been extirpated from an island varies substantially among islands (Fig. 1A). This variation has normally been ascribed to differences in the characteristics of individual islands, such as area, isolation, elevational range, or date of human colonization (9, 2124). Isolation may increase extinction susceptibility, because the avifaunas of isolated islands are likely to have evolved for longer in the absence of predators, and so their species are more likely to react naïvely to exotic mammalian predators when they arrive. Large islands support larger populations of native bird species, which are thus less prone to extinction (25). Larger islands (or those with a greater elevational range) are also more likely to provide refugia from forces that promote extinction. Islands colonized longer ago may have more unrecorded prehistoric extinctions and so appear to have fewer extinctions, because only species that are relatively resistant to extinction remain [the filter effect (21)]. However, islands also vary substantially in the number of exotic mammal species, from none to >20 species. This variation may also drive variation in extinction probability. Here, we test this hypothesis using data from 220 oceanic islands worldwide (26) (table S1). Most mammalian introductions have occurred in the period after European colonization of an island, so we focus on bird extinctions that have occurred in this historic period. This increases the reliability of our results because historic extinctions are well documented, whereas prehistoric extinctions suffer from the incompleteness of the (sub)fossil record.

Fig. 1.

Interisland variability in the probability of historic extinction and current threat for bird species. (A) The frequency distribution across islands (n = 220 islands) of the probability that a species in the historic fauna has become extinct from an island (the proportion extinct). Here and throughout the paper, extinction refers to the loss of a species from an island and so does not necessarily equate to global extinction. (B) The frequency distribution across islands (n = 220) of the probability that a species in the extant fauna is threatened with extinction (the proportion threatened). Threat refers to the risk of global extinction (30).

Our results show that islands with more exotic mammal predator species have lost a greater proportion of their avifauna since European colonization. The probability of extinction from an island's avifauna since European colonization increases with the number of exotic mammal predator species introduced (estimate ± SE = 0.24 ± 0.05, P < 0.01; the estimate is a linear parameter estimate from a general linear mixed model with binomial errors). The numbers of exotic mammalian predators and herbivores on an island are highly positively correlated (Pearson's r = 0.67, n = 220 islands, P < 0.01); however, the probability of extinction is unrelated to changes in the number of exotic mammal herbivore species (estimate ± SE = 0.02 ± 0.07, P = 0.51). This suggests that the predator relationship is not a simple consequence of more extensive environmental modification that leads to both more native extinction and exotic establishment. Introduced mammal predators are known to have caused specific island bird extinctions, but our results are consistent with introduced predators being a major cause of bird extinctions on oceanic islands around the world. Islands with more exotic predator species have suffered correspondingly greater losses.

We restrict analysis to historic extinctions, yet some exotic mammal species were established on islands before European colonization (e.g., the Pacific rat Rattus exulans), and many bird extinctions were prehistoric. A significant effect of predator species number is still observed when the analysis is expanded to include prehistoric mammal introductions and bird extinctions (multiple regression, predators: estimate ± SE = 0.24 ± 0.06, P < 0.01; herbivores: estimate ± SE = 0.01 ± 0.07, P = 0.81).

The apparent influence of predators could be an artifact of colinearity between the number of predator species introduced and biogeographic variables that have previously been shown to relate to extinction probability across islands, such as area or isolation (9, 2124). The number of predator species introduced to an island does indeed covary positively with island size (r = 0.63, n = 197 islands, P < 0.01) and maximum elevation (r = 0.42, n = 183, P < 0.01). However, because extinction probability should be lower on large and elevationally diverse islands, negative relationships would be expected if this colinearity were causing the predator effect. Moreover, multiple regression analyses that include island characteristics and number of exotic mammal predator species reveal that the effect of predators is robust to the inclusion of other variables that might determine avian extinction probability and that predator species number is the strongest predictor (Table 1).

Table 1.

The minimum adequate multivariate model (MAM) for historical probability of extinction in native island bird species. The MAM was derived by backward deletion of nonsignificant terms from a full model that also included island isolation, latitude, and human colonization date. Adding the number of exotic bird species to the full model does not alter the final MAM. The estimate is a linear parameter estimate.

Predictor Estimate SE t value
Intercept -4.25 0.67
Number of introduced predatory mammal species 0.41 0.07 5.75View inline
Log (area) -0.70 0.16 -4.40View inline
Log (maximum elevation) 0.69 0.27 2.48View inline
  • ** P < 0.01.

  • View inline* P < 0.05.

  • View inline*** P < 0.001.

  • Previous studies have found that islands with more exotic bird species have lost more native bird species (27). This relationship is argued to be an indirect consequence of extensive environmental modification that increases the habitat available for exotic bird species and negatively affects native species. Numbers of exotic bird species and native bird extinctions are also positively correlated across islands (r = 0.62, n = 220 islands, P < 0.01), as are numbers of both exotic bird and mammal species (r = 0.65, n = 220 islands, P < 0.01). However, the proportion of bird species extirpated from an island is independent of the number of exotic bird species in a multiple regression that includes the number of exotic mammalian predator species (Table 1).

    Islands with more exotic predatory mammals suffered higher probabilities of bird extinction presumably because more diverse predator assemblages target a wider range of prey species or increase the overall predation rate on each species. However, alternative possibilities are that extinction has been driven by the range of predator types or by the presence of one or two particularly damaging predators, such as rats or cats (17, 19, 20, 28), with the probability of these species being present on an island increasing with predator assemblage size [the sampling effect (29)]. We assessed these possibilities by comparing the fit to the data of different models in which the assemblage of introduced mammals was characterized in different ways (Table 2). Characterizing an assemblage by the number of predator species clearly provided the best fit to the data (the probability that this model was the best fitting out of the candidate set is 0.77), better than models in which the assemblage is characterized by the number of all mammal species or by the specific presence or absence of rats and/or cats. It was also a better fit than the number of taxonomic orders of predators introduced, which we used as a metric of number of predator types. These results strongly imply that each successive addition of an exotic predator acts to eliminate an additional proportion of an island's avifauna.

    Table 2.

    Comparison of models for the probability of historical bird extinction on oceanic islands (response variable) with the assemblage of introduced mammals characterized in different ways. Predictor variables in each model are log10 area, log10 elevation, and the variable in the first column; thus, the first model in this table is identical to that in Table 1, and subsequent models replace the number of predators in that model with the variable in the first column. Akaike's Information Criterion (AIC) values were calculated with PROC NLMIXED in SAS (32). ΔAIC is the difference between AIC values for each model and that with the lowest AIC (in this case, number of predators). W is the model's Akaike weight, which is the relative probability that the model is the best fit to the data of those tested (33).

    Predictor AIC ΔAIC W
    Number of exotic predator species 335.9 0.0 0.77
    Number of exotic herbivore species 341.0 5.1 0.06
    Number of exotic mammal species 339.7 3.8 0.12
    Number of exotic mammal orders 341.9 6.0 0.04
    Rats, presence or absence 348.9 13.0 <0.01
    Cats, presence or absence 345.3 9.4 0.01
    Rats and cats, presence or absence 347.4 11.5 <0.01

    The impacts of mammalian predators ought to be greater on bird species endemic to islands, because these are more likely to have evolved in the absence of predators. For a given number of exotic predators, the probability of extinction is higher for island endemics than for species with continental populations (endemism: estimate ± SE = 1.55 ± 0.28, P < 0.01). Moreover, although higher extinction probability for species with continental populations is associated with decreasing island area (estimate ± SE = –0.84 ± 0.22, P < 0.01), increasing island elevation (estimate ± SE = 0.79 ± 0.32, P < 0.01), and the number of predatory mammal species (estimate ± SE = 0.35 ± 0.07, P < 0.01), extinction probability for island endemics is solely associated with the number of predatory mammal species (estimate ± SE = 0.16 ± 0.05, P < 0.01).

    The probability that a species in the extant native avifauna is threatened with extinction also varies substantially across islands in our data (Fig. 1B). Is this variation also explained by the numbers of exotic mammal predator species? Certainly, the probability of threat is positively correlated with the probability that a species has already been extirpated from an island (r = 0.43, n = 220 islands, P < 0.01), and the proportion threatened is positively related to the total number of exotic mammal species (estimate ± SE = 0.04 ± 0.01, P < 0.01). However, in contrast to extinctions, there is no significant relationship between the probability of current threat and the number of exotic predators, although there is for the number of exotic herbivores (herbivores: estimate ± SE = 0.08 ± 0.03, P < 0.05; predators: estimate ± SE = 0.04 ± 0.05, P = 0.36). Moreover, in multiple regression, the probability of threat is significantly related only to biogeographic features of the islands (Table 3), with variation in the probability of threat across islands independent of the number of exotic mammal predator species.

    Table 3.

    The MAM for probability of threat in extant island bird species. The MAM was derived by backward deletion of nonsignificant terms from a full model that also included the number of exotic predator species, the number of exotic herbivore species, and the number of exotic bird species. The estimate is a linear parameter estimate.

    Predictor Estimate SE t value
    Intercept -4.83 1.32
    Log (area) 0.19 0.09 2.21View inline
    Log (maximum elevation) 0.39 0.17 2.24View inline
    Log (years since colonization) -0.76 0.24 -3.15View inline
    Log (isolation) 1.06 0.31 3.38View inline
  • ** P < 0.01.

  • View inline* P < 0.05.

  • View inline*** P < 0.001.

  • The introduction of mammalian predators has been a major cause of bird extinctions on oceanic islands worldwide. Each successive predator introduction increases the number of species lost (Table 1), and island endemic species have suffered the most. In contrast, the proportion of an island's avifauna currently threatened with extinction is unrelated to the current number of exotic mammal species, even though many individual bird species are at risk from such predators (30), and islands with many extinctions in the past are also those with high levels of current threat. However, the current threat to native birds is higher on larger, more elevated, isolated islands that have been colonized more recently, consistent with the well-known filter effect (21, 31). Presumably, most species susceptible to the current assemblages of exotic predators have already been driven extinct, especially on smaller, more easily colonized islands. However, these results do not imply that there is no longer any danger to island birds from exotic predatory mammals, because they relate to the current communities of exotic mammals on these islands. The establishment of each additional predator species is predicted to lead to additional extinctions among the native birds. This implies that exotic predators remain one of the major threats to island avifaunas, given that most islands currently have few predators (and some have none) and are likely to suffer progressively more extinctions if and when additional predators colonize. The likely consequences of future introductions are clearly presaged by the losses of the past.

    Supporting Online Material

    www.sciencemag.org/cgi/content/full/305/5692/1955/DC1

    Materials and Methods

    Table S1

    References and Notes

    References and Notes

    View Abstract

    Navigate This Article