Edge Effects and the Extinction of Populations Inside Protected Areas

See allHide authors and affiliations

Science  26 Jun 1998:
Vol. 280, Issue 5372, pp. 2126-2128
DOI: 10.1126/science.280.5372.2126


Theory predicts that small populations may be driven to extinction by random fluctuations in demography and loss of genetic diversity through drift. However, population size is a poor predictor of extinction in large carnivores inhabiting protected areas. Conflict with people on reserve borders is the major cause of mortality in such populations, so that border areas represent population sinks. The species most likely to disappear from small reserves are those that range widely—and are therefore most exposed to threats on reserve borders—irrespective of population size. Conservation efforts that combat only stochastic processes are therefore unlikely to avert extinction.

The contention that small populations are vulnerable to extinction through stochastic processes has a sound theoretical basis in both demography and population genetics (1). Management of small populations has therefore dominated both the theory and practice of conservation biology for nearly 20 years (2). However, most empirical evidence supporting this contention is indirect, because direct measures of size are rarely available for populations that have subsequently become extinct (3).

If small populations are vulnerable, large carnivores should be especially extinction-prone because their trophic position constrains them to living at low population densities. However, carnivore populations are also exposed to strong external pressures because their requirements conflict with those of local people. Where large carnivores survive outside protected areas, intentional or accidental killing by humans frequently limits their numbers (4). Even within protected areas, conflict with humans is usually the single most important cause of adult mortality (5). Most of this mortality occurs when carnivores range beyond reserve borders (5); such deaths account for proportions of mortality comparable with those known to cause decline in harvested populations of the same species (4, 5). Border areas of reserves may therefore become population sinks. Such sinks will have the greatest impact on overall population dynamics in small reserves with high perimeter:area ratios and in species that range widely and therefore come into frequent contact with reserve borders. In large carnivores, then, both stochastic processes and strong edge effects could contribute to the extinction of isolated populations.

We investigated the relative importance of these two factors by compiling data on population extinctions for 10 species of large carnivores (Table 1). For each species, we chose a geographic region within the species' historic range in which suitable habitat has become fragmented. In all of these regions, people kill large carnivores that range outside the protected areas (5). For each region, we identified protected areas that fell within the former geographic range of the species, treating complexes of contiguous reserves as single protected areas (6). We determined the presence or absence of the species in each of these protected areas, using a combination of published and unpublished data (7). Because none of the species has highly specific habitat requirements, and all have experienced range contractions within the last century, their absence from those protected areas that contain suitable habitat can be taken as evidence of local extinction. We excluded areas where evidence indicated that extinction had occurred before the reserves were designated. We recorded the size of each protected area (reserve size) and the time elapsed between the dates when the area was offically designated and when it was surveyed for carnivores (reserve age).

Table 1

Results of logistic regressions on the presence and absence of large carnivores in protected areas falling within their historic ranges (7, 8). Wherever possible, data on population densities and home range sizes are taken from the regions for which critical reserve sizes were determined (11). Population density refers to the density of adults, averaged across studies; home range size refers to the mean area used by each adult female (or social group for social species).

View this table:

We investigated the relation between reserve size, reserve age, and carnivore extinction, using logistic regression, a standard technique for the analysis of binary data (8). All 10 species were more likely to disappear from small reserves than from larger ones, but extinction was related to reserve age in only one species (Table 1). The statistical effect of reserve size was very strong for all species, but there was considerable variation in the size of the reserves from which each species had disappeared (Fig. 1). We derived a measure of critical reserve size by using the logistic regression models to predict the area at which populations persisted with a probability of 50%. This measure is analogous to the LD50 of a drug, the dose that, administered to experimental subjects, kills exactly half of them. Critical reserve size varied among species by over two orders of magnitude (Table 1).

Figure 1

Proportion of reserves of various sizes in which 10 species of large carnivores have persisted (7). Population persistence is related to reserve area for all species (Table 1). Curves show the probability of persistence predicted by logistic regressions fitted to the binary data (8); filled circles show the critical reserve sizes (±SE) for which the models predict a 50% probability of population persistence. Species: (A) black bear; (B) jaguar; (C) snow leopard; (D) tiger; (E) spotted hyena; (F) lion; (G) dhole; (H) gray wolf; (I) African wild dog; (J) grizzly bear.

If probability of extinction is determined primarily by population size, then critical reserve size should be related to average population density, because the size of a population at isolation will be determined by the population density and the area of the reserve. In contrast, if extinction is caused by edge effects, critical reserve size should be related to home range size, as long as reserve shape varies randomly with reserve area. Population density and home range size will not necessarily be correlated with one another, because carnivores that range widely tend to occupy overlapping home ranges (9).

For each species, we collected data from published reports to estimate average population density and average female home range size within the regions for which we investigated population extinction (Table 1). We avoided statistical nonindependence of measures from closely related species by analyzing phylogenetically independent contrasts, calculated from a composite phylogeny for the Carnivora (10). All contrasts were calculated with log-transformed data, and all regressions of contrasts on contrasts were forced through the origin.

After controlling for phylogeny, average female home range size was a good predictor of critical reserve size (Fig. 2) (r 2 = 0.84,F 1,8 = 42.1, P < 0.0005). The effect of population density was much weaker (r 2= 0.52, F 1,8 = 8.8, P< 0.05), and disappeared entirely after we controlled for home range size (multiple regression: overall, F 2,7 = 20.6,P < 0.005; effect of density, t = 0.82, P > 0.4; effect of home range size,t = 4.00, P = 0.005). As expected, contrasts for population density and female home range size were only weakly intercorrelated (r 8 = −0.69), partly because some species were social and partly because home range overlap was high in species with large home ranges (9, 11).

Figure 2

Relation between phylogenetically independent contrasts in log(critical reserve size) and log(female home range size) calculated for 10 species of large carnivore.r 2 = 0.84, F 1,8= 42.1, P < 0.005. The effect remains strong after controlling for the (nonsignificant) effect of population density (t = 4.00, P = 0.005).

These results show that, in a reserve of given size, wide-ranging carnivores are more likely to become extinct than those with smaller home ranges, irrespective of population density. Thus, population size is a relatively poor predictor of extinction among carnivores. Ranging behavior mediates contact with human activity, contact that accounts for a very high proportion of adult mortality in all of these species. Our results therefore indicate that human-induced mortality contributes more to the extinction of populations of large carnivores isolated in small reserves than do stochastic processes. Conservation measures that aim only to combat stochastic processes are therefore unlikely to avert extinction. Instead, priority should be given to measures that seek to maximize reserve size or to mitigate carnivore persecution on reserve borders and in buffer zones.

  • * To whom correspondence should be addressed. E-mail: rbw20{at}


View Abstract

Navigate This Article