Island Biogeographic Theory and Conservation Practice: Species-Area or Specious-Area Relationships?

We present statistical techniques to evaluate species-area regressions and models o f faunal and floral collapse and apply these techniques to several recent examples from the literature. The application of these models to the design of nature reserves is unwarranted. These models have low explanatory power; they typically explain only half the variation in species number. Their parameter estimates are sensitive to particular cases. Consequently, estimates from these models range over several orders of magnitude following the deletion of a single observation. Species-area and faunal collapse models give unreliable estimates; 95 °/o prediction intervals and inverse prediction intervals routinely span two or more orders of magnitude. These models should be subordinate to autecological considerations in policy formulation. I N T R O D U C T I O N Theoretical ecologists have offered species-area relationships and models of faunal collapse as analytical tools to aid conservation biologists in preserving species diversity. Species-area relationships, modelled as linear regressions, can estimate average species number for a given area, or, conversely, the minimum area sufficient to preserve a given number of species. Estimates from these regressions have mot ivated specific * Present address: Department of Biological Sciences, Northern Arizona University, Flagstaff, Arizona, 8601 l, USA. 63 Biol. Conserv. 0006-3207/84/$03.00 Elsevier Applied Science Publishers Ltd, England, 1984. Printed in Great Britain 64 William J. Boecklen, Nicholas J. Gotelli recommendations concerning optimal refuge size (Diamond, 1975; Diamond & May, 1976). Models of faunal and floral collapse are derived directly from equilibrium theory (MacArthur & Wilson, 1967) and forecast the efficiency of reserves in preserving species number following insularisation (Diamond, 1972; Terborgh, 1974; Soul(~ et al., 1979; Wilcox, 1980). Predictions from these models have been used to advance management policy (Soul6 et al., 1979) and inter-island transfers of endangered species (Temple, 1981). The suitability of species-area regressions and models of faunal collapse for conservation practice will depend upon the quality of the predictions that they generate. These models could be evaluated through implementation in test situations but the possibility of unforeseen, undesirable, and irrevocable consequences negates this approach with real reserves. Alternatively, statistical analysis is possible. A statistical examination can quantify the level of reliability and precision of these models. We believe that these models should be subjected to a critical statistical examination before conservation policies derived from island biogeographic theory become widespread. Here, we present statistical techniques to evaluate species-area and faunal collapse models and apply these techniques to several recent examples from the literature. Species-area relationships and the coefficients of faunal collapse models are typically estimated by linear regression. Therefore, we will evaluate these models according to three criteria that are commonly used to evaluate any regression model: Rsquared adjusted (the percent variation in species number that is explained by area), sensitivity of the parameter estimates to influential cases, and, most important for conservation practice, the precision of the estimates.