Demographic stochasticity does not predict persistence of gecko populations

We present a population viability model for an arboreal gecko (Oedura reticulata). This gecko needs a habitat of smooth‐barked Eucalyptus woodlands. In Western Australia its distribution has declined dramatically, largely through clearance of woodlands, but populations persist within woodland remnants. Evidence from extensive field data suggests that the gecko was formerly distributed through much of the original eucalypt woodlands, and that geckos show little movement between patches. The populations in all woodland remnants seem to be isolated. We ask whether the present distribution of the gecko across remnants could have been produced solely by the extinction of populations through demographic stochasticity. To test this possibility, we developed a stochastic, individual‐based model including environmental stochasticity and estimated the percentage of extinct populations of different size from known field characteristics and the time span since the clearing of the woodland. The model predicted a relationship between remnant size and gecko persistence, driven by demographic stochasticity, that is qualitatively similar to the observed pattern. Despite extensive testing, however, we found that the model predicted an incidence function much too optimistic for the observed distribution of populations in small remnants. This discrepancy between field data and our model is due to a series of implicit assumptions. Thus, our modeling exercise sheds light on the procedures commonly applied to population viability analyses of single populations of endangered species. The implicit assumptions involved in such models make many predictions vague. We suggest that for the study of declining species like O. reticulata it is essential to adequately test extinction models and therefore population viability analyses.