DEVELOPING PROBABILISTIC MODELS TO PREDICT AMPHIBIAN SITE OCCUPANCY IN A PATCHY LANDSCAPE

Human‐caused fragmentation of habitats is threatening an increasing number of animal and plant species, making an understanding of the factors influencing patch occupancy ever more important. The overall goal of the current study was to develop probabilistic models of patch occupancy for the mountain yellow‐legged frog ( Rana muscosa ). This once‐common species has declined dramatically, at least in part as a result of habitat fragmentation resulting from the introduction of predatory fish. We first describe a model of frog patch occupancy developed using semiparametric logistic regression that is based on habitat characteristics, fish presence/absence, and a spatial location term (the latter to account for spatial autocorrelation in the data). This model had several limitations including being constrained in its use to only the study area. We therefore developed a more general model that incorporated spatial autocorrelation through the use of an autocovariate term that describes the degree of isolation from neighboring frog populations (autologistic model). After accounting for spatial autocorrelation in patch occupancy, both models indicated that the probability of frog presence was strongly influenced by lake depth, elevation, fish presence/absence, substrate characteristics, and the degree of lake isolation. Based on cross‐validation procedures, both models provided good fits to the data, but the autologistic model was more useful in predicting patch occupancy by frogs. We conclude by describing a possible application of this model in assessing the likelihood of persistence by frog populations.