Modelling nesting habitat preferences of Eurasian Griffon Vulture Gyps fulvus in eastern Iberian Peninsula
Published: Volume 52(2), December 2005. Pages 287-304.
Keywords: Akaike´s information criterion (AIC), cliff-nesting raptor, generalized linear models (GLM), Geographic Information System (GIS), habitat preferences, habitat selection, information-theoretic approaches and ROC plots
Aims: To apply modern habitat modelling techniques using generalized linear models approach to generate not only explicative but also predictive habitat suitability models. Ecological factors that could affect the species’ nesting habitat preferences at two spatial scales, one referred to the colony as a sampling unit and another trying to reflect the landscape level, had been underlined.
Location: Castellón province, eastern Iberian Peninsula.
Methods: Occupied and unoccupied cliffs and 10x10Km U.T.M. squares were compared using univariate tests. A generalized linear model, employing logistic regression, was applied to explain and predict the occurrence of Griffon Vultures at two spatial scales. In order to select the most parsimonious model amongst a set of logistic models built with different explanatory variables, the Akaike´s information criterion (AIC) was performed. With the aim of testing the predictive performance of the previously selected models, the threshold-independent Receiver Operating Characteristic (ROC) plot was computed.
Results: 19 of 29 variables analyzed presented significant differences between occupied and unoccupied cliffs and between occupied and unoccupied squares. Colony parameters, climate, disturbance and vegetation variables, showed a good model performance in accordance with ROC plots and AUC. Despite geomorphological variables being significant on logistic model, model assessment techniques applied in this study suggested that they could not discriminate Griffon Vulture probability of occurrence by themselves. Trophic variables were not selected as good predictors by any method.
Conclusions: Model selection is useful for making inferences from field data. When coupled with Geographic Information System (GIS), habitat models generated by logistic regression approach can aid to develop occurrence maps displaying suitable habitat for nesting. This new applied technique has a broad potential application in conservation biology and management. Future research will involve the development of GIS-based predictive models to a finer scale of description and management application.