Modelling breeding habitat preferences of Bonelli’s eagle ( Hieraaetus fasciatus ) in relation to topography, disturbance, climate and land use at different spatial scales

Predictive models on breeding habitat preferences of Bonelli’s eagle (Hieraaetus fasciatus; Aves: Accipitridae) have been performed at four different spatial scales in Castellon province, East of Iberian Peninsula. The scales considered were: (1) nest site scale (1×1 km2 Universal Transverse Mercator (UTM) square containing the nest); (2) near nest environment (3×3 km2 UTM square); (3) home range scale (5×5 km2 UTM square); and (4) landscape level scale (9×9 km2 UTM square containing the above mentioned ones). Topographic, disturbance, climatic and land use factors were measured on a geographic information system (GIS) at occupied and unoccupied UTM squares. Logistic regression was performed by means of a stepwise addition procedure. We tested whether inclusion of new subset of variables improved the models by increasing the area under the receiver operator characteristic plot. At nest site scale, only topographic factors were considered as the most parsimonious predictors. Probability of species occurrence increases with slope in craggy areas at lower altitudes. At the 3×3 km2 scale, climate and disturbance variables were included. At home range and landscape level scales, models included climate, disturbance, topographic and land use factors. Higher temperatures in January, template ones in July, higher rainfall in June, lower altitudes and higher slope in the sample unit increase probability of occurrence of Bonelli’s eagle at broadest scales. The species seems to prefer disperse forests, scrubland and agricultural areas. From our results, we consider that there is a hierarchical framework on habitat selection procedure. We suggest that it is necessary to analyse what key factors are affecting Bonelli’s eagle nest-site selection at every study area to take steps to ensure appropriate conservation measures. The combination of regression modelling and GIS will become a powerful tool for biodiversity and conservation studies, taking into account that application depends on sampling design and the model assumptions of the statistical methods employed. Finally, predictive models obtained could be used for the efficient monitoring of this scarce species, to predict range expansions or identify suitable locations for reintroductions, and also to design protected areas and to help on wildlife management.