Nesting habitat preferences of snow petrels (Pagodroma nivea) and Wilson's storm petrels (Oceanites oceanicus) in East Antarctica : a modelling approach to predict species distribution

Although snow petrels are ubiquitous around the Antarctic, population estimates of this “not so charismatic” top predator are generally limited. Such information is highly valuable for the monitoring and management of Antarctic and Southern Ocean ecosystems, especially in a climate change context. There is a need to complement long–term temporal demographic information obtained at a limited number of monitoring sites with spatial distribution data.Systematic surveys of snow petrels and Wilson’s storm petrels were undertaken at Casey (2002-2003) and Mawson (2004-2005) in order to provide better regional population estimates and test the performance of predictive distribution models based on topographic and substrate variables for refining such estimates. As habitat selection modelling is rarely used in Antarctic regions, methodological developments focus on dealing with the peculiarities of a semi-colonial hollow-nesting species, testing habitat selection modelling approaches and comparing the output of four types of models (Generalized Linear and Generalized Additive Models, Classification Trees and Ecological Niche Factor Analysis) across a range of scales.Snow petrel nest distribution was directly related to the nature of the rock substrate and to major topographic/geomorphological parameters such as hill slope and the direction of the prevailing winds. Model performance varied with the scale at which models were implemented, suggesting that nest selection processes happen predominantly at the habitat unit and individual nest scales. Further study at the nest scale highlighted that the influence of biotic related parameters such as conspecific attraction (modelled as autocorrelation due to coloniality) may be of lesser influence than selection based on individual nest quality. An alternative modelling method, ENFA, which creates environmental envelopes for the niche of the species with presence data only was identified as valuable for Antarctic data sets, which often lack comprehensive records of species absence.The validation of the models created at Casey with nest data collected in the Mawson region returned satisfactory prediction rates in two different habitat types, coastal islands and inland mountains, suggesting that it may be possible to predict snow petrel distribution across East Antarctica using remotely sensed information on topography and geomorphology, for example high-resolution aerial photography to guide in the design of and complement ground surveys.Similar modelling procedures applied to Wilson’s storm petrels produced more mitigated results and selection for this species appeared to be based principally on nest microhabitat characteristics. However, modelling provided useful information on the large-scale habitat preferences and ecological requirements of both species.