Spatial ecology of AdevÉvÖlie penguin breeding colonies : the effects of landscape, environmental variability and human activities

Adelie penguins have been widely studied as an \indicator\" species for the health of the Southern Ocean ecosystem. However the effects of climatic variability and human activities on Adelie penguin populations are poorly understood. As many of the Adelie penguin colonies used for long-term demographic studies are located near research stations there is a need to be able to disentangle the effects of human activities and environmental variability on Adelie penguin populations. This study investigates the landscape properties that drive the locations of Adelie penguin colonies in the Windmill Is East Antarctica. It also examines whether potential changes in snow cover and/or proximity to human activities best explain the varying population trends of colonies in two breeding localities. While some colonies have been abandoned or have undergone strong population decreases the populations of others have grown by more than 1000% in the past 38 years. This study uses Geographic Information Systems to generate spatial data of landscape snow accumulation patterns and proximity to human activity parameters. Landscape parameters are derived from fine-scale digital elevation models (DEMs) and snow accumulation patterns are modelled using a complex physically-based GIS model. The parameters are then combined into multivariate statistical models to generate predictions of habitat suitability. Individually the landscape attributes such as elevation slope solar radiation and wetness index have little power to predict the distribution of colonies within a breeding locality. On the other hand multivariate models (discriminant analysis and decision tree) derived from these landscape attributes predict the presence or absence of colonies in test grid cells with up to 78.9% accuracy. General rules to describe the distribution of Adelie penguin colonies are not easily derived as habitat suitability appears to be driven by complex interactions between landscape attributes. At Whitney Pt the study site farthest from Casey modelled snow accumulation parameters explain most of the variation in population trends among colonies (up to 83.7% accuracy for five classes). At Shirley I 500 m from Casey models derived from proximity to human activity parameters correctly predict the trend classes for up to 83.8% of test cells while models derived from snow accumulation parameters correctly classify up to 57.8% of test cells. This suggests that while snow accumulation patterns are a primary driver of variation in population trends among colonies the effect of snow accumulation is outweighed by the effects of human activities near Casey."