MEPS 299:297-309 (2005)  -  doi:10.3354/meps299297

ABSTRACT: Predictive species–environment models have numerous ecological and management applications, and can provide insight into the functional relationships between a species and its environment. Generalised linear models were used to develop predictive models of crabeater seal Lobodon carcinophaga distribution during the breeding season in relation to physical-oceanographic and ice-cover variables. Models were constructed from data collected on shipboard surveys in the pack-ice off east Antarctica between 50 and 115°E from 1985 to 1987, and were tested from surveys in the same region from 1995 to 1997. Models of presence selected ocean depth as the only significant physical-oceanographic predictor variable for both breeding and non-breeding seals. Seals were most likely to be present on the ice at moderate ocean depths of approximately 2500 m. Unlike non-breeding seals, breeding seals were all but absent on the ice associated with depths of >4000 m. The predicted distribution of presence comprised a band delineated by the shelf-break to the south and extending northwards for 1.5 to 5° latitude, with non-breeding seals having a more northerly limit than breeding seals. This band is coincident with the known distribution of the crabeater seals’ primary food source (Antarctic krill), and also coincides with frontal features such as the Antarctic slope front and the southern boundary of the Antarctic Circumpolar Current, which are considered to be areas of enhanced primary and secondary productivity. The presence–depth models were improved by the inclusion of ice-cover variables. Breeding seals may avoid areas with a high proportion of all ice as floes <20 m in size, because the chance of successfully weaning a pup in such areas is low.

KEY WORDS: Crabeater seal · Distribution · Breeding · Generalised linear modelling · Antarctica