Scale-dependent correlations between the abundance of Brunnich's guillemots and their prey

1. The foraging ecology of Brunnich's guillemots Uria lomvia was studied du ring the breeding season in south-eastern Svalbard. In the region of Storfj orden there are two large breeding colonies comprising a total of about 540 000 individuals. These birds forage in the western part of Storfjorden and further to the south. Their main prey are polar cod Boreogadus saida, pela gic amphipods Parathemisto spp. and euphausiids Thysanoessa inermis. 2. A ship-based transect survey was used to record bird abundance and the a coustically determined biomass of presumed prey. The five transects were di vided into 33 segments, each 8-11 km in length. The resolution of the surve y was 150 m, and analyses of correlations between predators and prey were p erformed at length scales from 150 m to 9 km. We differentiated acoustic si gnals into aggregated and dispersed categories according to the estimated h orizontal distribution of presumed prey. 3. Foraging guillemots were consistently more strongly correlated with the aggregated prey than with dispersed prey over scales ranging from 150 m to 9 km. Correlations were weak at small scales (150 m - 1 km) and increased a nd stabilized at scales of 23 km. The spatial scale at which we obtained a shift from weak to strong correlations between guillemots and their prey wa s similar to the scale at which the spatial variances in both guillemot and prey abundance were high. 4. Guillemots showed low correlations with prey at low prey densities. Simi larly, correlations between guillemots and prey were low at low bird densit ies. The data support the hypothesis that the birds associate with prey pat ches with densities above a certain threshold, and that 'regional' prey abu ndances affect local use of patches. 5. The numerical aggregative response curves between guillemot and prey den sity were classified as being neither hyperbolic (type II) or sigmoidal (ty pe III) within the range of prey densities observed in this study. The aggr egative response curves were sensitive to spatial scales, which suggest tha t studies of response curves should be conducted at a range of spatial scal es.