Total foraging area of the Great Barrier Reef breeding seabird community

The total foraging area of the Great Barrier Reef breeding seabird community was estimated at 2,941,000 km2 by global foraging circles, of which 197,000 km2 was predicted unsuitable foraging habitat and excluded following transferability-supported refinement (Table 5). There were large differences in the total area of unsuitable habitat excluded from foraging circles between species due to radius size (Table 3), proximity of GBR colonies (overlap of foraging circles; Fig. 5a) and model transferability (Fig. 5b). We saw the greatest refinement in species with large foraging radii (e.g. wedge-tailed shearwater long trips, frigatebirds, tropicbirds) and species which had better model transferability (e.g. terns, Fig. 5b). The better local prediction of GBR colony-specific models over multi-colony models afforded locally tracked species greater transferability-supported refinement (Fig. 5). This led to greater total unsuitable habitat exclusion when locally tracked sites were well represented and transferability was higher (e.g. wedge-tailed shearwater short trips compared to brown booby and masked booby; Fig 5; Table 5). By contrast, noddies had the highest number of breeding colonies but unsuitable foraging habitat could only be excluded at one site with local tracking because their multi-colony model transferability was too poor (AUC<0.5) for foraging circle refinement. Sooty terns were the only species for which we were unable to use transferability-supported refinement for any foraging circles due to their poor multi-colony model transferability and absence of GBR tracking data. Refined foraging circles on the GBR produced through transferability-refined refinement had high refinement confidence, with a 96% average probability of including known foraging areas (Table 5). This was higher than the 89% average from our global validation exercise (Fig. 4) due to the contribution of locally tracked GBR colonies, which all had 100% inclusion of known foraging areas in refined foraging circles.
After transferability-supported refinement had been completed, six modelled species still contained some colonies with refined foraging circles that exceeded 100,000 km2 (Fig. 5b). Using area-limited refinement to drop these refined foraging circles below 100,000 km2 excluded a further 1,629,000 km2 of predicted unsuitable foraging habitat from the community (Table 5). However, this came at a cost of reducing refinement confidence, with the average inclusion of known foraging areas in refined foraging circles dropping from 97% to 55% in these species. Sooty tern, tropicbirds and frigatebirds all required large area-limited refinement to meet the 100,000 km2 limit, greatly reducing their refinement confidence (16-40% inclusion; Table 5). Wedge-tailed shearwater long trips required the greatest refinement to meet the area limit and highlighted the power of tracking data from the GBR; refined foraging circles delineated by the 90thpercentile of habitat suitability dropped predicted inclusion of known foraging areas to 6% in multi-colony models but only to 87% in colony-specific models (Table 5). Area-limited refinement of wedge-tailed shearwater short trips and red-footed booby foraging circles to meet the 100,000 km2 limit was achieved with minor reduction in known foraging area inclusion (77-95%). Applying the most appropriate foraging circle refinement approach to each modelled species (none for sooty tern, area-limited for wedge-tailed shearwater and red-footed booby, and transferability-supported for the remainder) created a network of refined foraging circles for the breeding seabird community of the GBR that balanced size against refinement confidence (Fig. 6).