Understanding ground-nesting habitat selection by waterbirds to prioritize invasive predator control on islands

The introduction of carnivores to islands affects the conservation status of native prey. For two decades, American mink (Neogale vison) has expanded its distribution in the sub-Antarctic archipelago of Austral Patagonia, while preying on nests of solitary ground-nesting birds. Here, upland geese (Chloephaga picta) and flightless steamer ducks (Tachyeres pteneres) are naïve to such predation risk because of the lack of native terrestrial predators. We used nest occupancy models to reveal preferred breeding habitats of those two ground-nesting waterbirds to prioritize areas for future mink control programs. We searched for nests along 80 transects on Navarino Island, southernmost Chile, and 11 transects on 10 nearby islets (3 ± 2.9 ha). Our results showed that islets were essential for breeding in comparison to the main island (5.3 nests/km versus 0.8 nests/km, respectively). The occupancy models revealed that islets were particularly important for flightless steamer ducks, endemic to western Patagonia. As breeding habitat, upland geese preferred medium-dense shrubland, and flightless steamer ducks preferred rocky coastlines. Nest detection was negatively affected by dense vegetation and, in the case of the upland geese, towards the end of the survey. Ten camera traps (n = 385 trap nights) revealed mink presence on all islets, highlighting their capacity to swim in cold water, up to 340 m distance from the main island. We conclude that islets represent breeding refuges for ground-nesting waterbirds and therefore advocate allocating resources for mink control during the warm season on islets. We also provide predictive maps of nesting habitat preference of upland geese and flightless steamer ducks for conservation managers to focus their often scarce human and financial resources on areas critical for breeding. Finally, we argue that detection probability should be an integral part of bird breeding studies, which can be methodologically achieved through targeting transects, not nests, as sample units.