Unexpected sources of uncertainty in projecting habitat shifts for Arctic shorebirds under climate change

AimThe rapidly changing Arctic is ideal for investigating uncertainties in climate projections. Despite the challenges of collecting data in this region, an unprecedented large-scale survey of shorebirds has been conducted over the last 30 years. Our study aimed to (1) develop probabilistic estimates for the change in suitable habitat for 10 Arctic shorebird species in Canada by 2075 and (2) assess the contribution of modelling decisions to the uncertainty in these estimates.LocationArctic Canada.MethodsTo evaluate uncertainty, we considered six classes of modelling decisions, yielding 216 unique projections for each species. We tested three decisions that are less commonly explored - the pool of candidate variables, a method for selecting variables, and the maximum distance of tree line dispersal, as well as the modelling algorithm, carbon emissions scenario, and global circulation model. We used a bootstrapping approach, creating a probability distribution for the proportional change in suitable habitat for each species.ResultsOur findings indicated a substantial risk for 8/10 species to lose over half of their suitable breeding habitat, but this projection is much less certain than has been described previously. While much uncertainty is unexplained, we were surprised that the largest source of uncertainty among our modelling decisions was from our choice of methods for variable selection, that the other modelling decisions were relatively small sources of uncertainty, overshadowing other modelling decisions.Main ConclusionsWhile most scenarios predict a northward shift and significant habitat loss for Arctic-breeding shorebirds, the Arctic Archipelago of Canada will remain an important refuge because in many other Arctic regions, there is no land farther north for these species to shift into. A comprehensive understanding of uncertainty is important for deciding if future projections can or should be used when planning climate-resilient protected area networks.