Climate variables are not the dominant predictor of Arctic shorebird distributions

are important predictors of plant distribution, are strongly influenced by underlying geology, irrespective of climate [20]. In this paper, we ask how much do variables in addition to climate help to explain habitat suitability? To do this we model species occupancy using path analyses, which allow us to esti- mate the indirect effects of climate on other predictor variables, such as land cover. We also use partitioning to quantify the total relative importance of climate versus additional predic- tors in explaining species occupancy. We addressed our question using data on the distribu- tion of Arctic-breeding shorebirds. Most papers studying habitat suitability for Arctic- breeding shorebirds have been conducted at smaller extents, and for the most part they do not include climate variables, likely because climate does not vary enough over smaller extents to be a relevant predictor [i.e., 21–23]. The few large-extent models that have been used to predict habitat suitability for these species include a model that largely relies on climate predictors [24], but also a report suggesting that additional predictors beyond climate were important for predicting current and future occupancy of Red Knot (Calidris canutus) and Semipalmated Sandpipers (Calidris pusilla) [25]. Based on previous studies of Arctic shorebird habitat associ- ations, the additional predictors we consider are land cover, snow cover, substrate chemistry, and elevation, and the standard deviation of elevation, as shorebirds prefer flat habitats [21, 24–27].

A priori, we expected that the probability of shorebird occupancy at sites distributed over large spatial extents can be estimated considerably better by explicitly including additional var- iables in habitat suitability models, rather than assuming that these additional variables are themselves driven by climate. If climate is the dominant factor influencing the occupancy of breeding shorebirds across the Arctic, we expect that the variation in shorebird occupancy is mostly explained by the combination of the direct effects of predictors from climate-only mod- els and their indirect effects through their effects on the additional predictors. If, independent of climate, additional predictors have important influences on the occupancy of breeding shorebirds across the Arctic, we expect that a substantial amount of the explained variation in shorebird occupancy is uniquely attributed to the additional predictors not found in climate- only models. Methods Overview To look at the predictors of shorebird occupancy, their interrelationships, and their relative effect sizes, we developed a hypothetical path diagram outlining our hypotheses about how cli- mate predictors and additional environmental predictors are interrelated (Fig 2). We identi- fied predictors that have been important in other models of habitat suitability for Arctic- breeding shorebirds [21, 24–27]. We used path analysis to verify the structure of our path dia- gram and assess the degree to which the effects of additional predictors are driven by climate.

We then used a deviance partitioning approach to test whether the total (direct + indirect) effects of climate predictors explained most of the variation in shorebird occupancy. All analy- ses were performed using R Statistical Software v4.2.1 [28]. Shorebird surveys Plots were surveyed for shorebirds across the Canadian Arctic as part of the Arctic Program for Regional and International Shorebird Monitoring (PRISM) [29], which covers all of Arctic North America as defined by the Circumpolar Arctic Vegetation Map [30]. The Canadian PRISM data used here includes 2336 plots across the Canadian Arctic (Fig 3), an area of 3.5 million km2. Given this massive area, surveys were completed over a number of years, from PLOS ONE Climate variables are not the dominant predictor of Arctic shorebird distributions PLOS ONE | https://doi.org/10.1371/journal.pone.0285115 May 17, 2023 3 / 14