Climate variables are not the dominant predictor of Arctic shorebird distributions

Sandpiper (Calidris pusilla), Stilt Sandpiper (Calidris himantopus), Wilson’s Snipe (Gallinago delicata), and White-rumped Sandpiper (Calidris fuscicollis). In our analyses we chose to model occupancy instead of density because shorebirds were observed at very low densities (average of <5 birds/survey for all 17 species, ranging from 0.03–0.68 birds for each individual species), with a frequent counts of 0 (no birds were observed in 54% of surveys). Surveys were observational, and involve no capture or sampling. Lands surveyed include Federal Crown Lands, Inuit Owned Lands, and Federal and Territorial Protected Areas. Access to these lands was granted through permits to JR including: Environment and Climate Change Canada (NWT-MBS-17-03, MM-NR-2022-NU-005, NUN-NWA-17-04, SC-NR-2001-NT-004, SC-NR-2022-NU-005), Government of Nunavut (WL2022-031), Kitikmeot Inuit Association (KTX115X007), Kivalliq Inuit Association (KVX15N05), Qikiqtani Inuit Association (Q14X001). Environmental predictors We used two categories of environmental predictors of shorebird occupancy: predictors found in ‘climate-only models’, and additional predictors in ‘climate+additional models’. For the cli- mate predictors we used the bioclimatic data available from WorldClim 2 at a 30 arc-second resolution (~1km) [31]. This data is typical of that used in climate-only habitat suitability mod- els because it is global in scope and easily accessible. We eliminated any predictors that had correlations higher than 0.7, selecting the 5 uncorrelated climate predictors that seemed most

relevant to shorebird occupancy: annual mean temperature, temperature seasonality (standard deviation of mean monthly temperature x 100), mean diurnal range (mean of monthly max temp–min temp), annual precipitation, and precipitation seasonality (coefficient of variation of monthly precipitation). Additional predictors in the climate+additional models of shorebird occupancy included topography, geology, snowmelt, and land cover variables. The topography predictors we included were elevation, the standard deviation of elevation, and distance to coast. We used elevation data available from WorldClim 2, also at a 30 arc-second resolution (~1km). From this, we derived the standard deviation of elevation over a 5km grid. We calculated distance to coast from coastline data available from Natural Earth [32]. To represent geology, we used sub- strate chemistry data from the Circumpolar Arctic Vegetation map [30], which includes 3 cate- gories: acidic (pH < 5.5), circumneutral (pH 5.5–7.2) and carbonate (pH > 7.2). Snowmelt timing was derived from an 8-day composite MODIS product, and we used the day of 50% snowmelt per 500m pixel [33]. We used land cover data from the Circa-2000 Northern Land Cover Map of Canada. This dataset derives 15 land cover classes from Landsat data 1999–2002 at a 30m resolution [34]. Statistical analysis Before building models of habitat suitability for each species of shorebird, we first determined the spatial extent at which day of snowmelt and each land cover predictor most strongly

affected each species, their "scale of effect" [35]. Habitat selection is scale dependent, therefore identifying the appropriate scale at which species respond is important for making correct inferences about the relative importance of environmental variables [36]. We did not complete this step for climate, topography or substrate because these variables were measured at much larger spatial resolutions, and show little variation at small spatial resolutions. To find the scales of effect, we first filtered the plots to a reduced set that were at least 10km from each other. We calculated the mean day of snowmelt and the proportion of each land cover type at multiple scales, centered on the middle of each plot. We tested spatial scales from a radius of 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 5 / 14