LAUSR

Abstract Climate change and biodiversity loss underscore the need for conservation planning, even in remote areas. Species distribution models (SDMs) can help identify critical habitat for reserve design and selection, and have quickly advanced to the fore of ecological inquiry. Such models are typically dominated by abiotic factors, following the Eltonian Noise Hypothesis (ENH) that physical features set the limits of species distributions. Nevertheless, recent studies challenge this notion and highlight the importance of biotic interactions. Resolving this discrepancy could have significant implications for conservation and ecological understanding. To test these ideas, we built distribution models for two large herbivores, muskoxen (Ovibos muschatus) and Peary caribou (Rangifer tarandus pearyi), systematically observed across a vast spatial extent – 65 islands spanning 800,000 km2 in the Canadian High Arctic. To test the ENH we fit SDMs with two sets of predictors: (1) abiotic only (i.e. topographic, climatic) and (2) abiotic + biotic (i.e. vegetation communities, distance-to-heterospecifics). We evaluated these models and spatially estimated habitat suitability for each species. We found both sets of models had good predictive ability, although biotic variables (i.e. proportion of grass-lichen-moss) improved model performance and substantially narrowed areas of high habitat suitability. Niche overlap between caribou and muskoxen was moderate and highly suitable areas were spatially disjunct between species and largely outside protected areas. These results fail to support the ENH. Our study implies that biotic features, although often overlooked, may be important to the performance of SDMs and vital in identifying priority areas for conservation. For these large herbivores, reflecting trophic interactions in SDMs was essential when estimating areas of conservation value. Our approach helps prepare the way for improved projections regarding the prospects for wildlife while laying the foundation for biologically relevant protected areas.