Scale relativity of species-habitat models

Abstract Scale has been identified as a central unifying concept in ecology, yet few empirical studies examine its importance per se when quantifying species-habitat relationships. Means and associated variances of ecological variables are known to aggregate unpredictably among observational scale sizes; however, the empirical implications of these scaling effects remain to be fully examined when building and interpreting species-environment models. Here we explore scale-based modeling implications using radio-telemetry data of adult and dispersing-juvenile northern flying squirrels linked to LiDAR-derived fine-grain forest structure data. We construct and rank the same set of candidate species-habitat models across a continuum of 14 biologically relevant observational scales to examine whether scale alone can affect species-habitat model selection. Our results demonstrate differential relative model support (via AIC weights) whereby upwards of seven different “best models” can be generated with varying levels of support entirely contingent upon the scale within which they were quantified. Further, we show this effect to be different for male, female, and dispersing-juvenile animals. We conclude that a continuum-based approach and an understanding of model relativity among scales is a fundamental but absent step in the building and interpretation of most multi-scale ecological models.