The interplay between animal location accuracy and the decorrelation length scale of environmental variables when investigating environmental selection in marine organisms

Abstract

Many large pelagic organisms appear to select specific oceanic conditions, probably due to physiological, energetic, reproductive, or other life history needs. However, prior to characterizing these dynamics and determining their underlying drivers, the selection itself must be reliably identified. Our ability to do so depends on the quality of animal locations and on the heterogeneity of the environmental conditions driving selection. To draw meaningful conclusions about environmental selection of large organisms and therefore about their basic ecology, distribution, and ultimately their potential exploitation and conservation, limits of selection detectability must be established. Here, we investigated how animal location accuracy and environmental variable decorrelation length scales impact the ability to detect environmental selection by marine organisms. We created synthetic tracks cuing on environmental variables potentially relating to life history demands, like sea surface temperature, chlorophyll a concentration, and Lagrangian coherent structures. By artificially imposing different animal selection strengths and location accuracies, we assessed how well environmental selection can be detected statistically in different cases. We found medium or strong selection to be reliably detected, even with relatively small samples and large position uncertainty, while weak selection presented similarly to no selection, especially for large sample sizes and position uncertainties. We therefore recommend using a selection strength threshold, which can significantly reduce the number of false positives while only increasing the risk of false negatives in cases of very weak selection, which are also less meaningful ecologically. We provide criteria to use when assessing confidence in environmental selection results for real marine organisms.