Using data from related species to overcome spatial sampling bias and associated limitations in ecological niche modelling

Abstract

Abstract Ecological niche modelling (ENM) is used widely to aid in conservation planning and management, often focusing on rare species characterized by the biased observations associated with restricted geographic ranges, habitat specialization, small population size and limited natural history information. Generating reliable ENMs for such species is a challenge, however, owing to issues that arise from spatial sampling bias, such as model inaccuracy and overfitting. Here, using virtual scenarios, we assess the utility of integrating occurrence data for closely related species with varying degrees of niche overlap into ENMs of focal species. We consider two approaches to merge related and focal species models: integrating occurrences of focal and related species directly as inputs with which to generate ENMs, vs. creating ENMs based on occurrences of focal and related species separately and merging results based on Bayesian inference approaches. Both single, integrated models and Bayesian inference approaches performed better than models based on focal species only when niche overlap between the focal and related species was large, across ENM algorithms examined. While assessing sensitivity and specificity separately, the performances of the two integration approaches over different ENM algorithms were complicated. The results of the study offer a novel way forward in managing the challenge of creating useful, predictive models even for the rarest species, taking advantage of the reasonably general property of niche conservatism over small‐to‐moderate amounts of evolutionary time.