Using Habitat Suitability Models Considering Biotic Interactions to Inform Critical Habitat Delineation: An Example with the Eastern Hog-nosed Snake (Heterodon platirhinos) in Ontario, Canada

Posted on Nov 7, 2014



Correspondence: Victor homasson, Department of Biology, University of Ottawa, 30 Marie Curie, Ottawa, ON, K1N 6N5, Canada. Email:

Submitted 14 August 2014 — Accepted 7 November 2014


Habitat suitability models have been used in a variety of fields, including conservation, and are considered a powerful tool to model the potential niche of species. Presence-only models have been particularly useful to define suitable habitat for rare species at a landscape scale. In Canada, the Species at Risk Act not only protects species at risk, but also their residences and critical habitats. It is thus necessary to identify which geographic areas species at risk depend on and which habitats can be considered suitable. In this study, we identify areas of high suitability for the eastern hog-nosed snake, Heterodon platirhinos, in Ontario, Canada. We employ three models – Maxent, Boosted Regression Trees, and the Genetic Algorithm for Rule Set Production (GARP) – to model the current distribution of the species. Because the eastern hog-nosed snake is a diet specialist, we also assess the importance of biotic interactions in habitat suitability models by including variables representing prey availability. The best models were combined using a consensus approach and categorical maps showing 4 conservation scenarios were built.  Maxent and Boosted Regression Trees performed better than GARP. While forest density is positively related to habitat suitability, cropland density limits the distribution of this snake. Climate also played an important role in shaping the distribution of this species. Biotic variables allowed better interpretation of the predictions made by the models by reflecting spatial bias in sampling. We discuss how habitat suitability models can help delineate the critical habitat of species at risk and whether variables representing biotic interactions should be included.

Key Words: Boosted Regression Trees, Conservation, Maxent Habitat Model.



View Full-Text    Download PDF