A practical framework for a theory-driven ecological niche modeling workflow

Distributional ecology provides a multidimensional understanding of the complex ecological, evolutionary, and biogeographic factors shaping species’ distributions. Distributional ecology uses ecological niche modeling (ENM) serving as a quantitative approach to estimate species’ ecological niches and their manifestation as likely geographic ranges. An important ongoing debate is how to choose a suitable algorithm and its parameters to perform models well. Nevertheless, a main question should be what ecological niche is being reconstructed, the realized or fundamental? Current protocols and emergent evaluation metrics have only focused on reconstructions of the realized niche, driven by the unbalanced credibility between present and pseudo-absence (or background) occurrences, which often prioritize fitting to the available data while overlooking species’ physiological and ecological constraints. Our findings indicate that generalized linear models (GLMs) effectively reconstruct most of the fundamental niche, whereas hypervolume methods, such as kernel density estimation (KDE) and Marble Algorithm (MA), tend to overfit the data and perform poorly. Similarly, Maxent exhibits limitations in characterizing the fundamental niche. We present a conceptual framework to guide assumptions and workflows in ENM applications to facilitate model selection and interpretation.