Poster

         Biogeography

Improving MaxEnt ecological niche models: assessment of some limitations and assumptions using virtual and real species

Presenting Author
Geraldine (Gerry) Allen
Description
Ecological niche models are widely used to identify ecological niches and project these in geographic space, but understanding their limitations and underlying assumptions is essential. The ENM method MaxEnt contrasts species occurrence points with sampled background points to estimate ecological niches. We generated virtual species with the R package virtualspecies using climate data from a region of western North America, and analyzed these with MaxEnt to explore the effects of various model parameters and assumptions. We considered the effects of different background areas (defined by radii of 10 to 500 km drawn around occurrence points), and found that larger backgrounds gave better fit of model predictions to species occurrences. Including ecologically divergent outliers led to overestimates of niche breadth, especially for species with narrower niches. Two assumptions implicit in MaxEnt and other methods are that a species is in equilibrium with its environment (it occurs in all suitable habitats) and that it has niche homogeneity (its ecological requirements are similar throughout its geographic range). We used virtual species with some occurrence points removed to simulate disequilibrium along a known directional gradient, and found that model predictions were sensitive to the background selected. We created composite species by combining pairs of virtual species, and showed that model predictions for these were similar to the combined predictions for the individual species pairs. These results can be used to investigate spatial ecological patterns in real species. We tested our conclusions using Quercus garryana, a western white oak with a well-documented geographic range.