Ecological niche models (also referred to as species distribution models) can be used for two broad purposes: to create an accurate prediction of a species spatial distribution and to infer the environmental correlates influencing this distribution. The former has received considerable attention in the literature and we can create niche models that accurately portray species distributions. It is less clear if the response curves created by a given model are ecologically informative in terms of understanding a species broad environmental niche; are we able to capture the relationship between species occurrence and niche parameters (environmental variables)? Basically, a model that can accurately predict the distribution of a species does not necessarily uncover the true relationship between environment and species occurrence. However, if the aim is to interpret the results from an ecological perspective, determining these underlying species-environment relationships is key. Further, if we hope to use niche models to predict how species will respond to different climate scenarios we should capture the true environmental relationship between predictor and species occurrence. To determine how well niche models capture true broad environmental dimensions of the niche we use a simulated species approach. This approach allowed us to create a known relationship between species occurrence and a given environmental predictor variable. We then asked how well different ecological niche modeling methods captured the shape and magnitude of this known environmental response. We compared GARP, general additive models, boosted regression tree and maximum entropy models to determine which could capture accurate species-environment relationships.Results/Conclusions
We found that maximum entropy models and boosted regression tree models could capture known relationships while GARP and general additive models did not perform as well across all environmental variables. These results indicate that niche modeling has the potential to capture the shapes of ecologically informative niche dimensions – an encouraging result given the potential application of these methods to both applied and theoretical research. Nonetheless, we stress that an increased focus on understanding why certain methods perform well is essential to continued development in the field.
