When is variable importance estimation in species distribution modelling affected by spatial correlation?

Species distribution models are generic empirical techniques that have a number of applications. One these applications is to determine which environmental conditions most important for species. The calculation this variable importance depends on assumptions, including the observations used estimate independent each other. Spatial autocorrelation, common feature factors confounds assumption. Besides, many species trained using explanatory variables different levels spatial autocorrelation. In study we quantified effects differences in autocorrelation and type responses gradients estimations models. We simulated data both predictors species, so were control true contribution every model could be estimated after fitting found inflated estimates, but only when response linear. This inflation effect was larger preferences coincided with dominant site. Additionally find unimodal yield systematically higher compared linear responses. conclude relative cause bias estimations. way, helps clarify systematic controlled approach how make proper inferences about