Here be dragons: A tool for quantifying novelty due to covariate range and correlation change when projecting species distribution models

摘要: Aim Correlative species distribution models (SDMs) often involve some degree of projection into novel covariate space (i.e. extrapolation), because calibration data may not encompass the entire interest. Most methods for identifying extrapolation focus on range each model individually. However, can occur that is well within univariate variation, but which exhibits combinations between covariates. Our objective was to develop a tool detect, distinguish and quantify these two types novelties: covariates. Location Global, Australia, South Africa. Methods We developed new multivariate statistical tool, based Mahalanobis distance, measures similarity reference domains by accounting both deviation from mean correlation variables. The method also provides an assessment detection most influential covariates leading dissimilarity. As example application, we modelled Australian shrub (Acacia cyclops) widely introduced other countries compared data, global against globally in Africa. Results The successfully detected quantified dissimilarity points were either outside or formed (correlations) still For A. cyclops, more than half (6617 10,785) found lie exhibited distorted correlations. Not all climate used modelling contributed novelty equally over geographical projection. Main conclusions Identifying non-analogous environments critical component interrogation. (ExDet) be as quantitative exploring interpreting projections correlative SDMs available free download stand-alone software http://www.climond.org/exdet.