A fuzzy classification technique for predicting species’ distributions: applications using invasive alien plants and indigenous insects

ABSTRACT A new predictive modelling technique called the fuzzy envelope model (FEM) is introduced. The technique can be used to predict potential distributions of organisms using presence‐only locality records and a set of environmental predictor variables. FEM uses fuzzy logic to classify a set of predictor variable maps based on the values associated with presence records and combines the results to produce a potential distribution map for a target species. This technique represents several refinements of the envelope approach used in the BIOCLIM modelling package. These refinements are related to the way in which FEMs deal with uncertainty, the way in which this uncertainty is represented in the resultant potential distribution maps, and the way that these maps can be interpreted and applied. To illustrate its potential use in biogeographical studies, FEM was applied to predicting the potential distribution of three invasive alien plant species ( Lantana camara L., Ricinus communis L. and Solanum mauritianum Scop.), and three native cicada species ( Capicada decora Germar , Platypleura deusta Thun. and P. capensis L.) in South Africa, Lesotho and Swaziland. These models were quantitatively compared with models produced by means of the algorithm used in the BIOCLIM modelling package, which is referred to as a crisp envelope model (the CEM design). The average performance of models of the FEM design was consistently higher than those of the CEM design. There were significant differences in model performance among species but there was no significant interaction between model design and species. The average maximum kappa value ranged from 0.70 to 0.90 for FEM design and from 0.57 to 0.89 for the CEM design, which can be described as ‘good’ to ‘excellent’ using published ranges of agreement for the kappa statistic. This technique can be used to predict species’ potential distributions that could be used for identifying regions at risk from invasion by alien species. These predictions could also be used in conservation planning in the case of native species.