Use of dispersal–vicariance analysis in biogeography – a critique
Analytical methods are commonly used to identify historical processes of vicariance and dispersal in the evolution of taxa. Currently, dispersal–vicariance analysis implemented in the softwaredivais the most widely used method. Despite some recognized shortcomings of the method, it has been treated as error-free in many cases and used extensively as the sole method to reconstruct histories of taxa. In light of this, an evaluation of the limitations of the method is needed, especially in relation to several newer alternatives. Methods
In an approach similar to simulation studies in phylogenetics, I use hypothetical taxa evolving in specific geological scenarios and test how welldivareconstructs their histories. Results
divareconstructs histories accurately when evolution has been simple; that is, where speciation is driven mainly by vicariance. Ancestral areas are wrongly identified under several conditions, including complex patterns of dispersals and within-area speciation events. Several potentially serious drawbacks in usingdivafor inferences in biogeography are discussed. These include the inability to distinguish between contiguous range expansions and across-barrier dispersals, a low probability of invoking extinctions, incorrect constraints set on the maximum number of areas by the user, and analysing the ingroup taxa without sister groups. Main conclusions
Most problems with inferences based ondivaare linked to the inflexibility and simplicity of the assumptions used in the method. These are frequently invalid, resulting in spurious reconstructions. I argue that it might be dangerous to rely solely ondivaoptimization to infer the history of a group. I also argue thatdivais not ideally suited to distinguishing between dispersal and vicariance because it cannot a priori take into account the age of divergences relative to the timing of barrier formation. I suggest that other alternative methods can be used to corroborate the findings indiva, increasing the robustness of biogeographic hypotheses. I compare some important alternatives and conclude that model-based approaches are promising.