Identifying global zoogeographical regions: lessons from Wallace
When dividing the world into zoogeographical regions, Alfred Russel Wallace stipulated a set of criteria by which regions should be determined, foremost the use of generic rather than species distributions. Yet, recent updates of Wallace's scheme have not followed his reasoning, probably explaining in part the discrepancies found. Using a recently developed quantitative method, we evaluated the world's zoogeographical regions following his criteria as closely as possible.
We subjected presence–absence data from range maps of birds, mammals and amphibians to an innovative clustering algorithm, affinity propagation. We used genera as our taxonomic rank, although species and familial ranks were also assessed, to evaluate how divergence from Wallace's criteria influences the results. We also accepted Wallace's argument that bats and migratory birds should be excluded (although he was contradictory about the birds) and devised a procedure to determine the optimal number of regions to eliminate subjectivity in delimiting the number of regions.
Regions attained using genera (eight for mammals and birds and six for amphibians) strongly coincided with the regions proposed by Wallace. The regions for amphibians were nearly identical to Wallace's scheme, whereas we obtained two new ‘regions’ for mammals and two for birds that largely coincide with Wallace's subregions. As argued by Wallace, there are strong reasons not to consider these as being equivalent to the six main regions. Species distributions generated many small regions related to contemporary climate and vegetation patterns, whereas at the familial rank regions were very broad. The differences between our generic maps and Wallace's all involve areas which he identified as being uncertain in his regionalization.
Despite more than 135 years of additional knowledge of distributions, the shuffling of generic concepts, and the development of computers and complex analytical techniques, Wallace's zoogeographical regions appear to be no less valid than they were when he proposed them. Recent studies re-evaluating Wallace's scheme should not be considered updates as such because they have not followed Wallace's reasoning, and all computer-based analyses, including this one, are subject to the vagaries of the particular methods used.