Studies investigating the theory of tropical refugia for vertebrates have previously focused on a handful of species or a single taxonomic group. We sought to identify the potential location of cross-taxonomic refugia of African birds and mammals in the Last Glacial Maximum, and used historic climate data to hindcast the location of past ranges of species based on their current distributions.
Mainland sub-Saharan Africa.
Using current distributions of 537 mammal and 1265 bird species, we modelled the past distribution of species, taking advantage of recently available reconstructions of climate for the Last Glacial Maximum. Modelled historical ranges were verified individually using standard techniques for evaluating the precision of bioclimatic envelope models. Potential refugia were identified as those areas with a higher overlap of climatically suitable ranges (i.e. levels of species richness) than expected based on randomizing of the modelled past climatically suitable ranges in the sub-Saharan domain and on the level of resource availability (by modelling past species richness patterns as would be expected given the water–energy theory).
Our models show that during the Last Glacial Maximum areas of high concentration of climatically suitable ranges of birds and mammals tend to aggregate, more than can be accounted for random placement of ranges and resource availability (ecological processes), in the same six areas: Upper Guinea, the Cameroon Highlands, the Congo Basin, the Ethiopian Highlands, the Angola–Namibia area and the south-east part of South Africa.
The unusually high aggregation of predicted suitable ranges for birds and mammals in six relatively small geographical areas corresponds to the location of some of the previously suggested refugia. We interpret this – and the similarity of patterns obtained for both birds and mammals – as a strong indication of the existence of refugia in those areas. The results also illustrate the usefulness of bioclimatic envelope models, coupled randomization techniques and macroecological models, for the reconstruction of geographical distribution patterns in the past.|