Bayesian analysis of combined chloroplast loci, using multiple calibrations, supports the recent arrival of Melastomataceae in Africa and Madagascar.

A new biogeographic scenario for Melastomataceae (Morley and Dick, American Journal of Botany 90(11) pp. 1638-1645, 2003) accepts an ndhF-based phylogeny for the family by Renner et al. (American Journal of Botany 88(7): 1290-1300, 2001), but rejects those authors' divergence time estimates. Morley and Dick concluded that Gondwanan vicariance, rather than the more recent long dispersal proposed by Renner et al. explains the presence of the family in Africa and Madagascar. To assess the strength of this conclusion, a Bayesian analysis was conducted on three times the amount of sequence data used before (ndhF, rbcL, rpl16; 3100 base pairs [bp], excluding all gaps). The Bayesian approach to divergence time estimation does not rely on a strict molecular clock and employs multiple simultaneous minimal or maximal bounds on node ages. Reliance on northern mid-latitude fossils of Melastomataceae for calibrations was avoided or reduced by using alternative fossil and tectonic calibrations, including all those suggested by Morley and Dick. Results reaffirm the relatively recent spread of melastome lineages among the southern continents and refute the breakup of Gondwana as a plausible explanation for the presence of Dissochaeteae/Sonerileae in Madagascar and Africa and the presence of Melastomeae in Africa and Southeast Asia. Melastomeae appear to have reached Africa around 17-15 million years (my) ago, while Dissochaeteae and Sonerileae apparently reached Madagascar at 17-15 and 20-18 my ago. I also explored the effects of constraining Melastomeae to minimally 76 my old (to have reached Africa by island hopping as postulated by Morley and Dick). This resulted in an estimate for their arrival in Africa of 35 my ago and for Dissochaeteae and Sonerileae in Madagascar of 28 and 33 my ago, still implying long-distance dispersal. The Bayesian 95% credibility ranges around these dates, however, are large. Regardless of the increasing sophistication of molecular estimates of divergence time, Gondwanan scenarios will remain untestable as long as biases in the fossil record can justifiably be invoked to explain away the absence of fossils.