Mid-Cenozoic climate change, extinction, and faunal turnover in Madagascar, and their bearing on the evolution of lemurs

Abstract

Background: Was there a mid-Cenozoic vertebrate extinction and recovery event in Madagascar and, if so, what are its implications for the evolution of lemurs? The near lack of an early and mid-Cenozoic fossil record on Madagascar has inhibited direct testing of any such hypotheses. We compare the terrestrial vertebrate fauna of Madagascar in the Holocene to that of early Cenozoic continental Africa to shed light on the probability of a major mid-Cenozoic lemur extinction event, followed by an "adaptive radiation" or recovery. We also use multiple analytic approaches to test competing models of lemur diversification and the null hypothesis that no unusual mid-Cenozoic extinction of lemurs occurred.

Results: Comparisons of the terrestrial vertebrate faunas of the early Cenozoic on continental Africa and Holocene on Madagascar support the inference that Madagascar suffered a major mid-Cenozoic extinction event. Evolutionary modeling offers some corroboration, although the level of support varies by phylogeny and model used. Using the lemur phylogeny and divergence dates generated by Kistler and colleagues, RPANDA and TESS offer moderate support for the occurrence of unusual extinction at or near the Eocene-Oligocene (E-O) boundary (34 Ma). TreePar, operating under the condition of obligate mass extinction, found peak diversification at 31 Ma, and low probability of survival of prior lineages. Extinction at the E-O boundary received greater support than other candidate extinctions or the null hypothesis of no major extinction. Using the lemur phylogeny and divergence dates generated by Herrera & Dàvalos, evidence for large-scale extinction diminishes and its most likely timing shifts to before 40 Ma, which fails to conform to global expectations.

Conclusions: While support for large-scale mid-Cenozoic lemur extinction on Madagascar based on phylogenetic modeling is inconclusive, the African fossil record does provide indirect support. Furthermore, a major extinction and recovery of lemuriforms during the Eocene-Oligocene transition (EOT) would coincide with other major vertebrate extinctions in North America, Europe, and Africa. It would suggest that Madagascar's lemurs were impacted by the climate shift from "greenhouse" to "ice-house" conditions that occurred at that time. This could, in turn, help to explain some of the peculiar characteristics of the lemuriform clade.

Keywords: Colonization; Diversification; Eocene-Oligocene transition; Evolutionary modeling, transoceanic dispersal; Fossils; Mid-Cenozoic extinction; Primate.

Fig. 1

Phylogeny of lemurs used in this analysis, modified from [1]

Fig. 2

Changes in global temperatures during the Cenozoic. a Global temperature from 50 Ma to today. Smoothed values (red line) are generated by a 300-knot generalized additive model (GAM) with cubic splines of temperature over time. b Rate of change of global temperature from 50 Ma to today

Fig. 3

Modeling lemur diversification and probability of survival using TreePar. a Changes in diversification over time, when the model is forced to include mass extinction. b Probability of survival over time, when the model is forced to include mass extinction

Fig. 4

Composite modeling results (BAMM, RPANDA, Tree Par, and TESS) with credible intervals from BAMM (grey) and TESS (light blue). Scst = speciation rate constant (for the model); Ecst = extinction rate constant; Evar = Extinction rate variable

Fig. 5

Comparison of profiles. a AICc profile of tests for large-scale extinction in RPANDA. AICc scores from models that feature constant speciation rates and allow extinction rate to vary as a function of an environmental variable. The model that had a sharp increase in extinction rate at 34 Ma was best but was statistically indistinguishable from other models with extinctions at similar times (black points on grey shading) and moderately better supported than all other candidate models (dashed lines indicate ΔAIC of 4). Grey points show AICc of models with temperature and its rates of change influencing extinction rate and null model; white circles show AICc of models with time-varying speciation and extinction rates. b AIC weight profile for models with extinctions compared to null model. c Bayes Factor profile for timing of extinctions with 95% extinction (95% CI: 91–99%). d Bayes Factor profile for timing of extinctions with75% extinction (95% CI: 50–93%). e 2 Bayes Factor profile for timing of extinctions with 5% extinction (95% CI: 6–52%). Grey shaded areas indicate times at which moderate amount of support for extinction is found