Middle Miocene climate and vegetation models and their validation with proxy data

The Miocene is a relatively recent epoch of the Earth's history with warmer climate than today, particularly during the middle Miocene Climatic Optimum (MMCO, approximately 17–15Ma). Although the cause of the warming is probably not only attributable to CO2, but also to changes in orography and configuration of ocean gateways, this time interval represents an ideal case study to test the ability of climate models to simulate warm climates comparable to those that the Earth may experience in the near future. However, even with higher than present-day CO2 concentrations, the MMCO warming inferred from terrestrial proxy data has been difficult to reproduce in climate models.Since fossil flora do not provide direct information on climate, but on flora and vegetation, climate model results are generally compared to climate reconstructions obtained from the fossil flora. In this study, we apply an alternative method by simulating palaeovegetation from the outputs of the climate model, using a dynamic vegetation model. Model vegetation reconstruction can then be compared to the vegetation cover indicated by the fossil flora record at the various localities, provided that a common classification of plant functional types (PFTs) is used for the data and the model. Here, we reconstruct the vegetation of the middle Miocene with the global dynamic vegetation model CARAIB, using the climatologies derived from five atmospheric general circulation models. The reliability of the simulations is examined on a presence/absence basis of PFTs by comparison of vegetation reconstructions to palaeoflora data recorded in the Northern Hemisphere and the Tropics.This comparison provides an overall agreement around 60% between model and data, when all sites and tree types are considered. Three model simulations out of five show to be better at predicting the absence than the presence. The presence of warm-temperate mixed forests in the middle latitudes, dominated by broadleaved deciduous warm temperate and subtropical trees is generally well reproduced in CARAIB simulations. However, poor agreement is obtained for the presence of tropical PFTs out of the Tropics and for warm PFTs at latitudes northward of 50°N, where climate models remain too cold to produce assemblages of trees consistent with the data. Nevertheless, the model–data comparison performed here highlights several mismatches that could result not only from missing feedbacks in the climate simulations, but also from the data. The results of the likelihood analysis on presence/absence of PFTs illustrate the uncertainties in the PFT classification of the Neogene floral records. The coexistence of some PFTs in the palaeovegetation data is impossible to reproduce in the vegetation model simulations because of the climatic definition of the modern PFTs. This result indicates either a bias in the identification of modern analogues for fossil plant taxa, or a possible evolution of environmental requirements of certain plants.