Applying regional climate modelling to the Guineo-Congolian region, with the MAR model © X. Fettweis

To perform paleovegetation reconstruction with dynamic vegetation models (DVM), climate simulation outputs are needed. The mid-Holocene (6 ka), LGM (21 ka), Eemian (125 ka) and Pliocene (ca. 3 Ma) snapshot simulation results of the various models implied in the PLIOMIP/PMIP inter-comparisons will be tested for Equatorial Africa. Downscaling will be applied to the climate simulation outputs for Holocene time slices. Future scenarios of the IPCC Coupled Modelling Intercomparison Project will also be selected.

Example of distribution modelling for Milicia © A. Hambuckers et al.

To study the equilibrium (potential) distributions of a set of focus rainforest tree species at different times in the past using niche-based modelling (SDM), observed distribution data will be assembled for selected rainforest tree species from various databases like GBIF or atlases of African floras. The potential distributions of these species will be derived from niche modelling for the present using present-day gridded climatological data, with also downscaled climate simulations for past time periods.

Architecture of the CARAIB vegetation model © L. François et al.

The CARAIB DVM will benefit from current improvements relating to tropical forests. These consist in upgrading the model for inclusion of focal rainforest tree species, as well as other plants of African forests and savannahs that are not modelled in BIOSERF program. Other planned improvements relate to the fire module of CARAIB and a seed dispersal/plant migration module.

An example of landscape mosaic in the Ivindo National Park, Gabon © O. Hardy

Here, we will run the DVM without the dispersal module (i.e., assuming equilibrium with climate) for the present, the pre-industrial, and all planned snapshot paleosimulations. We will compare the results for all simulated species to niche-based modelling. The distributions of the focal tree species obtained for the Eemian and the Holocene will be compared. Combined with data from WP3, the results for the LGM will allow defining glacial forest refugia. A transient CARAIB DVM model simulation will be run over the Guineo-Congolian region from the LGM to pre-industrial time using, for a subset of the focal trees, the “dispersal module” previously developed. Probability of fires will also be examined.