An assessment of interactively coupled paleoclimate-vegetation models

Vegetation is a key component of climate models. Failure to include realistic vegetation-climate interactions can cause large discrepancies in predicted surface temperature and precipitation. In this study, four different vegetation models—BIOME4, EVE, IBIS and SIVM—are interactively coupled to the GENESIS GCM and simulations run for four paleoclimatic time periods ranging from the Miocene to the Last Glacial Maximum. The simulated vegetation from each model is compared to observed paleo-distributions derived from pollen and plant macrofossils. The BIOME4 model is found to best reproduce the paleovegetation. Within each time period, changing the vegetation model causes a maximum difference in globally averaged surface temperature of 1.5°C. Regional differences approach 4°C for annual means and 7°C for seasonal means. In addition, a multiple regression analysis using surface albedo, leaf area index, soil moisture and surface roughness as predictors indicates that differences in surface temperature are generally related to differences in surface albedo, which are caused by differences in simulated vegetation. Differences in climate-vegetation feedbacks are assessed and found to be less than 0.5°C on a globally averaged basis and less than 2°C at any latitude.