The knowledge of tropical palaeoclimates is crucial for understanding global climate change, because it is a test bench for general circulation models that are ultimately used to predict future global warming. A longstanding issue concerning the last glacial maximum in the tropics is the discrepancy between the decrease in sea-surface temperatures reconstructed from marine proxies and the high-elevation decrease in land temperatures estimated from indicators of treeline elevation. In this study, an improved inverse vegetation modeling approach is used to quantitatively reconstruct palaeoclimate and to estimate the effects of different factors (temperature, precipitation, and atmospheric CO 2 concentration) on changes in treeline elevation based on a set of pollen data covering an altitudinal range from 100 to 3,140 m above sea level in Africa. We show that lowering of the African treeline during the last glacial maximum was primarily triggered by regional drying, especially at upper elevations, and was amplified by decreases in atmospheric CO 2 concentration and perhaps temperature. This contrasts with scenarios for the Holocene and future climates, in which the increase in treeline elevation will be dominated by temperature. Our results suggest that previous temperature changes inferred from tropical treeline shifts may have been overestimated for low-CO2 glacial periods, because the limiting factors that control changes in treeline elevation differ between glacial and interglacial periods.biome model ͉ biome pollen scores ͉ palaeoclimatology ͉ pollen ͉ vegetation model inversion P ollen data show a lowering of treeline elevations during the last glacial maximum (LGM) by Ϸ1,000-1,700 m in mountains at a wide range of tropical and subtropical locations (1, 2). Assuming this decrease corresponds to a decrease in mean temperature and given a lapse rate of 5-6°C km Ϫ1 , this would imply a substantial cooling of Ϸ5-10°C in these areas, particularly at high elevations in the tropics (1, 2). However, a new faunal reconstruction of sea surface temperature (Multiproxy approach for the reconstruction of the glacial ocean surface: MARGO) for the tropics indicates a more limited cooling of Ϸ2°C at low elevations during this period (3, 4), which is slightly more than the Ϸ1.5°C estimated by using CLIMAP (5). These reconstructions are clearly not consistent, and although the differences could in principle be explained by a steeper atmospheric lapse rate, this parameter needs further validation by using more reliable data (6-8). Nonetheless, this lack of agreement leaves considerable uncertainty about what the tropical climate was really like during the last ice age.Physiological data and models have demonstrated that the processes that modify carbon and water uptake in plants are highly dependent on CO 2 concentrations (9, 10). This suggests that modern plant-climate relationships are not representative of interactions between plants and climate in the past (9 -11), because the atmospheric CO 2 concentration has f luctuated ...