Abstract. In temperate regions the chemistry of the lower troposphere is known to be significantly affected by biogenic volatile organic compounds (VOCs) emitted by plants. The chemistry of the lower troposphere over the tropics, however, is poorly understood, in part because of the considerable uncertainties in VOC emissions from tropical ecosystems. Present global VOC models predict that base emissions of isoprene from tropical rainforests are considerably higher than from savannas. These global models of VOC emissions which rely mainly on species inventories are useful, but significant improvement might be made with more ecologically based models of VOC emissions by plants. Ecosystems along a successional transect from woodland savanna to primary rainforest in central Africa were characterized for species composition and vegetation abundance using ground surveys and remotely sensed data. A total of 336 species (mostly trees) at 13 sites were recorded, and 208 of these were measured for VOC emissions at near-optimal light and temperature conditions using a leaf cuvette and hand-held photoionization detector (PID). A subset of 59 species was also sampled using conventional VOC emission techniques in order to validate the PID technique. Results of ecological and VOC emission surveys indicate both phylogenetic and successional patterns along the savanna-rainforest transect. Genera and families of trees which tend to emit isoprene include Lophira, Irvingia, Albizia, Artocarpus, Ficus, Pterocarpus, Caesalpiniaceae, Arecaceae, and Moraceae. Other taxa tend to contain stored VOCs (Annonaceae and Asteraceae).Successional patterns suggest that isoprene emissions are highest in the relatively early successional Isoberlinia forest communities and progressively decrease in the later successional secondary and primary rainforest communities. Stored VOCs appear to increase along the savanna-rainforest succession, but these data are more tentative. These findings are consistent with successional patterns of isoprene and terpene fluxes in North American forests and highlight the feasibility of constructing better predictive models of VOC emissions.