A soil vegetation atmosphere transfer scheme (SVAT), describing the fluxes of heat, water vapour and CO 2 between a multi-component vegetated land surface and the atmosphere, has been developed. The SVAT has been calibrated and tested using 30 days of micrometeorological and physiological measurements collected over a sparse savannah in the Sahel. The model explicitly represents the four major functional components of the savannah: shrubs (C 3 photosynthesis), grasses (C 4), herbs (C 3) and bare soil. The leaf conductance model for the shrub, grass and herb components was calibrated against porometry measurements made in the field. The performance of the SVAT was tested against independent surface flux and temperature measurements. Agreement between the model predictions and measurements of the total net radiation, sensible and latent heat fluxes, net CO 2 exchange and surface temperatures was good: in all cases at least 80% of the variance in the measurements was explained by the model. Separate flux and surface temperature predictions for the four surface components were satisfactory, although complete verification was difficult as data were lacking for some variables. Predicted water use efficiency (WUE) of the vegetation showed a strong, non-linear dependence on vapour pressure deficit, especially for the bushes and the grasses. WUE for the grasses was about three times as large as the values found for the bushes and herbs. A SVAT model like this can be employed to address the possible effects of CO 2 enrichment and climate change on the competitive balance between different species in a plant community or in research concerning productivity and water use efficiency in mixed crops, such as agroforestry systems.