Eucalyptus plantations have been introduced since 1978 on savannah soils of the coastal plains of Congo, but there is still little information on the effect of silvicultural practices on soil organic carbon dynamics after afforestation on these savannahs. The objectives of this study were to assess the effects of two experimental site preparation treatments on soil CO2 efflux, tree growth and soil carbon balance during the first year following plantation establishment. One treatment involved mechanical soil disturbance with disk harrowing (D), whereas in the second treatment (H), savannah grasses were killed by herbicide application before planting, without mechanical soil disturbance. Soil respiration and soil water content were monitored for 1 year following treatment application, at 2-week intervals. We hypothesized that mechanical soil disturbance would increase soil CO2 efflux, but the results did not support this hypothesis. The cumulated soil CO2 efflux over 1 year was not significantly different in the two treatments and averaged 658 g C m-2. In contrast, tree growth was significantly increased by disk harrowing, maybe as a result of decreased soil penetration resistance. Carbon inputs to the soil from savannah residues (428 g C m-2) were outweighed by the annual carbon outputs through heterotrophic respiration (505 and 456 g C m-2 in the H and D treatments, respectively) leading to a slightly negative soil carbon budget in both treatments 1 year after afforestation.