The African equatorial rain forest is known to have expanded over peri-forest humid savannas during the last centuries, north as well as south of the equator, despite frequent savanna fires. This has been shown by botanical studies (e.g. Swaine et al. ABSTRACTAim To understand the persistence of a forest-savanna mosaic in places where rainfall data suggest that forest take-over should take place. To study the various modes of forest encroachment, and the role of human activities to hamper it.Location Data were collected on several forest-savanna ecotones in the coastal region of the Republic of Congo. The sites were chosen to illustrate the differing principal modes of forest expansion, corresponding to different levels of anthropic pressure. MethodsThe study sites were situated on five transects perpendicular to the ecotone (total sampled area: 1.7 ha) and 10 forest clumps in savanna (with diameters from 3 to 20 m). Along the transects botanical identification, diameter measurement and cartography were performed, while leaf area index was measured at a high resolution (every metre) along two of them. Collected data were analysed using a continuous quantification approach, which is much more useful than classical quadrat analysis. Time calibration of progression rates was performed using a simple model of the growth of the characteristic pioneer species, Aucoumea klaineana. ResultsThe two main different modes are reflected in different successional patterns. The edge diffusion is slow (its rate is evaluated to c. 1 m year )1 ) and is characterized by a progressive increase in large-diameter tree density and shadetolerant tree density away from the ecotone. Conversely, savanna to forest phase transition by coalescence of clumps exhibits high tree density remnants distributed in established forest. The composition of these remnants is compatible with that of the forest clumps in savannas.Main conclusions Three functional groups of pioneer trees are distinguished: some occupy the edge (edge pioneer), others establish clumps of forest in savanna (clump pioneers) and the longer-living A. klaineana ensures the transition to 'mature' forest. The two different observed patterns (linear edge progression and clump coalescence) can be understood with the use of a model of forest-savanna dynamics, 'FORSAT'. The two control parameters are the annual rainfall and the frequency of man-made fires in each savanna. In particular, an increase in the fire frequency can lead to a shift from the coalescence regime to the edge progression one.

An important property of plant communities is the Leaf Area Index (LAI), which is the vertically integrated surface of leaves per unit of ground area. Leaves are the primary sites of photosynthesis and transpiration, thus the LAI, which conditions the light interception by the canopy, is directly related to carbon and water exchange with the atmosphere at the stand scale (McNaughton & Jarvis 1983). LAI also has an impact on tree growth through the interception of light. Light availability below canopies is the principal limiting factor of tree recruitment and growth in forests (Denslow et al. 1990). Several methodologies have been used for measuring LAI in the field. These can be classiffed in four categories (Marshall & Waring 1986): (1) direct measurements by litterfall collection or destructive sampling, (2) allometric correlations with variables such as tree height or tree diameter, (3) gap-fraction assessment (e.g. with hemispherical photographs), (4) measurement of light transmittance with optical sensors.