The possible use of fire for the management of the Ankarafantsika Reserve in the northwest of Madagascar and of its surrounding area is studied. Within this savanna landscape large parts of the remaining dry forests still exist with a unique biotic diversity, both in terms of total number of species and endemism. Unfortunately, mainly man-induced uncontrolled fires threaten these forests.Actual and former fire regimes of the local communities are analysed. The use of fire is an integrated part of land use and is also governed by socio-cultural traditions. The impact of fire on the dynamics of dry forests and grass savannas is studied considering the specifics of different fire regimes. We propose that a deliberate and controlled use of fire respecting the vegetation stage and the defined objectives could be an appropriate management tool. The strategy of a fire management is elaborated considering both the conservation of biodiversity and improvement of the livelihood of the local population depending upon the Reserve's resources. Obviously, a sustainable management of the natural resources requires a substantial participation of the community.

Question: What are the genesis and development of thicket clumps within a savanna landscape at geomorphically different locations and what are the driving forces? Location: The Kagera Region, in the border area of Rwanda, Uganda and Tanzania. Methods: The vegetation of 32 dry evergreen thicket clumps and their surrounding savannas have been analysed at different geomorphic locations. At each vegetation plot Na + , K + , Mg 2+ , Ca 2+ , Al 3+ , Fe 2+/3+ , H + , P, C, N, bulk density and particle size were determined for each soil horizon. The impact of soil and termite mounds on thicket clump dynamics on seasonally waterlogged plains, gentle slopes and stony hillsides were assessed. Results: Thicket clumps and their surrounding savannas have a distinct structure and floristic composition. They also have distinct soil properties although parent materials are the same. On seasonally waterlogged plains, new thicket clumps can develop on Macrotermitinae mounds; on stony hillsides, Trinervitermes and Macrotermes show a uniform distribution pattern and may initiate the genesis of thicket clumps. Conclusions: Geomorphology broadly determines the significance and interactions of the main factors affecting site-specific vegetation dynamics. On seasonally waterlogged plains, thicket clumps are restricted to termite mounds. Since intra-species competition dictates a minimal distance between neighbouring Macrotermitinae colonies, thicket clumps do not coalescence. By contrast, on stony hillsides, the vegetation mosaic is highly dynamic and determined by the interplay of several factors. The growth of thicket clumps is mainly a function of the fire regime and the browsing intensity. At the present time, frequent cool, early dry season fires and the near absence of large browsers have favoured the advance and coalescence of thicket clumps and forest patches on stony hillsides.Typical pioneer species are marked with an asterisk; Most frequent nucleus trees are written in bold; Nucleus trees in thicket clumps on stony hillside from Karama could not been determined as they have been cut.

No abstract

Aim  The spatio‐temporal dynamics of dry evergreen forest patches in the savanna biome of the Kagera region (north‐western Tanzania) are largely unknown owing to a lack of pollen and macrofossil evidence. Our aims were to reconstruct local‐scale shifts of the forest–savanna boundary in order to determine whether the forests have been expanding or retreating on a centennial and millennial time‐scale. Location  The Kagera region of north‐western Tanzania, East Africa. Methods  The vegetation reconstruction was based on analysing δ13C signatures in soils along a transect spanning both C4 open savanna and C3 forest vegetation. Furthermore, we fractionated soil organic matter (SOM) according to density and chemical stability to analyse δ13C values of soil fractions with distinct radiocarbon ages. Results  We found sharp changes in δ13C signatures in bulk SOM from the forest to the savanna, within a few metres along the transect. The forest soil profiles carried a persistent C3‐dominated signature. Radiocarbon dating of the oldest, most recalcitrant forest soil fraction yielded a mean age of 5500 cal. yr bp, demonstrating that the forest has existed since at least the mid‐Holocene. The savanna sites showed a typical C4 isotopic signature in SOM of topsoils, but subsoils and more recalcitrant SOM fractions also contained signals of C3 plants. The dense soil fraction (ρ > 1.6 g cm−3) carrying a pure C4 label had a mean age of c. 1200 cal. yr bp, indicating the minimum duration of the dominance of grass vegetation on the savanna site. At the forest edge, the older C4 grass signature of SOM has steadily been replaced by the more negative δ13C fingerprint of the forest trees. As this replacement has occurred mainly in the 10‐m‐wide forest–savanna ecotone over the last c. 1200 years, the forest expansion must be very slow and is very likely less than 15 m century−1. Main conclusions  Our results suggest that forest patches in the Kagera savanna landscape are very stable vegetation formations which have persisted for millennia. During the last millennium, they have been expanding very slowly into the surrounding savanna at a rate of less than 15 m century−1.