Résumé
Questions related to conservation biology (e.g. biodiversity) and climate change often require a long-term historical perspective in which quantitative reconstructions of past vegetation abundance/land cover at local and regional spatial scales play an important role. The Landscape Reconstruction Algorithm (LRA; Sugita 2007 a, b) estimates vegetation abundance/plant cover at the regional scale (100 km x 100 km) and local scale (area of ca. 1-5 km radius) using the models REVEALS and LOVE, respectively. The LRA is the most widely tested, validated, and applied pollen-based quantitative reconstruction approach to date. However, the method has been used primarily in the northern hemisphere. Duffin and Bunting (2008) published estimates of relative pollen productivity (RPP), a parameter needed in LRA aplications, for five plant taxa of southern Africa. Within the French research project C3A “Causes and consequences of the last great environmental crisis 3000 years ago in forests of Atlantic Equatorial Africa”, we test the LRA in the mountain woodland and savannah vegetation of Cameroon. We use modern pollen assemblages from soil samples and related vegetation surveys from 31 sites and the ERV-model to estimate RPPs of major taxa in the study region. We apply the LRA (REVEALS and LOVE models) using Holocene pollen records from two crater lakes/ponds (16.5 ha and 198 ha) in mountain woodlands and three lakes (46.7 ha, 25 ha and 8 ha) in savannahs. The RPP estimates suggest that, relatively to Gramineae (RPP=1), taxa such as Schefflera and Combretaceae have high RPP values (>1), while e.g. Podocarpus, Asteraceae, and Cyperaceae have RPPs close to Gramineae, and e.g. Alchemilla and Celtis have low RPPs (< 1). LRA reconstructions of regional and local vegetation cover indicate more wooded landscapes than suggested by pollen percentages over the entire Holocene at sites located today in mountain woodland and mixed woodland-savannah landscapes. Moreover, the decreases in woodland cover at 3 ka BP and from 1.5 ka BP are generally more pronounced in the LRA reconstructions than in pollen percentages. LRA reconstructions in combination with land-use reconstructions inferred from archaeological data will be an important step towards more realistic descriptions of past anthropogenic land-cover and land-use change in western Africa for e.g. earth-system modelling.