Abstract
Tree communities vary at different spatial scales and are influenced by environmental factors. To understand the role of environmental factors in tree communities requires a scale-wise analysis. The overall objective of this study was to analyse the influence of environmental drivers (soil and climate) on three aspects of forest community structure (floristic patterns, tree height–diameter relationship and leaf functional traits) at regional (>100 km2) and local (0.5 km2) spatial scales.This study was conducted in Cameroon, Gabon and the Democratic Republic of the Congo, all belonging to the lowland (< 800 m) forests of Central Africa. At the regional scale, 133 non-contiguous permanent plots of one ha separated by at least 500 m were established in mature forests. Data from the 50 ha forest block were obtained for the local scale analysis. The later was split into 50 single one ha plots and each plot was subdivided into 25 quadrats of 20 m x 20 m. Ten soil and ten climatic variables were collected and analysed from 78 plots; five leaf functional traits (leaf area, specific leaf area, leaf phosphorus nitrogen content and nitrogen: phosphorus ratio) were measured only in Korup. Principal component analyses permitted to determine soil and climate gradients, and correspondence analysis with instrumental variables enabled to decompose their influence on floristic patterns. The Second-order polynomial model permitted to model tree height-diameter relationship while linear regression and correlation were employed for trait-gradient analysis with the help of R statistical software.At the regional scale in the moist forests of eastern Cameroon, floristic patterns were characterised by few “abundant species” (> one individual ha-1 per site) and by many “scarce species” (≤ one individual ha-1 per site). The gradients in soil composition (soil texture, phosphorus and organic fertility) explained 15.42 % (76/493) mostly scarce tree species while the climatic gradients (rainfall, temperature and precipitation seasonality) explained 26.37 % (130/493) mostly abundant tree species. Tree basal area and the climatic gradient improved the prediction of tree height with only 18.9 % model error as opposed to 20.2 % and 21.7 % model error from two existing pantropical height–diameter models (Chave and Feldpausch, respectively). The new height–diameter models also improved aboveground biomass prediction and revealed that the pantropical models consistently overestimated biomass by 25 and 55 Mg ha-1. At the local scale in Korup, soil fertility was the main gradient and significantly correlated with the leaf functional traits (leaf area, specific leaf area, leaf phosphorus and nitrogen contents). Up to 33 % of these traits were significantly either over-dispersed or under-dispersed (non-random distribution) confirming the influence of abiotic filtering by the soil fertility gradient.In the prospect of possible climate change, there may be a shift in species dominance for tree species in eastern Cameroon forests while the effects of soil on the scarce species pool may limit the risk of extinction. This study thus presents an improved height-diameter model to predict the heights of Central African forest trees by integrating the effects of some easy-to-retrieve ecological predictors. At the local scale, leaf functional traits reveal the effect of soil gradient in the distribution of species and community assembly.