Résumé
For obvious reasons, niche modelling is clearly of low relevance when considering rare species. It also displays limitations in presence of weak environmental gradients. Moreover autocorrelation in both floristic and environmental variables is likely to unduly inflate type I error associated to niche models while methods designed to acknowledge for autocorrelation are generally of limited help when a plotbased floristic sampling is very uneven in space. Spatially implicit neutral models (SINM) may provide answers to the aforementioned problems by providing baseline abundance distributions that integrate both low equitability in species abundances and species aggregation within plots. The estimation of the two main parameters of the SINM (i.e. the fundamental biodiversity and migration numbers) is moreover straightforward from species abundance data. In addition, SINMs are generally good at fitting observed species abundance distributions (SAD) as well as reproducing classical biodiversity statistics which are derived from SADs (species richness, Shannon and Simpson-Gini indices). As such, SINMs can provide arrays of simulated non-trivial species by plots abundance data against which it is possible to test statistics obtained from nested schemes of diversity apportioning with respect to environmental variables. The breaking-down of Simpson diversity corresponding to a network of plots is particularly straightforward and provides measurements of beta-diversity for different subsets of the network (from geographic or environmental classes down to individual plots). Related beta-diversity statistics can be tested against simulations of the neutral null model as to detect “supra-neutral” patterns that cannot solely result from migration limitation and SAD characteristics. From the presented modelling and simulation framework it is possible to compare the behaviour of scarce species with the entire floristic pool and assess to what extent rare species may display specific features in relation to geography or environment. Examples will be provided from data collected in tropical forests in Central Africa