Local-scale structuration of species-habitat association in the semi-deciduous tropical rainforest in eastern Cameroon

Abstract

Tropical forests are hyperdiverse, and although broad-scale variation in diversity can be reasonably well explained, local-scale variation in diversity is harder to understand. It has been revealed that species may show species-habitats associations. However, how the assembled species are associated with local habitats and how community-habitat associations vary regarding species abundance and class size remains unclear. Here, we analyzed the community-habitat associations by applying the multivariate regression trees (MRT) technique on data of 22,064 trees distributed across 30 1-ha plots varying in topography and soil conditions in a semi-deciduous tropical rainforest of Cameroon. Further, the indicator species analysis was used to determine the most important species for structuring assemblages.The MRT divides the whole tree community inventoried into four habitats types: Habitat 1 (H1) sandy loamy, acidic soil with a low Na concentration (pH ≤ 6.245 and [Na] < 4.535 cmol kg−1); H2 equal to neutral sandy soil with a small Na concentration (7 ≤ pH > 6.245 and [Na] < 4.535 cmol kg−1); H3 equal to Loamy sandy soil with average Na concentration (4.535 ≤ [Na] < 6.243 cmol kg−1) and; H4 equal to Loamy sandy soil with high nutrients concentration ([Na] ≥ 6.243 cmol kg−1). Coincidentally, the same habitat types were also generated for the understorey and large trees classes taken separately, although a difference was somehow recorded for the small stems class. Among investigated environmental variables, soil Na content and, to a lesser extent, soil pH were identified as the primary responsible for the first split for habitat types for understorey and large trees classes. By contrast, the most determinant for discrimination habitat types for small stems was soil cation exchange capacity, followed by soil moisture content. Overall, 29 species were found as indicators characterizing the four habitat types for the whole tree community, with index values varying between 33% and 99 %. But specifically, the number of indicators species for large trees was 28. Intriguingly, some indicator species within the large trees class changed to either non-indicator or intermediate indicator species in the understorey trees class that was best represented by as less as 16 species. A comparable species richness was recorded across all plots of the studied area; however, significant variability was found regarding species abundance, with habitats H1 and H4 with higher average abundance values. This study highlights that soil variables adding to topographical habitat filtering were important in shaping the composition of local species.