Unravelling the factors determining tree species distributions is a key issue in tropical ecology. In this study, we analysed the extent to which soil properties and spatial aggregation influence the distribution of individual canopy species (DBH > 10 cm) in a 5-ha terra firme plot in Amacayacu National Park in the Colombian Amazon. The plot was divided into 125 quadrats of 400 m² each (20 m x 20 m), and we selected all species present in ten or more quadrats. In order to quantify the effect of the environment and geographic space, we used a logistic regression model based on species presence-absence data. Soil nutrient data were reduced a Principal Component Analysis (PCA), and spatial variables was analysed with a third degree polynomial. We used backward elimination first only with environmental data (significant axes of the PCA), and then only with the spatial variables. The final model, which was submitted to a variance partitioning, was run with all significant variables previously from the backward procedures.

Results/Conclusions

Species distribution showed a higher correlation with spatial variables than with environmental. In most species, a high proportion of the variation remained unexplained. The results obtained in this study show little importance of soils on species distribution patterns than reported before at local spatial scales, with a high random component. However, it is important to highlight the possible effect of species undersampling due to the high local diversity recorded in this forest.