Abstract
Aims: To analyse the elevational richness pattern of tree species in New Caledonian tropical rain forests and investigate how the richness pattern is affected by standardizing on area vs standardizing on number of individuals and by sample coverage. Location New Caledonia (SW Pacific). Methods: We re-analysed data from the New Caledonian Plant Inventory and Permanent Plot Network (NC-PIPPN). In this plot network, all trees (DBH ≥ 5 cm) were inventoried from 201 plots measuring 20 × 20 m and located between 5 and 1292 m a.s.l. We tested whether the number of trees inventoried in these plots affected the elevational richness pattern, and if so, whether plots with large trees have a low species density because of a low number of inventoried trees. This was done by applying individual-based and coverage-based rarefaction procedures with different cut-off threshold values and re-testing the relationship between estimated species richness and elevation. Results: About 40% of the variability in species density is explained by variation in the density of trees, which in turn had no obvious pattern with elevation. Plots with large trees (high quadratic mean DBH) contained fewer individuals and thus fewer species. Applying rarefaction changed the observed elevational pattern from a peak in species richness approximately in the middle of the studied elevational gradient to a peak at lower elevations. Both the choice of the rarefaction procedure and the cut-off threshold value used to estimate species richness may dramatically change the elevational pattern. Conclusions: When the local species pool is large, the use of small plots relative to tree size induces a high correlation between species richness and the number of trees inventoried per plot. To analyse appropriately an elevational pattern for such small plot sizes, species richness should be rarefied to provide additional perspectives. We show that rarefaction changes the elevation at which species richness is expected to be maximal and that increasing the cut-off threshold value sharpens the observed humped pattern.