AIM : We used a hierarchical fractal-based sampling design to test how sampling scale influences i)
beta diversity and ii) inferences on the modelled contribution of niche- versus dispersal-based
assembly processes in structuring tree and bird assemblages.
Location Coastal forest fragments, South Africa
METHODS : We surveyed 103 tree and 267 bird points within eight forest fragments and partitioned
beta diversity (βsor) into its turnover (βsim) and nestedness (βnes) components. We evaluated how
sampling at fine, intermediate and coarse scales influenced beta diversity components and
compared how tree and bird beta diversity respond to sampling grain variation. We then explored
the relative contributions of niche- and dispersal based assembly processes in explaining spatial
turnover as a function of sampling grain and/or study taxon by using multiple regression modelling
on distance matrices and variance partitioning. RESULTS : Beta diversity (βsor) of trees and birds was mainly explained by spatial turnover (βsim) at all
sampling scales. For both taxonomic groups, βsor and βsim decreased as sampling scale increased.
Beta diversity differed among trees and birds at fine, but not at coarse sampling scales. Dispersalbased
assembly processes were the best predictors of community assembly at fine scales, whereas
niche-based assembly processes were the best predictors at coarse scales. Most of the variation in
tree community composition was, however, explained at fine scales (by dispersal-based assembly
processes), while most of the variation in bird community composition was explained at coarse
scales (by niche-based assembly processes).
MAIN CONCLUSIONS : Our study shows that inferences from beta diversity are scale dependent. By
matching the grain of the data with the grain at which predictor variables and associated processes
are likely to operate, multi-scale sampling approaches can improve biodiversity conservation and
should be part of incentives directed at ecological sensible conservation plans.