Testing the Effect of Ecosystem Composition/Structure on Its Functioning

Biodiversity can be divided into three components : composition, structure and function. A major challenge for ecologists is to quantify to what extent ecosystem functioning is affected by a change in ecosystem composition or structure. The limitations of 'natural experiments', microcosms, retrospective studies and modelling are noted. An appropriate research design requires monitoring replicated field sites before and after manipulation, alongside equally replicated control sites. Manipulations involve deletion or addition of species or alteration of their abundances, and possible candidates include dominants, ecosystem switchers, and apparent redundant and keystone species. Suitable indices of ecosystem functioning are considered. Interaction effects are paramount in analysing the data with emphasis on the biological significance of any change in functioning detected. Suitable approaches include curve fitting and univariate and multivariate analyses of variance with their non-parametric equivalents. Ways to improve statistical power to avoid type II errors and deal with asymmetrical and poorly replicated designs are noted. The long term nature of this new task for ecology necessitates careful attention from the outset to clarifying the conceptual issues and choice of experimental design and suitable statistical analyses.