Studying the effects of plant species richness on ecosystem functioning:: does the choice of experimental design matter?

We established two parallel greenhouse experiments to empirically test the effect of experimental design on the conclusions made of the effects of plant species richness on ecosystem functioning. The experiments included an identical group of six grassland plants and were performed under identical environmental conditions, but were set up according to two different designs. In the richness design (RD), which specifically aims at testing the effect of species richness, each replicate at each richness level was chosen by a separate random draw from the total pool of species, whereas in the richness and composition design (RCD), which aims at testing the effect of both richness and composition, richness levels included deliberately replicated monocullures (at the one-species level) and mixtures (at other levels) of constituent species. When regression analysis was applied, both experimental designs found a positive effect of plant species richness on primary production (estimated using shoot mass accumulation), with species richness explaining 34% and 16% of the total variation in production in RD and RCD, respectively. Based on an overyielding analysis, this positive overall effect of species richness on primary production was in both experiments caused by the sampling effect rather than complementarity. When analysis of variance was applied, RCD also found that shoot production was greatly affected by the presence of one species, the legume Trifolium hybridum, and that 98-99% of variation in primary production within richness levels, i.e. the variation not explained by species richness, was explained by species composition. It appears that while the RD experiment suggests that plant species richness was a significant controller of primary production in our experimental grassland community, the RCD experiment suggests that primary production was mainly determined by plant species composition, i.e. the identity of the species present. Our results are consistent with earlier mathematical simulations in that experimental designs differ in their ability to discriminate the effects of species richness and composition on ecosystem functioning and may therefore lead to different conclusions of the role of species richness in functioning. We propose that this may partly explain the different relative role of plant species richness in ecosystem functioning in earlier investigations using RD and RCD.