Ecosystem effects of biodiversity: a classification of hypotheses and exploration of empirical results

The increasing evidence for effects of biodiversity on ecosystem properties and processes indicates that effects may vary strongly over the range of studied systems. In order to explore patterns and draw some preliminary conclusions on biodiversity effects in different ecological contexts, we have compiled empirical data from original studies and reviews on effects of biodiversity in natural, agricultural, and forestry systems. We classified hypotheses according to observed ecosystem variables and trophic contexts; trophic contexts were distinguished according to the trophic position of the observed ecosystem variable and the trophic position of the organismic group whose diversity is hypothesized to cause an effect. With four basic ecosystem variables and 14 trophic contexts we obtained 56permutations or specific hypotheses. Of these, 20 have so far been experimentally tested. We review the results reported from experimental tests of each hypothesis and examine how the reported effects are associated with characteristics of the studied systems. In natural communities, positive effects were most consistently observed for plant diversity affecting plant productivity and stability thereof and, in a broad sense, nutrient retention. Positive diversity effects were most prominent in grassland-type ecosystems, in experimental diversity gradients, with intermediate-to-high species numbers and wide ranges of the diversity gradient, or with randomized species composition. While some of the studies found negligible effects of biodiversity, effects that are beneficial to humans were found in a wide range of ecosystem contexts, even at the limited time scale of ecological experiments. However, our analyses indicate that general statements or predictions about effects of species diversity can only be made with reference to specific variables and trophic contexts. Our classification of hypotheses should help to identify unexamined research questions and to group the variety of hypotheses into classes in which potential diversity effects are mediated by similar mechanisms, thus yielding a more consistent explanatory pattern of experimental results.