This week we gave the insightful analysis treatment to Wardle et al’s 1994-97 experiment in a New Zealand pasture, in which they removed all the species from 360 20cm patches and allowed them to grow back in, while selectively pruning out different mixtures of functional species groups. Good stuff. Especially cool was their careful measurment of below-ground biota and below-ground ecosystem functioning. Especially especially cool was the fact that they collected a time series of data across the 3 year times span of the experiment. Most of these diversity experiments have to destructively sample to collect their data and so they end up with a single snapshot of the state of the ecosystems at the arbitrary end. These folks sampled just a few of their plots per time step over a bunch of steps, so they had lower sample sizes at any time but a real dynamic story to study.

The paper itself is a whopper, 34 pages in Ecological Monographs.

David A. Wardle, Karen I. Bonner, Gary M. Barker, Gregor W. Yeates, Kathryn S. Nicholson, Richard D. Bardgett, Richard N. Watson, Anwar Ghani. Plant Removals in Perennial Grassland: Vegetation Dynamics, Decomposers, Soil Biodiversity, and Ecosystem Properties. Ecological Monographs, Vol. 69, No. 4 (Nov., 1999), pp. 535-568, doi:10.2307/2657230

scholar.google link

Thankfully the insightful analysis is the usual one page, below the fold.

This paper answers several of our concerns about earlier experiments, and in doing so highlights the difficulty in interpreting experiments which are as intricate as we would like.

Realism/Generalizability: In our discussions we have identified that some of the “random assemblage” experiments are unconvincing because they don’t necessarily represent particular processes of the real world. This experiment, even more than the Cedar Creek examples, is located in a very specific place. The design (round 20cm plots, grazing) is meant to replicate actual conditions, and the experimenters seem familiar with the local species and the way they interact in non-experimental conditions. This is a strength of the paper. It allows the authors to translate their statistical outcomes into believable narrative. They can confidently go beyond correlative observation and identify causative mechanism, as Bengstton would wish them to. On the other hand, we are left wondering: do their conclusions translate to other ecosystems? There is no effort to reproduce the general pattern of their results under differing conditions or with differing species assemblages. The authors make few over-reaching claims about such generalizability, and this may have been a prudent choice. In Individual-based Modeling and Ecology, Grimm and Railsback argue that ecology’s goal may not be to create theories which are applicable in any context, but rather to create theories which are locally robust and robustly identify their contextual limits. In a small way, Wardle et al’s results highlights the tension between these goals.

Expected results: The above-ground aspects of the study shared some of the weaknesses of our earlier papers. In particular, much of treatment response was driven by the presence or absence of one or two canopy-dominant species (L. perenne, P. annua). Total biomass and canopy cover was reduced when these high biomass, large canopy species, were excluded. This could be, as Hodgson claimed of Naeem et al, a “forseeable consequence of the experimental design”. Likewise, the authors highlight that final species richness was decreased when high-species-count functional groups were removed, and that shoot:root ratios were higher when rooty L. perrene was excluded. This does at least support the hypothesis that one of the strongest effects on ecosystem function of biodiversity may simply be the presence or absence of a few dominants.

Unexpected results: This contrasted with some of the responses observed in the below-ground functional groups. In these cases, below-ground response to above-ground perturbation was perhaps not as directly forseeable, and the localized knowledge of the authors allowed them to confidently link the above ground and below ground dynamics.

Time and equilibrium: The experimental design ingeniously allowed both for full destructive sampling of the plots and a robust time series of data collection. This answered my concerns with several of the previous experiments that the results obtained might have been significantly a function of the time they were collected. The authors explicitly caution that “it is unclear whether the system had reached an asymptotic phase with regard to these plant properties by the end of the study”, and indeed, their time-trend graphs suggests that had they sampled at a single time several of their conclusions might have been different.