What constitutes evidence of community structure? A reply to van der Maarel, Noest & Palmer

Abstract

Abstract. Van der Maarel et al. (1995) — hereafter VNP — criticize our analyses of Öland limestone grassland. They have four general objections, addressed below.1. VNP attribute our significant results to randomness (‘volatility’); however, significant deviation from randomness cannot be explained by randomness. VNP's conclusion of volatility derives from results that are inconsistent with ours. Their ecological interpretation assumes spatial and temporal near‐constancy in the vegetation; we demonstrate that these assumptions are not correct.2. We discussed physical limitations to plant module packing. VNP give estimated module sizes. We appreciate this information, though data on actual module sizes and overlap patterns would be required before the real module packing effects could be determined. If module packing were really the main cause of deficits of variance in richness, the effect would not fluctuate between sites and years, as VNP admit it does.3. The shape of the richness frequency distribution is a potentially interesting additional form of analysis, albeit one that we had chosen not to include in our analysis. However, it is surprising that VNP dismiss analysis of richness variance, but then interpret a more subtle aspect of the richness distribution — the skewness. VNP's redefinition of the terms‘Niche facilitation’for a deficit of low richness values, and ‘Niche limitation’ for a deficit of high values, leads them to misrepresent our arguments. They are mistaken in suggesting that niche limitation will necessarily lead to a skewed curve: they do not allow for the fact that the species frequencies, on which the null model is based already, incorporate effects of species interactions.‘Niche facilitation’, as defined by VNP, would lead to a variance excess, not a deficit as they assume.4. VNP's criticism of a priori guild classifications had already been met by our use of the intrinsic guild approach. Guild analyses offer the best way forward.VNP use principally the methodology of 1987. There have been a number of methodological advances since that time. We used these advances in our original paper specifically to circumvent the kinds of problems that they identify. We agree with VNP that it would be very useful to explore the mechanisms behind assembly rules with experiments, but the logical first step is to identify potential assembly rules, as we have been attempting to do. We conclude that, whilst VNP make some interesting observations, none of their criticisms invalidate our results or conclusions. Our original approach stands as the best known approach, we believe, for searching for community structure in such data, and we reaffirm the validity of our ecological conclusions.