Alternative classifications in the intrinsic guild structure of a New Zealand tussock grassland

Guild structure was determined in the inter-tussock vegetation of a New Zealand grassland. Twenty-one sites were sampled by shoot presence/absence in 10 x 10 cm quadrats, for all macroscopic plants. Using these data, intrinsic guilds were derived (intrinsic guilds are those that are defined by the observed restrictions on species co-occurrence, not by a priori characters). Such guilds were found by heuristic searches for the guild classification that was optimal, as measured by an index of community structure, RV(gp). The optimisation was for minimum RV(gp), i.e maximum guild proportionality, relative constancy in guild representation. One hundred searches were carried out on the data, each starting from a different initial random configuration. When the quadrats were split at random into an Optimisation subset and a Test subset, a guild classification that showed significant guild proportionality in the Test subset was found in a significantly greater number of searches than expected by chance (28 out of 100). The ten of those 28 classifications that gave the tightest community structure comprised three groups. Further optimisation of representatives of these groups using the whole dataset confirmed that the community contained at least two genuinely independent, alternative guild classifications. It is concluded that two or more guild classifications can exist within the same set of species in a community. These classifications can be orthogonal in the sense that they are unrelated to each other and operate simultaneously. Attempts to correlate the demonstrated guild membership with known characteristics of the species showed some, but limited, relation to growth form/height.