Abstract. A tussock grassland, in Blackrock Reserve, New Zealand, was sampled thoroughly at scales ranging from 0.01 m x 0.01 m to 20 m x 20 m, to investigate species-area relations of relevance to plant community structure, and to offer a pointer to reserve design. In total, 96 species were found. Of the native vascular species among these, 20% were new records for the reserve.
For the total and the native vascular flora the observed points fell midway between the Arrhenius and Gleason fitted curves. Cryptogams fitted the Arrhenius model well, with z close to the 0.26 value expected for isolates under Preston's Canonical hypothesis. Extrapolation of the Arrhenius curve to the whole of the New Zealand uplands gave values far too large; the Gleason curve gave values much too low.
A General Root model is introduced. It fits the observed species richnesses at various quadrat sizes considerably better than previous models. When biogeographic limitations on species richness are included, extrapolation of the General Root curve accurately predicts the size of the upland flora. This fit, the lack of an asymptote, and the rarity structure, are compatible with a random and individualistic model of community structure. However, there are differences in species-area relation between vascular and cryptogamic plants, which cautions against expecting any universal type of community structure.
Extrapolations using the General Root model suggest that if the ideal is a 10 km x 10 km reserve, a reserve one tenth that size would contain 81% of the native species in that ideal, and the present Blackrock Reserve contains 67%.