Mesoscale distribution patterns of Amazonian understorey herbs in relation to topography, soil and watersheds

1 Many authors have suggested that topography and soils are the major determinants of species distributions and community patterns at small or regional scales, but few studies addressed the patterns at mesoscales. We used Reserva Ducke (100 km(2)) as a model to analyse the effects of soil, topography and watersheds on the variation of the herb community composition, and to determine the relative importance of the environmental factors on species composition. 2 Taxonomic groups are frequently used as surrogates in studies of biodiversity distribution and complementarity, but their efficacy is controversial. We therefore studied the correlations between the distributional patterns of three different herb groups (Marantaceae, pteridophytes and 'others') and their responses to environmental predictors. 3 Terrestrial herbs were sampled in 59 plots of 250 x 2 m, systematically distributed over the reserve. Plots followed isoclines of altitude, to minimize the internal variation of soil. Composition of the total herb community and of the three herb groups was summarized with PCoA. 4 Soil structure, represented by PCA axes, was the main determinant of the variation in herb composition for all groups, but slope affected only pteridophytes. Soil and topography explained less than one-third of the variance in community data. Herb composition was significantly different between watersheds, but watersheds differ only slightly in soil parameters. Our results indicate high turnover in species composition, on spatial scales of 5-10 km in central Amazonia, which is not necessarily associated with soil change. 5 Compositional patterns of the three groups analysed were significantly correlated, but with low values for the correlation coefficient. Although composition was correlated, the responses to environmental predictors differed among groups, and the use of one group as a surrogate will miss around 50% of the variation in other groups. 6 Although important, soil and topography alone cannot predict herb community structure. Knowledge of geographical, historical or other landscape features, such as watershed morphology, may therefore be necessary to predict the turnover patterns over mesoscales. Moreover, the same factors may not have the same effectiveness as predictors of the structure of seemingly similar biological groups.