The relationship between land use and physicochemistry, food resources and macroinvertebrate communities in tributaries of the Taieri River, New Zealand: a hierarchically scaled approach


COLIN C.R. TOWNSEND Department of Zoology, University of Otago, PO Box 56, Dunedin, New Zealand


1. The relationship between land use and stream conditions was investigated, including physicochemistry, the availability of primary food resources and species richness, species composition and trophic structure of stream macroinvertebrate communities. The survey involved eight subcatchments of the Taieri River (New Zealand) encompassing reasonably homogeneous examples of four major land uses: native forest, native tussock grassland, plantations of introduced pine and agricultural pasture.

2. Each land use was represented by two subcatchments, each subcatchment by two to four tributaries, and each tributary by two to three sampling sites. These three sampling scales each represent typical designs for stream community studies. By recording responses at all scales, it can be determined explicitly whether the scale of sampling influences interpretation of community structure.

3. Elevation, riffle length, proportion of large substrata in the bed, total phosphorus and alkalinity were significantly related to land use, as were canopy cover and the relative abundance of leaves and wood in the streams. Principal components analysis of invertebrate density data identified nine orthogonal community types, the distributions of two of which were significantly related to land use. The role played by browsers and shredders in the stream community depended on land use.

4. Primary analysis was at the level of the tributary. When it focused on sites within tributaries, more variables were related to land use and at a higher level of significance. This was largely a result of enhanced statistical power due to increased replication. When whole subcatchments were the focus of attention, statistical power was so low, even with six to eleven subsamples to generate overall means, that few significant patterns could be identified. However, the community patterns that were revealed were similar whatever the scale of sampling.