1. Restoration of riparian forests has been promoted as a means of mitigating urban impacts on stream ecosystems. However, conventional urban stormwater drainage may diminish the beneficial effect of riparian forests. 2. The relative effects of riparian deforestation and catchment urbanisation on stream ecosystems have rarely been discriminated because urban land use and riparian degradation usually covary. However, land use at three scales (channel canopy cover along a 100-m site, riparian forest cover within 200 m of the channel for 1 km upstream, and catchment imperviousness) covaried only weakly along the lowland Yarra River, Victoria, Australia. 3. We tested the extent to which each land use measure explained macroinvertebrate assemblage composition on woody debris and in the sediments of pools or runs in the mainstem Yarra River in autumn and spring 1998. 4. Assemblage composition in both habitats and in both seasons was most strongly correlated with proportion of catchment covered by impervious surfaces. Sites with higher imperviousness had fewer sensitive taxa (those having a strong positive influence on indicators of biological integrity) and more taxa typical of degraded urban streams. Sensitive taxa rarely occurred in sites with >4% total imperviousness. However, within sites of similar imperviousness, those with more riparian forest cover had more dipteran taxa. Channel canopy cover did not explain assemblage composition strongly. 5. Riparian forest cover may influence richness of some macroinvertebrate taxa, but catchment urbanisation probably has a stronger effect on sensitive taxa. In catchments with even a small amount of conventionally drained urban land, riparian revegetation is unlikely to have an effect on indicators of stream biological integrity. Reducing the impacts of catchment urbanisation through dispersed, low-impact drainage schemes is likely to be more effective.