Stream macroinvertebrate assemblages are diversely influenced by the upstream and overland arrangement of forest within a stream catchment. Past studies have shown near-stream forest cover exerts greater effect on in-stream macroinvertebrate assemblages than more distant forest. In the greater Melbourne region, the availability of more accurate and extensive stream network maps, finer-grained urban surface cover data, and joint-taxon modelling approaches permits review of past studies by evaluating the effect of improved data and methods on past conclusions (Walsh & Webb, 2014). We fitted hierarchical linear models to joint macroinvertebrate species presence-absence and to Stream Invertebrate Grade Number Average Level (SIGNAL) in response to 7 environmental variables including exponentially-weighted forest cover. Using first a new high-resolution stream network, and then the network pruned to a similar density to a past study in the same region, we identified the optimal weighted forest cover by comparing the plausibility of models with varying half-decay distance (HDD) for upstream flow (W) and overland flow away from the stream (L). The most plausible L using the full network was 5-15 m (narrower than the 35 m determined in the past study) and W was 400-1200 m (consistent with optimal W of 1000 m determined in the past study). Optimal L using the pruned network was more consistent with the past study (10-35 m). The narrower optimal L determined in this study suggests that the extent of stream network maps can influence the inferred effects of overland forest cover on stream ecosystems. This highlights that narrower riparian buffers (than previously inferred) can protect in-stream biota from land-clearance impacts, while emphasizing the importance of buffers along small tributaries and headwater streams that were unmapped in the earlier study.