• boulder additions;
  • channel reconfiguration;
  • effectiveness monitoring;
  • in-stream habitat restoration;
  • large woody debris;
  • macroinvertebrates;
  • meta-analysis

The assumption that restoring physical habitat heterogeneity will increase biodiversity underlies many river restoration projects, despite few tests of the hypothesis. With over 6,000 in-stream habitat enhancement projects implemented in the last decade at a cost exceeding $1 billion, there is a clear need to assess the consistency of responses, as well as factors explaining project performance. We adopted an alternative approach to individual case-studies by applying meta-analysis to quantify macroinvertebrate responses to in-stream habitat restoration. Meta-analysis of 24 separate studies showed that increasing habitat heterogeneity had significant, positive effects on macroinvertebrate richness, although density increases were negligible. Large woody debris additions produced the largest and most consistent responses, whereas responses to boulder additions and channel reconfigurations were positive, yet highly variable. Among all strategies, the strength and consistency of macroinvertebrate responses were related to land use or watershed-scale conditions, but appeared independent of project size, stream size, or recovery time. Overall, the low quality and quantity of pre- and post-project monitoring data reduced the robustness of our meta-analysis. Specifically, the scope and strength of conclusions regarding the ubiquity of macroinvertebrate responses to restoration, as well as the identification of variables controlling project performance was limited. More robust applications of meta-analysis to advance the science and practice of river restoration will require implementing rigorous study designs, including pre- and post-project monitoring replicated at both restored and control sites, collection of abiotic and biotic variables at relevant spatiotemporal scales, and increased reporting of monitoring results in peer-reviewed journals and/or regional databases.