Rivers in the Great Lakes basin are highly fragmented due to the presence of thousands of in-stream barriers (dams and road-stream crossings). The removal or modification of barriers can restore migratory pathways for river-spawning fishes, but the costs (financial, species invasions) and benefits (access to breeding habitats) differ among potential mitigation projects. We are undertaking a three-phase project with the goal of providing a transparent method for comparing these costs and benefits to assess which barrier removal projects would offer the greatest return on investment. First, we collated several existing barrier databases to create a single, comprehensive database for the Great Lakes basin. Second, we used field surveys (n=1088 barriers) of two components of barrier passability (vertical drop, water velocity) to create a statistical model predicting probability of fish passage for each barrier in the basin. Third, we developed mathematical optimization models to determine the most efficient barrier repair/removal strategies to maximize the amount of available breeding habitat. Our database of the location of dams (n=7,091) and road-stream crossings (n=268,818) is the most comprehensive inventory to date for the GL basin. Barrier passability was positively related to upstream drainage area and negatively related to stream channel slope. Across the Great Lakes basin, spatial patterns in drainage area and slope drive several characteristic patterns in barrier passability. Barrier removal strategies generated by the optimization models are characterized by a nonlinear relationship between budget and return on investment (i.e. access to breeding habitats). We will discuss key factors that drive barrier prioritization, future data needs, and the strengths and limitations of an optimization-based approach to river restoration planning.