During migration, whooping cranes (Grus americana) roost along the central Platte River in Nebraska, selecting for wide, unobstructed channels free of tall, dense vegetation. Reductions in river flow from water development caused channel incision and vegetation encroachment throughout the central Platte starting in the mid-1900s. As a result, a stable yet narrowed river system was observed by the early 2000s. Narrowed channels greatly decreased the availability of roosting habitat. Long duration (>40-days), high peak flows averaging 8000 cfs in 2011 and then 13000 cfs in 2015 widened narrowed river channels, scouring vegetation and eroding sand bars and channel banks. Since 2015, widened channels have persisted in the absence of peak flows. We hypothesize that moderate flows in June (averaging at least 1500 cfs) that inundate most of the active channel during the early growing season, likely suppress germination and reduce encroachment of vegetation including cottonwoods (Populus spp.) and willows (Salix spp.) into the channel. As managers of central Platte land and water to provide suitable whooping crane habitat, our objective was to test whether annual channel-inundating maintenance flows could be used to maintain wide unobstructed channels in the absence of high peak flows. Prior to costly implementation, we developed a machine learning Random Forest model using historical data to predict annual changes in unobstructed channel width in response to flow and mechanical management. Flow metrics associated with channel widening (maximum 40 day mean peak discharge) and maintenance (average June discharge) were included along with area of active channel disked and treated with herbicide each year to explain annual channel width changes. Model results were used to develop water management plans for upstream annual maintenance flow releases of 1500-2000 cfs during a 30-day period in June-July starting in 2020. Following four years of implementation, channel width dynamics will be evaluated to determine effectiveness of inundation flow releases for maintaining wide, unobstructed channels for whooping cranes and the value of the model for predicting outcomes of alternative management strategies.