Identifying thresholds in flow ecology relationships is a goal of many environmental flows studies and decision making. We compiled landscape-scale, georeferenced species occurrence datasets to examine and compare patterns of fish and macroinvertebrate assemblage and species-level turnover across flow regimes. A gradient forest machine learning approach was used to quantify multi-species threshold responses along hydrologic and watershed-scale disturbance gradients in groundwater, runoff, and intermittent streams in the Ozark and Ouachita Interior Highlands and Gulf Coastal Plains, USA. We identified numerous nonlinear threshold responses where significant changes in assemblage composition occurred along gradients of watershed fragmentation, water withdrawals, timing and frequency of high flows, and other spatial, seasonal, and stochastic properties of stream flow. The results of our analysis provide detailed information on important environmental and disturbance thresholds driving patterns in fish and macroinvertebrate assemblages and species in streams to better manage environmental flows, watershed fragmentation and land use.