Detecting water quality improvements following watershed management changes is complicated by flow-dependent concentrations and nonlinear or threshold responses that are difficult to detect with traditional statistical techniques. In this study, we evaluated the long-term trends (1997-2009) in total P (TP) concentrations in the Illinois River of Oklahoma, and some of its major tributaries, using flow-adjusted TP concentrations and regression tree analysis to identify specific calendar dates in which change points in P trends may have occurred. Phosphorus concentrations at all locations were strongly correlated with stream flow. Flow-adjusted TP concentrations increased at all study locations in the late 1990s, but this trend was related to a change in monitoring practices where storm flow samples were specifically targeted after 1998. Flow-adjusted TP concentrations decreased in the two Illinois River sites after 2003. This change coincided with a significant decrease in effluent TP concentrations originating with one of the largest municipal wastewater treatment facilities in the basin. Conversely, flow-adjusted TP concentrations in one tributary increased, but this stream received treated effluent from a wastewater facility where effluent TP did not decrease significantly over the study period. Results of this study demonstrate how long-term trends in stream TP concentrations are difficult to quantify without consistent long-term monitoring strategies and how flow adjustment is likely mandatory for examining these trends. Furthermore, the study demonstrates how detecting changes in long-term water quality data sets requires statistical methods capable of identifying change point and nonlinear responses.