The low-flow characteristics of thirty-seven gauging stations in North Carolina are analyzed on the basis of duration curves, theoretical probability distributions and techniques of time-series analysis. The sensitivity of flow magnitudes to duration and return period is considered. There is wide variation in the low-flow behavior of individual streams. The conclusion is reached that the Weibull distribution and autoregressive schemes of the first or second order can adequately represent annual minimum flows. A study of weekly flow on four gauging stations reveals that sophisticated, state-of-the-art models are required to preserve the duration distribution of historical flows. The relevance and suitability of the existing water quality data to the emerging needs of water quality management are investigated. The relationships of water quality parameters to stream and flow factors are considered. A possible framework for a conceptual water-quality management model is presented. A realistic model should permit simultaneous consideration of several factors. It is suggested that decomposition might make the model computationally tractable. Simultaneous consideration of flows and temperatures on several streams shows that critical conditions are not necessarily coincident with lowest flows. (Author)