Measures of stream flow traditionally have been based on averages and the extreme levels during floods and periods of lowest flow. For many streams with widely variant levels of flow, these statistics may be unrepresentative of the flows at most other times. To correct for this problem, hydrologists, geographers, and other scholars concerned about stream flows have developed probabilistic models to describe the entire range of flow levels for a streams. This doctoral dissertation research project will extend these types of models through development of a flow-duration curve to represent the cumulative volume of discharge in a stream. By refining this procedure, the distribution of flows at any level will be estimated, and procedures for estimating flow regimes at ungauged sites along streams will be developed. Following refinement, the models will be tested using data on daily stream flows for rivers in British Columbia in western Canada. Development and demonstration of probabilistic stream flow models that permit assessment of flow distributions at all levels will provide several significant benefits. From a scholarly standpoint, enhancement of the models will permit more accurate representations of continual stream flow, thereby providing a more complete empirical basis for inquiries into patterns and processes of hydrologic variability. Practical benefits from this project also will be realized, because the refined models will provide more reliable estimates of the frequency of flows at any level. These estimates will aid in making decisions about flood control, water management, and other public policy issues.