Probabilistic streamflow forecasts of flow variables on monthly to seasonal time scales are often made using an ensemble approach. First, historical weather data are used to simulate streamflow time series (traces) conditioned on the current hydroclimatic state. The ensemble traces are then weighted to produce a forecast of the conditional probability distribution of a streamflow variable. For water resources applications, the quality of forecasts for rare events (low and high flow extremes) is critically important. However, many commonly used approaches for assessing ensemble forecasts focus on the correspondence of the ensemble average with the observations, rather than the quality of probabilistic forecasts of rare events.

In this investigation, a distributions-oriented approach is developed for verification of ensemble forecasts. An advantage of the proposed approach is that it is able to quantify attributes of forecast quality over the entire range of flows, imposing insightful structure on the verification problem. An application of the approach is presented for ensemble streamflow predictions from an experimental system for the Des Moines River basin. Alternative forecasts are made using synthetically generated "skillful" climate forecasts to weight ensemble traces based on the correspondence between the meteological conditions used to generate the trace and the forecast climate conditions (climate-weighting). These alternative forecasts are compared with the forecasts made with no climate forecast information (equal weighting of traces), using the samples of monthly and seasonal streamflow forecasts over a one-year horizon, issued sequentially for each month from 1949 through 1997. The results show that while the conditional distribution forecasts have skill over a range of flows, the forecasts for extreme low and high flows have very little skill. Forecast quality is improved using climate forecast information, but high skill climate forecasts are required to make significant improvements in streamflow forecast quality for the Des Moines River basin with this climate-weighting approach.