The Lower Mississippi River Forecast Center has been issuing ensemble streamflow predictions for several of its forecast points for over three years as part of the National Weather Service's Advanced Hydrologic Prediction Service. There have been many questions raised concerning the accuracy and validity of these forecasts since no real time verification has been computed for these ensemble forecasts. The purpose of this study was to address those accuracy and validity questions.

Monthly ensemble forecasts of maximum 90 day river stages for 2004-2006 were compared to observations to determine the improvement of the ensemble forecasts over climatology. Various verification methods were applied including Ranked Probability Skill Score, Talagrand and reliability diagrams. The Ranked Probability Skill Scores showed that the streamflow forecasts were a significant improvement over climatology. Mixed results were found based on the Talagrand and reliability diagrams.

The ensemble streamflow forecasts were also compared to the percentage of normal rainfall for the same period. Although rainfall was below normal, the maximum observed stages generally exceeded the 50th percentile forecasts. That suggests the rainfall intensities were above normal, even though the rainfall totals were below normal.