Diagnostic Verification of Regional Hydrologic Ensemble Forecasting Systems

Ensemble streamflow forecasting is a predictive system that makes forecasts in the form of a conditional probability distribution for continuous streamflow. Different weather inputs are used to generate multiple numerical predictions that take into consideration the initial conditions and recent observations. Verification of streamflow forecasts for a region is a complicated process due to form of the forecasts themselves, the large number of forecast products, the large number of locations where forecasts are made, and the small size of the available forecast record for each case. The main challenge is to find an effective way to interpret available data sets for a region in a meaningful way to forecasters and forecast users. In this study experimental forecasts from the North Central River Forecast Center (NCRFC) will be used to assess the forecasting process. The Des Moines River basin and other forecast groups managed by the NCRFC will be used in this study. A data set of approximately 50 years of observed streamflow for the chosen forecast groups will be used to interpret the ensemble forecasting system. Different verification statistical measures are used to assess the predictive process. To aid in the comparison between sites and different forecast product, relative measures are used to assess forecast quality. Forecasts will be analyzed in terms of temporal and spatial variations of the observed streamflow as well as the seasonality of the errors. Graphical representations will be generated to explain which factors influence regional forecast quality.