Hydrologic Re-Calibration Requirements for Ensemble Precipitation Forecasts Following NWP Model Changes

Hydrologic models amplify biases in precipitation forecasts as a result of the nonlinear nature of surface hydrologic processes. This effect is important at all forecast lead times, but at large lead time even small biases may have very large cumulative effects. Biases not only in the mean but also in higher moments of ensemble distributions are of concern. Accordingly, NWP model outputs must be adjusted before being used as input to hydrologic models. These adjustments must be calibrated using archived model forecasts and corresponding observations. When NWP models are changed, the appropriate adjustments change as well. Potentially, a very long forecast archive may be needed to re-calibrate the adjustments. However, if parallel forecasts before and after model changes are sufficiently well correlated, this correlation can be used to reduce the length of period the changed model must be operated to re-calibrate the adjustments need for hydrologic application of the model outputs. The required length of archive to estimate NWP ensemble climatological statistics is illustrated with simulated ensemble forecasts using historical gage data.