Reliable ensemble streamflow prediction (ESP) requires unbiased ensemble precipitation forecasts as input to hydrologic forecast models. Because meteorological forecast information may contain a variety of different kinds of biases, an ESP preprocessing system is used in NWS River Forecast Center operations to remove these biases. The preprocessing techniques have parameters (coefficients and exponents) that must be calibrated. The calibration varies spatially and seasonally throughout an RFC area of responsibility.. In mountainous areas the calibration may vary with each hydrologic sub-basin. A historical meteorological forecast archive of forecasts representative of current operational forecasts is needed to provide data to calibrate preprocessor parameters. Typically the length of available archive is quite limited. This paper analyzes the effect of archive duration on the accuracy of parameter estimates and on the reliability of the adjusted ensemble precipitation forecasts. Strategies to extend the effective length of the historical archive through spatial averaging also is examined.