The NCEP regional spectral model (RSM) ensemble system was performed to forecast daily precipitation during winter 2002-2003 over the Southwest United States. The equivalent grid spacing is 12 km. Quantitative precipitation forecasts (QPFs) and probabilistic quantitative precipitation forecasts (PQPFs) from eleven ensemble members present large wet biases over the California Nevada River Forecast Center (CNRFC) and the Colorado Basin River Forecast Center (CBRFC). Using a simple linear algorithm, preliminary results show that the ensemble mean of daily precipitation on the catchment scale could be calibrated to reduce root mean square error based on the forecasting in early months. Less improvement of heavy precipitation events indicates that a larger training sample size is desirable.

Longer historical NCEP RSM data (since August 2000) on a nearly 48-km national grid from five ensemble members is used to calibrate QPFs and PQPFs. In order to facilitate hydrological applications of ensemble precipitation forecasting, 6-hour and 24-hour precipitation over hydrologic regions are calibrated for the cool season. An artificial neural network is applied to conduct this post-processing. The NCEP Stage II/IV precipitation analyses are used for verification. The impact of resolution, location, and precipitation thresholds on the calibration processing is discussed.