J3.3
Calibration of QPF/PQPF forecasts based on the NCEP global ensemble
Yuejian Zhu, NOAA/NWS/NCEP, Camp Springs, MD; and T. Zoltan
NCEP/GFS model Quantitative Precipitation Forecasts (QPF) have been generated daily for up to 16 days. Ensemble based Probabilistic Quantitative Precipitation Forecasts (PQPF) have been operationally created on a daily basis at NCEP since 1997. The PQPF forecasts are based on the relative frequency of ensemble members indicating a particular event. Since neither the numerical weather prediction model nor the ensemble generation technique is perfect, these QPF and probability values are not necessarily reliable. To ameliorate this problem, a calibration algorithm to reduce the bias in the first moment of the ensemble-based probability distribution has been developed. The calibration is based on a comparison of the frequency distributions of observed and forecast precipitation amounts over a preceding 30-day period over the continental US (CONUS). Verification results indicate that the reliability and skill of the calibrated forecasts is significantly enhanced. Further improvement is expected from an additional adjustment where the second moment of the ensemble is calibrated. The QPF/PQPF calibration procedure has been operationally implemented in NCEP by May 4th 2004, which expect to provide reliable and more skillful QPF/PQPF forecasts for the hydrologic and other user communities. .
Joint Session 3, Probabilistic hydrometeorological forecasting and acceptable uncertainty (Joint between the Limited Water Supply Symposium and the 19th Confernce on Hydrology) (parallel with Session 1 and Joint Session 4)
Monday, 10 January 2005, 1:30 PM-2:30 PM
Previous paper Next paper