Evaluation of Ensemble Predictions of Blocking in the NCEP Global Spectral Model

Abstract

Ensemble forecasts from the National Centers for Environmental Prediction (NCEP) Global Spectral Model (GSM) have been used to develop a probabilistic scheme for the prediction of blocking over the Northern Hemisphere. An evaluation of these forecasts during three recent cool seasons revealed that they underpredicted the frequency of blocking at all ranges considered. Probabilistic forecasts for blocking over a sector on a particular day were constructed via binary logistic regression from all ensemble forecasts verifying on that day. Forecasts from two cool seasons served as the developmental dataset; the probabilistic forecasts generated from these data were tested on data from a third, independent cool season. The resulting calibrated (bias corrected) ensemble forecasts were compared to the climatology of blocking derived from the 40-yr NCEP?NCAR reanalysis data, and to the probability of blocking expected from the uncalibrated ensemble forecasts. The calibrated forecasts had higher skill, relative to climatology, over the Atlantic than over the Pacific. The calibrated forecasts also produced more accurate predictions of blocking than did the uncalibrated forecasts over the Atlantic, but not over the Pacific. The results demonstrate that the ensemble forecasts, when calibrated, can provide useful predictions of blocking at extended ranges.