Abstract
In real time since 1990, the National Meteorological Center (NMC) has been running a system to predict the forecast skill of the medium-range forecasts produced by the NMC global spectral model. The predictors used are the agreement of an ensemble consisting of operational forecasts from various centers, the persistence in the forecast, and the amplitude of the anomalies. These predictors are used in a stepwise regression scheme, with the last 60 days used as training period, and the regional anomaly correlation of the 0000 UTC NMC global forecast is predicted from days 1 to 6. By far the most important predictor of skill is the agreement between the NMC global forecast started at 0000 UTC, out to 6 days, and four other 12-h “older” forecasts (Japan Meteorological Agency, United Kingdom Meteorological Office, and the European Centre for Medium-Range Weather Forecasts, as well as the average of the NMC forecast at 0000 UTC with the previous day's forecast). The other predictors have been selected to add to the predictive capability of the agreement alone, and together they quantify the factors that forecasters use subjectively when evaluating the available forecasts. These predictions are available to NMC forecasters on workstations and to outside users through the Internet.
The predictive ability of this system compares favorably with recent theoretical and experimental studies. The correlation between predicted and verifying forecast skill seems to be best in regions where forecast skin varies significantly. The seasonal variation in predicting the skill is small expect in the Tropics. The overall performance shows that these predictors include enough information about forecast skill to justify further development of skill predictions based on large forecast ensembles and on more sophisticated statistical techniques.