Medium-Range Forecast Skill Variation and Blocking Transition. A Case Study

Abstract

Temporal variability in skill of operational medium-range forecasts in the winter of 1988/89 is examined. The primary concern is to examine the performance of the Japan Meteorological Agency (JMA) global spectral model, which began daily 8-day forecasts in 1988. The model's forecast skill exhibits considerable low-frequency (∼a week or longer) temporal variability. During the period under study, root-mean-square error (rmse) of 500-Mb height showed a pronounced temporal maximum at the end of January 1989, when the atmosphere over the North Pacific underwent a remarkable transition from zonal to blocked circulation. The consecutive forecasts during this period showed large dispersion among one another. A linear measure of forecast spread, or phase-space divergence of atmospheric trajectories, is also evaluated after Lorenz using a nondivergent barotropic model. It shows a temporal maximum concurrent with rmse. Therefore, it is likely that the zonal-to-blocking transition was associated with higher-than-average instability of the atmosphere. This impression is strengthened by the fact that the models at two other centers, the European Centre for Medium- Range Weather Forecasts (ECMWF) and the U.S. National Meteorological Center (NMC), showed similar features. The forecast dispersion prior to the blocking is studied in detail. A set of forecasts initialized every 6 h is used to complement the operational dataset. The principal spatial pattern of the spread showed sensitivity over the blocking high. In addition, it is noted that all three models examined show considerable “reluctance” to enter the blocked regime unless they are initialized 5 days or less prior to the blocking. Once the block is established in the model, however, it tends to persist.