The predictability of a minor warming and the subsequent major warming in the winter of 2003/04 is examined in comparison with that of the major warming without preceding minor warmings in December 2001 by the use of operational ensemble 1-month forecast data produced by the Japan Meteorological Agency. In the case of the major warming in December 2001 caused by amplified zonal wavenumber 1 planetary waves, the predictable period based on zonal mean temperatures in the polar stratosphere is estimated to be at least 16 days, while the warmings in the winter of 2003/04 are predictable at most 9 days in advance. Such relatively reduced predictability for the latter warmings is considered due to the rather complicated time evolution of the warming episodes with a significant contribution of smallerscale planetary waves during the period prior to the warmings.

Practical predictability of the stratospheric circulation in the boreal winters during the period [2001][2002][2003][2004][2005][2006] has been examined using the archive of 1-month ensemble forecast datasets provided by the Japan Meteorological Agency. To investigate the predictability limit, two measures of the Root-Mean-Square Error (RMSE) and Anomaly Correlation (AC) have been used for the 10-hPa geopotential height field. In the winter stratosphere, an intermittent character of planetary wave activity causes two specific periods of the stratospheric circulation, i.e., an undisturbed phase with inactive planetary waves and a disturbed phase with active ones. Therefore, the predictability needs to be evaluated taking account of this feature. On the basis of careful consideration, the mean predictable period can be estimated to about 10 days during the disturbed phase; it is longer than the tropospheric predictable period of about 7 days. However, it exhibits large variability because of different growth rates of forecast errors caused by contributions of both wave and zonal-mean fields. During the undisturbed phase, the predictable period based on the AC is almost the same as that of the disturbed phase, although the predictable period based on the RMSE is extremely long, since the RMSE predictability is measured against the climatological standard deviation affected by disturbed phases. Therefore, the horizontal pattern of the stratospheric circulation is less predictable even though the forecast error is quite small.

No abstract