| description abstract | AbstractIn this study, the contribution of low-frequency (>100 days), Madden?Julian oscillation (MJO), and convectively coupled equatorial wave (CCEW) variability to the skill in predicting convection and winds in the tropics at weeks 1?3 is examined. We use subseasonal forecasts from the Navy Earth System Model (NESM); NCEP Climate Forecast System, version 2 (CFSv2); and ECMWF initialized in boreal summer 1999?2015. A technique for performing wavenumber?frequency filtering on subseasonal forecasts is introduced and applied to these datasets. This approach is better able to isolate regional variations in MJO forecast skill than traditional global MJO indices. Biases in the mean state and in the activity of the MJO and CCEWs are smallest in the ECMWF model. The NESM overestimates cloud cover as well as MJO, equatorial Rossby, and mixed Rossby?gravity/tropical depression activity over the west Pacific. The CFSv2 underestimates convectively coupled Kelvin wave activity. The predictive skill of the models at weeks 1?3 is examined by decomposing the forecasts into wavenumber?frequency signals to determine the modes of variability that contribute to forecast skill. All three models have a similar ability to simulate low-frequency variability but large differences in MJO skill are observed. The skill of the NESM and ECMWF model in simulating MJO-related OLR signals at week 2 is greatest over two regions of high MJO activity, the equatorial Indian Ocean and Maritime Continent, and the east Pacific. The MJO in the CFSv2 is too slow and too weak, which results in lower MJO skill in these regions. | |
