| description abstract | This study follows a recent paper on the predictability of low-frequency modes on the time scale of 30?50 days. By filtering out the high-frequency modes, we are able to delay the contamination of low-frequency modes for periods of the order of 1 month in global forecasts. A multilevel global model forecast is carried out to predict a wet spell over central China. It is shown that an initial state consisting of time-mean state, a low-frequency mode, and a specification of the sea surface temperature anomaly provides useful forecasts for the occurrence of dry or wet spells. All of these tests are carried out with a global model; however, only the monsoon region is examined in some detail. This study includes the results of a number of experiments where the sensitivity to the definitions of the mean state, the sea surface temperatures, and initial datasets are explored. The main finding of this paper is that the prediction of monsoonal low-frequency modes and the related dry and wet spells can be extended beyond the usual numerical weather prediction (NWP) predictability limit of 6 or 7 days. It appears that if the contamination from high-frequency modes is suppressed by an initial filtering, then the prediction of low-frequency motion through one cycle, a period of roughly a month, is possible. The forecasts are shown to be quite sensitive to the definition of the initial time-mean state and the sea surface temperature anomaly. | |
