| description abstract | The intraseasonal (40?50 day) oscillation in convection over the tropical Southern Hemisphere (0°?15°S) is examined using two years of ECMWF analyses. The initial period investigated was 1 May 1984?30 April 1986. This diagnosis revealed that the oscillation was essentially absent in the Southern Hemisphere during the winter months. Therefore, the paper focuses on two subperiods, 1 November 1984?30 April 1985 (Year 1) and 1 September 1985?15 April 1996 (Year 2), when the oscillation could be detected. Although several variables were examined, the velocity potential at 200 hPa (?2) and outgoing longwave radiation (OLR) were found to be the best indicators of the oscillatory convective activity; consequently, these variables are the only ones presented. One of the unique features of this study is that the data were not temporally filtered, except for removing the time mean and linear trend, until after it was established that statistically significant peaks occurred on the intraseasonal time scale. This was an important step in this case because the dominant spectral peaks for the oscillation in each year were considerably different. In Year 1 the significant intraseasonal period was between 50 and 67 days, while in Year 2 it was centered near 33 days. Based on this, a recursive bandpass filter of 40?80 days was applied to Year 1 and 27?44 days to Year 2. If the data was temporally filtered at the onset (e.g., 30?60 day band pass), the proper conclusions may not have been reached. For the most part, the findings agree with those of previous investigators. The oscillation propagated eastward, and its convective activity in both years was more intense over the Indian Ocean-Indonesia-western Pacific region than elsewhere. Furthermore, the ?2-wave could be followed continually around the globe, but the convection (OLR) associated with the oscillation was weak and difficult to track over much of the Western Hemisphere. The primary difference between the two years, besides the period of oscillation, was that the correlation between ?2 and OLR was much greater in Year 1. | |
