CR 29:111-127 (2005)  -  doi:10.3354/cr029111

ABSTRACT: The aim of this study was to classify the most frequently observed atmospheric circulation anomaly patterns in eastern Africa and the adjacent Indian Ocean. As an example of the usefulness of such a classification, the second objective was to test whether these patterns account for intraseasonal rainfall anomalies in the region. A partitioning algorithm, known as dynamical cluster analysis, was therefore applied to the zonal (U) and meridional (V) components of the wind anomalies, obtained from the NCEP-NCAR Reanalysis R-2 at the pentad (5 d) timescale. The 3 geopotential levels 850, 700 and 200 hPa were combined. Focus is on the transition seasons (March to May and October to December), which correspond to the 2 East African rainy seasons. The classification domain is large enough (40° latitude × 40° longitude) to describe large-scale anomaly patterns, but it focuses on equatorial and subequatorial atmospheric dynamics. The results consist of 7 major wind anomaly clusters, for which associated atmospheric fields (geopotential height anomalies, precipitation anomalies) and the seasonal distribution were explored. Among them, 2 clusters strongly resemble the anomaly patterns associated with the Madden–Julian Oscillation. The other major configurations show similarities with recurrent previously observed patterns involved in wet or dry spells over East or Southern Africa. Statistical tests reveal that this partition is able to significantly discriminate rainfall anomalies over the eastern part of the African continent; 4 of the 7 clusters correspond to abnormally wet or dry events over Equatorial East Africa.

KEY WORDS: Dynamical cluster analysis · Eastern Africa · Intraseasonal rainfall variability · Circulation anomaly patterns · Madden–Julian Oscillation