An improved understanding of interannual and intraseasonal variability of southern African rainfall using NCEP/NCAR Reanalysis data

With the purpose of understanding interannual and intraseasonal variability of rainfall in central Southern Africa, this paper analyses data taken from the latest and most complete Hulme rainguage dataset (1908-1998) over the austral summer (Jan-Feb-Mar) season. Varimax rotated Empirical Orthogonal Functions (EOFs) were created for this period and yielded a small number of stable EOFs that characterised the major components of rainfall variability in central Southern Africa.

The time co-efficients of these leading EOFs were correlated with a range of variables at a global scale derived from the NCEP/NCAR Reanalysis data for the period 1958-1998 including Sea Level Pressure (SLP), Outgoing Longwave Radiation (OLR), Velocity Potential (at a .2101 sigma level), Omega, columnar Precipitable Water and zonal + meridional winds as well as geopotential heights. The time-coefficients were also correlated with data from the latest version of the United Kingdom Meteorological Office Global Ice and Sea Surface (GISST) dataset. A suite of correlation maps were created for both simultaneous relationships and lead/lag relationships upto 12 months in advance (i.e. Feb-Mar-Apr through to Jan-Feb-Mar) in order to quantify the predictability of interannual and intraseasonal rainfall variability. Additionally Canonical Correlation Analysis (CCA) has been undertaken using the rainfall, SST and NCEP/NCAR circulation data.

Initial results confirm links between central Southern African rainfall variability and lower boundary forcing from ocean-atmosphere interaction within the tropics. The correlation analyses also identify the important relationships with the transport of moisture throughout the atmosphere both within the tropics and from the extratropics. As expected the strongest links are with the Indian and Pacific Ocean basins however links elsewhere across the globe were also uncovered.

It is suggested that this increased understanding of the physical forcing mechanisms will allow improved forecast skill in seasonal forecasts of Southern African rainfall variability with lead times of upto 12 months. This is borne out by the significance of the lead/lag correlations presented here.