Assessing The Predictability Of Rainfall In Southern Africa With Focus On Eswatini Using Global Ocean And Atmosphere Indicators

Abstract

Southern Africa and Eswatini in particular rely heavily on rainfall for agricultural production which supports socio-economic activities. Thus, skilful and timely rainfall predictions are very imperative, since it is only the prediction information that can be used to trigger actions that lead to reduction of climate related hazards. The predictive potential of rainfall in southern Africa and more especially in Eswatini using dominant ocean and atmosphere indicators was established while seeking to investigate the potential for predicting rainfall in Eswatini with increased time leads. The study was motivated by the significant role of rainfall variability in influencing social and economic activities in the sub-region and the need to avail skilful climate forecasts in advance of the target season. Rainfall data from stations in Eswatini and gridded rainfall data from the Global Precipitation and Climatology Centre were used in the analysis, together with global wind, mean sea level pressure and sea surface temperature data sets drawn from the NCEP/NCAR reanalysis. The study used principal component analysis (PCA) or empirical orthogonal functions (EOFs) as the main analysis method while principal component regression was used for developing statistical rainfall prediction models. The ENSO phenomenon was found to cause the second greatest variability in the global SSTs after the general warming of the global oceans which had the largest variance. A notable outcome from the study is that circulation at upper levels of the atmosphere have an important contribution to the predictability of rainfall in Eswatini. Using the station rainfall data, models for predicting December-January-February (DJF) rainfall anomalies were developed and a lead time of 3 months using June-July-August (JJA) predictors was found to be viable with good and useful results.