Prediction of rainfall in the southern highlands of Tanzania

Abstract

Previous studies on the variability of March to May (MAM) and October to December (OND) rainfall in East Africa have mainly covered the whole region without considering the local climate zones such as the Southern Highlands of Tanzania. Results from the regional studies have shown that the Indian Ocean Dipole (IOD) and Outgoing Longwave Radiation (OLR) have strong association with OND rainfall and weak one with MAM rainfall in East Africa. The present study was aimed at determining the potential predictors for the MAM and OND rainfall by considering IOD, El Ni o Southern Oscillation (ENSO), Quasi-biennial Oscillation (QBOu30 and QBOu50) and OLR. . Results suggest that rainfall in Southern Highlands of Tanzania is associated with IOD, ENSO, QBO and OLR. It is found through a Principal Component Regression that IOD, OLR and QBOu30 are potential predictors of both MAM and OND rainfalls. QBOu50 is a potential predictor for MAM and ENSO for OND rainfall, but as previous studies showed, ENSO is a weak predictor for MAM rainfall.QBOu50 is also a weak predictor for OND rainfall. It is therefore recommended that IOD, QBou30, OLR, QBOu50 and ENSO should be considered in rainfall forecasting in the Southern Highlands of Tanzania.

Key words: Collinearity, Multicollinearity, PCR model, Rainfall Variability, Southern Highlands of Tanzania