Determinants of Seasonal Rainfall And Forecast Skills in Semi-arid South-east Kenya

Abstract

Determination of seasonal rainfall predictors with local ramification offers an opportunity to improve climate forecast. This study makes a contribution towards meeting the challenges of more local level studies by identifying sea surface temperatures influencing seasonal rainfall and forecast skill for the main growing season of October-December in southeast Kenya. The study was based on nine key rainfall stations located in different climatic zones in southeast Kenya, and used monthly rainfall for the period 1961-2003. Stepwise regression results show that sites in southeast Kenya are influenced by different SSTs, with the southern oscillation index and the Atlantic Ocean emerging as key seasonal rainfall predictors. Niño1, Niño3.4, Niño3 and Niño4 SSTs subsequently emerged as predictors in the region. Forecast verifications scores generated from Climlab2000 software show a significant association between observed and forecast seasonal rainfall for six out of the nine stations. Stations with a higher hit score skill also showed a significant correlation between observed and predicted rainfall. Sites in semi-arid environment (UM4 and LM5) had the highest skill of predicting dry events, while high altitude zone (LH2) had the highest skill of predicting wet events.