Prediction of Ethiopian seasonal rainfall using enso indices

Abstract

Water is important for human life. Most of the water comes from rain. Hence, precise prediction of rainfall is necessary. The objective of this study was to develop a statistical model for forecasting seasonal rainfall over Ethiopia. .' . Monthly rainfall records at 37 stations spanning within the years 1968 - 1997 were used in the study. Quality control tests were used to test consistency and homogeneity of the data set. Some stations were found to have missing data and they were estimated using the arithmetic mean method. The spatial and temporal distribution of Ethiopia rainfall was examined using principal component analysis (peA) and time series analysis. The result shows rainfall is highly variable in time and space. The Rotated Principal component analysis (RPCA) was used to delineate Ethiopia into homogeneous climatological rainfall zones. The result showed that there are 12 zones for short (Belg) rainy season and 14 zones for long (Kiremt) rainy season .The station with highest communality was used as representative for each zone. Correlation analysis indicated the significant of correlation between regional rainfall and SST over some specific ocean regions. These formed the fundamental base for the predictions, which were used in this study. The results from the study further showed that SOl has a sig-iificant lag correlation with seasonal rainfall over some zones. 1\' Stepwise regression methods obtained the best regression .equation between the predictand (rainfall) and the various predictors. The period 1968 - 1992 was used for .model calibration while the period 1993 - 1997 was used for testing the skills of the developed model or model verification. While the SOl was dropped in the stepwise regression, The SST over the El-Nino regions featured in the model, indicating the .'. important role played by the ENSO in the interannual variability of Ethiopian rainfall. It was further noted that predicting the short rainy season using January SST has a better skill than ND] SST. The MAM SST predictors were found to have a good skill in forecasting the Kiremt (long) rainy season. The May SST predictors had a better skill for some zones.