Predictability of east African seasonal rainfall with sea surface temperature gradient modes

Abstract

The overall objective of this study is to improve the skills of seasonal rainfall prediction in East Africa through the use of sea surface temperature gradient modes. The sea surface temperature data used to generate the sea surface temperature gradient modes were for the period 1960-2006. The sea surface temperature gradient modes were generated for the Indian, Atlantic and Pacific oceans separately, and for the Indian and Atlantic oceans combined. The study assumed that the combined Indian and Atlantic Ocean sea surface temperature modes could provide other unique predictors that cannot be picked by the Principal Component Analysis modes of the individual oceans. Other data used in the study included monthly station rainfall,· global wind and satellite derived outgoing long-wave radiation for the same period 1960-2006. The satellite derived outgoing long-wave radiation data is, however, only available from 1974-2004. The data were quality controlled and normalized using standard methods before they were used in the study. The methods used in the study included Correlation Analysis, Composite Analysis, Regression Analysis, Canonical Correlation Analysis and Artificial Neural Networks.The sea surface temperature gradient predictors were searched from all standard seasons of the year, namely December-February, March-May, June-August, and September-November, since the best predictor is that associated with sea surface temperature variability that leads seasonal rainfall. Rainfall prediction studies were, however, restricted to the main rainfall seasons of the region that are concentrated within March-May and September-December months. Quality control analyses indicated that most of the data used in the study were of acceptable quality. The time series analyses of the standardized data indicated inter-annual recurrences of large and low values of anomalies in all records. The large SST anomalies are evident during the EI Nino/Southern Oscillation and Indian Ocean dipole years. The extreme high and low rainfall values, on the other hand, reflected the major flood and drought years. The results from Principal component Analysis of sea surface temperatures indicated that both the Atlantic and Indian Oceans have significant influence on regional rainfall. The first four Principal Component Analysis modes of variability that formed major sources of influence of sea surface temperature on regional rainfall represented ocean wide warming, and zonal and meridional sea surface temperature variability associated with El