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
Citation
Doctor of Philosophy in Meteorology,Sponsorhip
University of NairobiPublisher
Department of Meteorology, University of Nairobi