Numerical simulation of the influence of the sea surface temperature anomalies on the east African seasonal :rainfall
Abstract
In this study, the relationship between the Sea Surface Temperature (SST) anomaly
patterns and the interannual variability of both the Long Rains (March to May) and the
Short Rains (October to December) over East Africa was examined. The study was carried
out on both monthly and seasonal time-scales.
The statistical methods used to establish the rainfall-SST relationships included
Empirical Orthogonal Functions (EOFs), simple correlations and composite analyses. The
physical processes through which the SST anomalies may influence the interannual rainfall
variability over East Africa were investigated using numerical simulation. The data sets
used in the Study included rainfall, SST, surface wind over the oceans and Sea Level
Pressure (SLP).
The results from the analyses of the Long Rains season records indicated low intermonthly
correlations. The Short Rains on the other hand were more spatially and
temporally coherent. This suggested a higher (low) degree of persistence in the factors that
influence rainfall variations during the Short Rains (Long Rains) season.
The Short Rains indicated significant positive correlation with the SST over the
Arabian sea, central and east tropical Pacific. The rainfall during each of the Long Rains
months was correlated with different Oceanic regions. The degree of association between
the rainfall variations and the ·SST was also different for each month. The strongest
(weakest) rainfall-SST relationship was observed in May (April). In March, the rainfall
over most parts of East Africa was negatively correlated with the SST over the eastern
Pacific, central Pacific and southern Atlantic. In April, the rainfall over the northern parts
were
negatively correlated with the SST over east Pacific and the south west
Indian Ocean. In May, the western regions were negatively correlated with the SST over
south eastern Atlantic Ocean while the coastal region was positively correlated with the
SST over the north western Indian Ocean.
The first global SST EOF mode represented the general global warming/cooling
pattern. This mode of variation was not significantly correlated with the rainfall over East
Africa. The second and third global SST EOF modes were linked with the El-Nifio
Southern Oscillation (ENSO) variations.
The Short Rains over most parts of East Africa was positively correlated with both
the second and third global SST modes. The highest correlation values were, however,
concentrated over the western and coastal regions. During the Long Rainy season the
regions located over the northern (southern) part of East Africa were positively
(negatively) correlated with the third global SST mode.
The first SST EOF mode for both the October-December and March-May Indian
Ocean SST represented the general warming/ cooling over the Ocean. An east-west SST
anomaly pattern was observed during both rainfall seasons. Most of the East Africa regions
were significantly correlated with this east-west pattern during the Short rains season.
During the Long Rains season, however, only the western and coastal regions were
significantly correlated with this pattern.
The patterns of the composites of the near-surface ocean-atmosphere variables
were 'stronger for the extreme wet cases when compared to the extreme dry cases during
the Short Rains season. During the Long Rains season the patterns for the 'dry' composites
were more well defined as compared to the 'wet' composites. These results were not
surprising since the floods (droughts) are more (less) wide spread during the Short Rains
season. The reverse scenario is true for the Long Rains season.
A version of the U.K. l l=level GCM ( General Circulation Model) was used to study the
sensitivity of East Africa seasonal rainfall to both global and regional SST anomalies. The
model simulations replicated fairly well the observed climatology over eastern Africa. The
observed and simulated climatology however differed in fine detail due to the inability of
the GCM model to resolve the meso-scale features which substantially contribute to the
observed spatial rainfall distribution over East Africa. The skill of simulation was higher
over the northern part of East Africa (between 5°N and 50S) compared to the
southern part of East Africa (south of 50S). Also, the simulation skill tended to be better
during the Short Rains season. This was attributed to the high degree of correlations,
between the rainfall and the SST during the Short Rains season as was evident from the
empirical studies.
During the Long Rains season, the higher (lower) skill of simulation was observed
in May (April). Again, this was attributed to the higher (lower) degree of association
between the May (April) rainfall variations and the global SST patterns during the March
to May rainfall season.
The use of regional SST anomalies generally had no significant skill in the simulation
of the East Africa rainfall. The effect of superpositioning of the regional SST anomalies
was observed to be nonlinear, which suggested a complex interaction between the
influence of the SST anomalies over various Oceanic regions.
The major feature associated with the above (below) normal rainfall over East Africa
was identified as an anomalous low level wind
convergence (divergence) over the western Indian Ocean. This feature was
associated with the variations in both the east-west Walker and the
local Hadley circulations. The fluctuations in the Walker circulation were linked with the
ENSO mode of variation in the SST. The east-west circulation was stronger during the
Short Rains season as compared to the Long Rains season which explained the observed
stronger link between the ENSO and the Short Rains. The southern local Hadley
circulation showed a higher degree of variability when compared to the northern cell.
The impact of the variation in-the southern Hadley cell was higher during the Long Rains
season when much of the rainfall was associated with the south easterlies.
It was concluded from the study that significant linkages exist between the
variability of rainfall during both the Short Rains and the Long Rains seasons over East
Africa and the global! regional SST anomaly patterns. Given the high degree of
persistence of the SST anomalies, the results from the study show that SST anomalies have
significant predictability potential for seasonal rainfall over East Africa. The forecasting
skill is, however, higher during the Short Rains season when the rainfall-SST relationship
is stronger.
The results from the present study can therefore be used to develop models for
forecasting seasonal rainfall anomalies for the East African states. The forecast of the
seasonal rainfall anomalies is important because most of the social and economic activities
over East Africa are rain-dependent. The extreme rainfall anomalies often lead to loss of
life and property. Hence, skillful forecast of the seasonal rainfall anomalies may be used
as an early warning system to help the policy makers and planners to manage and mitigate
the negative socio-economic consequences that are most likely to arise from events like
droughts and floods.