| dc.description.abstract | The occurrence of extremes in East African seasonal rainfall cause severe social economic
hardships to the communities and governments. Timely long-range rainfall
prediction and early warning products can be used to mitigate the negative impacts, and also to
take maximum advantage of the positive impacts. Previous studies have shown that the El Nino/Southern Oscillation (ENSO) introduces worldwide signals in the general circulation that
persist for several months and have great potentials as predictors of seasonal rainfall. Most
studies in the region have largely addressed simple linkages between ENSO and seasonal
rainfall. This study investigated more complex rainfall-ENSO linkages with a specific focus
on the impacts of the evolution of the specific phases of ENSO for the specific rainfall seasons,
with a prime goal of improving the long range prediction potential over region.
The specific objectives that were addressed to achieve the overall objective included
diagnosis analyses to enhance the knowledge base on the teleconnections between East Africa
rainfall,and the specific ENSO evolution phases including the Onset, Maturity and Withdrawal
of both cold and warm events, and an investigation of the temporal stability of the derived
ENSO teleconnections. The core of the study involved general circulation model (GCM)
simulations of El Nino and La Nina signals for the specific rainfall seasons in an attempt to
study the physical reality of the ENSO-Seasonal rainfall linkages that were derived from
diagnosis studies. The ECHAM GeM was adopted in the study and the physical reality of the
ECHAM GCM was first investigated by simulating some known seasonal and interannual
climatologies. Attempts were also made in the study to investigate whether the ECHAM model
could capture the major spatial patterns over the region that were derived from Empirical
Orthogonal Functions (EOFs) analysis.
The data used in the study were monthly station rainfall observations, gridded rainfall
data set, global SSTs, and southern oscillation index (SOl) for the 50 years from 1950 to 1999.In order to achieve the overall and specific objectives of the study, the data were
subjected to various analyses including quality control and standardization to enable realistic
time and space comparisons to be made. The standardized rainfall indices were used to identify
all wet and dry events at the specific locations and for both long and short rainfall seasons.
Pearson correlations between rainfall and ENSO indices were then computed for the evolving
ENSO modes at the onset, peak, and withdraw of both warm and cold ENSO events. The
stability of the derived ENSO-rainfall linkages were further investigated using time lagged
correlation techniques that included cross-validation technique. Canonical correlation analysis
(CCA) was also used to study the more complex rainfall-ENSO space and time variability
modes. The core methodology involved the use of ECHAM GCM to investigate the physical
reality of the results that were obtained from the diagnoses studies. The ability of the ECHAM
model to provide realistic simulation of the regional climate dynamics especially the mean
annual patterns, interannual cycles, and spatial rainfall modes that were delineated from
empirical orthogonal functions (EOFs) were first investigated. The model was then used to
simulate the regional rainfall and circulation patterns using some representative El Nino and La
Nina years, and for the specific ENSO onset, peak and withdrawal phases.
Results from quality control tests indicated that the data used in the study were of good
quality. The standardized rainfall data were used to compare year to year rainfall patterns and
also to compute some spatial composite rainfall maps. It was evident from the normalized
rainfall indices and the composite maps that occurrences of droughts and floods are common
over East Africa, and that some of the extreme rainfall event years coincided with certain
phases ofthe ENSO phenomenon.
The study has observed that since the ENSO events tend to be at the onset and decay
phases during March-May, and at maturity during October-December respectively, one would
expect highest degree of linkages between ENSO and seasonal rainfall during the October
November season compared to the March-May season. The two periods correspond to the short
and long rains seasons respectively. It was however observed from both composite and
correlation analyses that during the long rainfall period, the region was dominated by weak
positive ENSO signals during the onset phase of warm events, and a negative correlation with
the maturity, and withdrawal phases of the warm events. It was however noted that during the
maturity phases of cold ENSO event, significantly negative rainfall anomalies are common
over the region, especially over the Inland areas, while the withdrawal of cold phase was
associated with strong wet rainfall anomalies. The potential predictive time lead with the
ENSO phases was about one to two months for the long rainfall season. Case studies using the
1997/98 and 1999/2000 ENSO events showed these types of signals on the long rainfall
season.
Correlation analyses further showed that rainfall during the short rainfall period had
stronger and uniquely opposite signals for warm and cold ENSO events. Onset and maturity
phases of warm ENSO correlated positively with rainfall during the October-December period
and maximum signals occurred at the maturity, in agreement with the floods observed during
warm phase, the best example being the 1997 floods. Decaying phase of warm ENSO showed
low positive correlations, while the onset and maturity phases of cold ENSO showed
significant positive correlations over some regions. From the interannual characteristics of
ENSO-rainfall linkages, the recent decades showed a significant increase in ENSO-impacts on
seasonal rainfall. This was especially evident for the short rainfall season, and may be
explainable by the high frequency occurrence of strong ENSO events during the 1980s and
1990s. The ENSO prediction signals on short rainfall period extend back to the preceding June
to August season, and about 6 months on a month by month basis.
CCA mode 1 results showed that the leading wet mode of long rainfall season was
associated with positive SST weights over Indian, western Pacific and northern Atlantic
oceans. CCA Mode2 of long rainfall period indicated a wetter northern and drier southern subregions
linked with positive SST modes over Indian, central Pacific and Atlantic and it was
also the warm ENSO sensitive mode. The leading CCA mode of short rainfall season was
negative in consistency with negative SST weights over the Indian, central Pacific and eastern
equatorial Atlantic. CCA Mode2 of short rainfall season indicated a wetter northern and a drier southern sector associated with positive SST modes off the East African coast! Arabian sea and
tropical-eastern Pacific. CCA results showed a typical warm ENSO wet-signal in short rainfall
season at mode2 and positive weights over Indian Ocean appeared to enhance the impact. The
CCA analyses indicated that the Indian Ocean dipole SST mode had significant impacts on the
seasonal patterns of rainfall over the region.
The results from the ECHAM GCM simulations indicated that the model provided
realistic simulations of annual and seasonal cycles of rainfall and circulation over the region.
The magnitudes of model rainfall and anomalies were however underestimations of
observations especially during the wet seasons. The results further showed that the simulation
of the interannual variability of the rainfall was more skilful during the short rainfall season
centred on October to December season. This could be due to normal occurrence of peak
ENSO signals during these months. Most of the GCM simulations were therefore concentrated
to this particular season. The short rainfall simulations indicated that although the observed wet
and dry conditions associated with the warm and cold ENSO events were well captured by the
GCM, the circulation patterns and rainfall over the region during the warm ENSO events were
simulated much better than for the cold events. The study also indicated that even for
individual warm ENSO events, there were significant differences in the observed regional
rainfall anomaly signals as was in the case of regional impacts of the 1982/83 and 1997/98 EI
Niiio events. The study revealed that localized regional circulation features could have been
responsible for the less wet 1982 in contrast to torrentially wet 1997. The 1997 event also
exhibited the strongest Indian Ocean dipole when compared to 1982. These results further
signified the important roles that are played by both Indian and Atlantic oceans in the
modulation of the impacts of the global scale climate anomaly signals such as those linked to
ENSO.
Our study has therefore for the first time isolated some unique space-time evolution
patterns of rainfall anomalies that may be associated with the Onset, Maturity, and Decay
phases of both the cold and warm ENSO events with significant time lags for seasonal rain
prediction. The results can be used to enhance the prediction of seasonal rainfall expectations
over the region. Such products are critical in the planning and management of all rainfall
sensitive social and economic activities together with improvement of the existing early
warning and weather driven disaster preparedness efforts in the region | en |
