Diagnosis and predictability of east african rainfall on intraseasonal to interannual timescales
Abstract
Almost all facets of societal and economic activities in East Africa are critically dependent on
the variability of seasonal rainfall which mostly occurs during March-May (long rains) and
September/October-December (short rains). However, the societies are often unprepared to adjust
quickly to dramatic deviations from normal rainfall regimes (both seasonal total and the frequency
and intensity of extended wet/dry spells within the season) and valuable resources are often wasted.
The fundamental goal of the present study is to understand the mechanisms that govern the
intraseasonal and interannual rainfall variability and hence improve existing climate monitoring and
forecasting in East Africa.
It is known that there is a good degree of predictability in the large-scale area-average
seasonal mean rainfall total, in particular during October-December (OND). This research first
assesses the representativeness of the predictability of the large-scale rainfall at smaller spatial scales
within East Africa. In addition, the validity of using satellite-observed outgoing longwave radiation
(OLR) anomalies as proxy for rainfall anomalies is evaluated. Next, diagnostic analysis is made of the
relationship between East African rainfall and ocean-atmosphere structures associated with El Nino -
Southern Oscillation (ENSO) and non-ENSO variability, with a view to establishing the physical basis
for remote teleconnections with sea surface temperature (SST) and therefore improving reliability and
confidence in SST -based prediction schemes for East Africa. Having defined the teleconnection
structures for the seasonal mean, this study then takes a first look at the role of extended wet spells
over East Africa in the October-November rainfall anomalies and associated teleconnection structures
which in addition to enhancing understanding, sheds light on the potential for anticipating intraseasonal
rainfall events. Finally, the study looks at the large-scale boundary layer moisture relative to rainfall
variability, and this also leads to a better understanding ofthe evolution of wet spells.
Data utilized for the study include satellite-derived OLR (a proxy for large-scale tropical
convection anomalies), ship observed and satellite-derived SST, and circulation kinematics and
boundary layer moisture inferred from the National Centers for Environmental Prediction - National
Center for Atmospheric Research (NCEP-NCAR) reanalysis. These modem datasets with high
resolutions and complete space-time coverage yielded insightful results with regard to East African
rainfall. In addition, both individual station and grid-box rainfall data (estimated from station
observations) are used. The methods employed include empirical orthogonal function (EOF) analysis,
area-average rainfall anomaly, correlation analysis, composite techniques, multiple linear regression,
vertical integration of boundary layer moisture, and atmosphere-ocean dynamics especially the response
of the tropical atmosphere to heating anomalies.
It is shown that the large-scale rainfall anomaly and associated teleconnection signals are usually
representative of the rainfall anomalies observed over most sub-regions in East Africa [in particular during
the OND season, though less so in March-May (MAM) and September], but with some caution raised for
predictability in certain sub-regions. These spatial variations in skill suggest interaction with orographic
features may modulate the large-scale ENSO and other coupled ocean-atmosphere signals in the region.
Substantial strong relationships were found between the rainfall and OLR anomaly indices over East
Africa except for September, confirming OLR to be a good proxy for the rainfall signal. The OLR signal
for OND is better than for MAM.
In the OND season, ocean-atmosphere teleconnection structures associated with OND East
African rainfall are amongst the strongest found for tropical regions remote from the Pacific. In the
seasonal mean and in each individual month, a sequence of three horseshoe structures are evident in the
tropical convection anomalies (inferred from OLR) for both the SO and East African rainfall host indices.
The anomaly sign is in phase over the central Pacific and East Africa, and out-of-phase over the Maritime
continent. Associated circulation kinematics and SST fields are consistent with the tropical convection
anomalies. These signals are shown more clearly with the new datasets than in previous studies. They give
rise to the positive ENSO association with East African rainfall. The clear SST, OLR and circulation
kinematics signal lead to confidence in the role of equatorial dynamics linking the poles of equatorial
convection, in turn related to the SST forcing fields. In addition, it is shown for the first time that the
atmospheric horseshoe structures in the Indian Ocean are absent in September and through much of the
long rains in MAM, though the three horseshoes are weakly evident again in May. It is suggested that the
presence or absence of this teleconnection structure is related to the state of the background annual cycle.
When the ENSO variance is removed (by linear regression) from the datasets, there emerges more
strongly a positive correlation between East African rainfall in the OND season and enhanced convection
through Equatorial Africa and into the equatorial Atlantic and Amazon region, in turn associated with
warm equatorial and tropical South Atlantic SST. These features had not been seen before and provide
strong evidence for Atlantic SST influence on Equatorial and East Africa rainfall in OND.
The lead-lag structure of intraseasonal teleconnections with East African rainfall suggest that 5-10
days before a rainfall event, near-surface westerly wind anomalies often start to develop in the equatorial
Atlantic and these penetrate across equatorial Africa and into East Africa during the event itself, at which
time anomalies in the Pacific and Indian Oceans, which were strong 15 days before the event, are now
weaker. While the local signal was known before, the large-scale precursor patterns through the Tropics
were not known. This establishes a sequence that can now be monitored to assist anticipation of major
rainfall events. The results suggest that these Atlantic source events tend to be much more common during
the warm SST phase in the equatorial and tropical South Atlantic. Five days after the rainfall event, 200
hPa divergence over East Africa usually pulls off to the east into the Indian Ocean and shows structures
that resemble the Madden-Julian Oscillation (a dominant mode of intraseasonal variability in the Tropics).
For above normal rainfall in East Africa, the seasonal mean teleconnection across the Indian Ocean
resembles this intraseasonal picture with strong convection particularly just off the East African coast,
prompting discussion of the interaction between the intraseasonal and seasonal anomalies.
The [mal activity of the thesis was to study the boundary layer moisture fields in the NCEPNCAR
reanalysis data. These studies proceeded with caution, because the moisture fields are given lower
confidence than wind fields by the creators of the NCEP-NCAR reanalysis. Nonetheless, good signals
were found in the seasonal mean fields: warm events in the equatorial Atlantic coincide with collocated
increases in boundary layer moisture, while warm ENSO events during October-November coincide with
increased moisture in the central/eastern tropical Pacific and western Indian Ocean close to East Africa
(regions of anomalously warm SST). On the intra seasonal timescale, a good local signal was found
with increasing moisture during extended wet spells in East Africa. Some propagation of moisture
also appeared from the Indian and Atlantic Oceans, but the signals were weak and require further
evaluation given the uncertainties in the moisture data.
The study therefore provides insight into rainfall variability over East Africa, in view of
global Tropics ocean-atmosphere climate patterns and underlying mechanisms. These results will
feed into real-time monitoring and forecasting at intraseasonal-to-interannual timescales to enhance
early warning and disaster preparedness activities and minimize the impacts of climate-related
catastrophes that are prevalent in the region.
Citation
Doctor of Philosophy in MeteorologySponsorhip
University of NairobiPublisher
Department of Meteorology University of Nairobi