Intra-seasonal characteristics of daily rainfall over Uganda during the wet seasons
Abstract
Over 80% of the population in Uganda lives in rural areas whose most of
socio-economic activities are rain-fed, yet rainfall displays largest variability in both
space and time which is reflected in either the amounts observed or its distribution or
both. Many applications require good knowledge of both rainfall amounts and the
distribution of wet and dry spells within a season. Therefore proper understanding of
the intraseasonal characteristics of daily rainfall over Uganda could minimise the
negative impacts of such extreme climatic events and take advantage of their positive
impacts and hence reducing climate-related risks for sustainable development of the
country.
The main objective of the study was to investigate the intraseasonal
characteristics of daily rainfall over Uganda during rainfall seasons. Specific
objectives included examination of the frequency, probability and statistical
distribution of wet/dry spells and rainfall amounts; investigation of the existence of
trends and cyclical patterns in the patterns of the wet / dry spells; examination of the
differences in the space-time patterns of the wet/dry spells during evolutions of strong
EI Nino Southern Oscillation (ENSO) and Indian Ocean Dipole (lOD) years, and
determination of the predictability of wet/dry spells over Uganda.
The datasets used in the study included daily-observed rainfall records for 18
stations from 1961 to 2000, Southern Oscillation Index (SOl) and the Indian Ocean
Dipole (IOD) indices from1960 to 2000. The methods used included
frequency/probability analyses; Markov chain models; Gamma distribution models;
graphical; Spearman Rank Correlation; ~pectral and Wavelet analyses; Composite
and Correlation analyses and lastly regression analyses to examine the predictability
of spells.
Results from the frequency distribution of wet/dry spells showed that the
.'"
frequency of occurrence of one-day wet/dry spell was highest at all locations and
reduces exponentially as the length of the spells increases. Wet spells of longer
VI
durations were more common over highlands and close to large water bodies that
normally have more reliable rainfall. The study also indicated that conditional
probability of wet/dry spells decreased from one day to higher spells. Markov Chain
Models results showed that the First-Order Markov Chain model performed better in
fitting the occurrence of wet/ dry spells than the higher order chain models at all
locations and seasons. The study further showed that the gamma model fitted well the
daily rainfall amounts over most parts of the country, although some individual
variability of the day to day peaks could not be well represented at some locations.
Trend analysis results showed that no station in Uganda had statistically
significant trends however at some regions the observed values were close to the
critical limit of 10.351at 95% confidence levels. Spectral analysis results indicated the
presence of a family of spectral peaks that were grouped into 2.0 - 4.5 days; 5.4 - 7.6
days; 8.3 - 10.1 days; 11.4 - 15.2 days; 18.2 - 22.8 days; and 30.3 days. Wavelet
analysis results showed that three major wavelet bands namely; the less than 10 days,
10-20 days and 20-32 days were common in all the rainfall seasons over Uganda
although the variance accounted for and the time of occurrence by the individual
bands varied from season to season at different regions.
Composite analysis results indicated some linkages of the wet/dry spells with
ENSO and laD in all months however the stronger linkages were recorded during
October to December months. There were however significant variations not only
from one location to: another but also from month to month. These differences in
impacts may be attributed to the complex local and regional systems such as the large
water bodies and diverse topography. Results from correlation analysis between laD
and the number of wet spells at various lags showed that the magnitude of the
correlation ranged from 0.35 to o. 60. It was noted that the correlation values
decreased with the I increase in the time lags. High correlation values were
concentrated over Lake Victoria region and south-central western Uganda during
September to December (SOND). Results from the regression analysis indicated that
fitted regression models could not provide realistic estimation of wet/dry spells
variability for most of the months with exception for the sand rainfall season at
some locations.
In general, the study showed that the wet/dry spells displayed different spatial
and temporal characterisrics during each of the rainfall season. These characteristics
are controlled by global and regional systems such as Inter tropical Convergence
Zone (lTCZ), ENSO and laD. The information obtained from this study can be used
in monitoring, building prediction scenarios, proper planning and management of
many weather sensitive activities including agricultural and water resources sectors
that will lead to economic growth and sustainable development of the country.
Citation
Master of Science in 'Meteorology,Publisher
Department of Meteorology, University of Nairobi