| dc.description.abstract | This study aimed at enhancing our understanding of the spatial and temporal
characteristics of wet and dry rainfall spells during the main rainfall seasons over Kenya
namely the long and short rainfall seasons in order to develop tools that can be used to
reduce vulnerability of the agricultural and other rain-dependent sectors to the negative
impacts of extreme rainfall events within the season.
The dataset used in the study was daily rainfall observations from 12 stations
distributed over Kenya and extending within the period 1961 - 2000. Three different
rainfall thresholds were used in the study to define a wet day. These were 1.0mm, 3.0mm
and S.Omm thresholds. Statistical methods were mainly used to attain the overall and
specific objectives of the study.
Results from quality control analysis indicated that the data used in the study was
of good quality and therefore suitable for further climatological analysis. The quality
controlled data formed the major foundation of this study. Results from the frequency
distribution of the wet / dry spells indicated that the I-day wet / dry spells were the most
dominant for all thresholds. In all cases, the frequency of the wet / dry spells decreased as
the length of the spells was increased. Longer wet / dry spells were dominant in the wet /
dry parts of the country. This was however not true during some of the wet years when
some longer wet spells were recorded in some dry parts of the country which are often
known for short duration and heavy episodic rainfall events. It was also noted that if more
and more days were considered for wet / dry spells of a given length M, the conditional
probability of the wet and dry spells were found to decrease in all locations and for all
thresholds.
Results from Markov chain models showed that the first-order Markov chain
model described the occurrence of wet / dry spells quite well. Simulated patterns of the
wet / dry spells from the developed model compared well with the observations
especially over the highlands and close to large water bodies. The results of fitting
gamma distribution to the rainfall amounts indicated that an exponential distribution best
fitted daily rainfall amounts for the 1.0mm threshold. The shape parameter K had a value
close to 0.6 with other thresholds.
The results from trend analyses indicated significant trend in the interannual
patterns of the wet / dry spells and rainfall amounts at some locations during some
seasons. A statistically significant decreasing trend was observed at Nairobi for the
1.0mm wet spells during the long rainfall season when arithmetic mean method was
applied. However no significant trend was notable at all locations and seasons when the
slopes of the regression trend lines were tested for significance at 95% confidence levels.
Results from spectral analysis showed several spectral peaks concentrated around
2 - 4.8 days, 5.2 - 7.7 days, 8.3 - 10.2 days, 11.5 - 15 days, 18.4 - 23 days and 30.7
days. However, the 2 - 4.8 days was the most dominant spectral mode during the two wet
seasons.
Results from the wavelet analysis- showed three major wavelet bands namely; the
less than 10 days, 10- 20 days and 20 - 32 days. Wavelet band of less than 10 days can
be associated with synoptic systems that influence weather over our region, such as
easterly / westerly waves, the tropical cyclones and Inter-Tropical Convergence Zone
(lTCZ).
The 10 - 20 days periodicity or the quasi-biweekly oscillations could be
associated with the westward moving waves, monsoonal wind circulation, while the 20 -
32 days wavelet band can be associated with the Madden-Julian Oscillation (MJO) that
have been linked with the globally eastward moving wave numbers 1 and 2 in the
equatorial regions.
Results from wavelet analysis further showed that the time of occurrence of the
above wavelet bands vary from year to year. These are indications of the inter-annual
variability in the occurrence of the intraseasonal oscillations over our region. Unique
characteristics were discernible during the EI Nifio / Southern Oscillation (ENSO) and
Indian Ocean Dipole (lOD) events. Years classified as dry showed extended wavelet
band thus not depicting any specific periodicity especially over the arid and semi-arid
areas.
In conclusion, the study has delineated some major modes of variability of the wet/ dry spells that could be used to enhance our understanding of temporal variability of wet/ dry spells in Kenya that are critical in the planning and development of all rainfall dependent activities. | en |
