Characteristics of the wet and dry spells during the wet seasons over Kenya

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.