Analysis of some seasonal rainfall characteristics in the Lake Victoria region of Kenya

Abstract

Unlike any other part of Kenya, the Lake Victoria Drainage Basin is a complicated area both in terms of its geography and weather systems. The interplay of geography and the meso-scale lake-induced circulation systems, in the presence of the lower tropospheric flows, gives rise to seasonal rainfall regimes which have previously been variously defined. Knowledge of the seasonal rainfall regimes that occur in the area and some aspects of the characteristics of rainfall in each seasonal rainfall regime is important for water resources planning in general and for agriculture in particular. This study was an attempt to define and delimit the seasonal rainfall regimes of the Kenyan part of the Lake Victoria Drainage Basin. On the basls of the identified seasonal rainfall regimes, a few sample stations were selected and whose data were used to study the frequency of occurrence and distribution of wet and dry spells of various lengths. The seasonal rainfall regimes were determined from a 67 x 67 inter-station correlation matrix whlch was based on the lO-day median rainfall of each station. The correlation matrix was subjected to Principal Components Analysis (PCA) the outcome of which was the extraction of 4 components (based on Kaiser's criterion of an eigenvalue of 1.0) which explained 89.2% of the system variance. Only the first two components showed spatial coherence and meaning; the first and second Varimax rotated principal components explained 50.8% ~nd 33.1% of the total variance. This led to the identifl.cation and delineation of the "Kibos" and "Eldoret" seasonal rainfall re~imes. At the boundaries of each regime, there were stations that had almost equal loadings on the first and second Principal Components ~or seasonal rainfall regimes) These were considered transitional areas.

An examination of the annual march of the 10-day or "decade" rainfall for selected stations in each seasonal rainfall regime revealed that the "Kibos" regime is mainly bimodal, the rainy periods being March-May and October-December; for some stations, there is a third but relatively unimportant peak in August. The "Eldoret regime" has two rainfall peaks falling within a single rainy season which extends from April to September. The transitional zones around Kakamega, Nandi and Kericho have rainfall almost throughout the year. For each seasonal rainfall regime, as well as for the transitional zones, a few stations were selected for the study of wet and dry spells. Two criteria were used to define a wet (and also dry) day i.e. 1.Omm and S.Omm. The latter limit was chosen because of its agricultural importance given that potential evapora-

tion in the study area is in the order of 4-6mm day . For each station, for the common 30-year period (1941- 1970), a count was made with the aid of a computer programme, of the wet and dry spells of various lengths. From these data, it was found that the frequency of both wet and dry spells fell off exponentially from the shortest duration to the longest. Wet and dry spells of short duration were more frequent in the "Kibos" regime. The data further showed that wet spells of longer durations were more frequent around Kericho and Tambach. For the 1.Omm criterion, the dry spells were longer than the wet spells except at Kakamega, Jamji Estate, Kericho and Sotik. These data help to emphasize the important nifferences between the main seasonal rainfall regimes. Near the lake, rainfall tends to fall in storms lasting anything from some minutes to a few days, while, particularly in the "Eldoret" regime and in the transitional areas, rainfall tends to persist for longer durations. - vii - For some stations, the wet spells data were well described by both the Markov chain and Exponential distribution models according to the X2 goodness-of-fit test. Only in one case did the observed dry spells data fit the Exponential distribution model well: in all other cases, the fit of the dry spells data to the two models was poor. From the cumulative frequencies of wet and dry spells, the percentage of times a period of a given type of weather (wet or dry) was followed by the same type of weather was evaluated. It was found that the probability of a type of weather recurring is higher the longer that type of weather has persisted. The Principal Components Analysis (PCA) method clearly brought out the seasonal rainfall regimes existing in the study area. The delineation of the different seasonal rainfall regimes forms a basis for future planning of a more representative agro- and' hydro-climatological network in the region. From the study of the wet and dry spells data, the spatial contrasts in the pattern of rainfall occurrence between the different rainfall regions were shown. This information is important in any consideration of water resources utilization in general, and in agriculture in particular.