An analysis of the within-season rainfall characteristics and simulation of the daily rainfall in two savanna zones in Ghana

This paper has investigated the within-season rainfall variability at selected sites (Accra and Tamale) in two savanna zones of Ghana. The average duration of the dry and wet spells was estimated for each site from the long-term rainfall probability parameters. At Accra, the average duration of dry spells decreases from March (7 days) to June (2 days) and rises again in July (5 days). At Tamale, the average length of a dry spell during the rainy season is about 3 days. The observed patterns of rainfall at the two sites seem to suggest that Tamale would offer more favourable conditions for rainfed crop production. However, a complete answer to such an important question needs to consider the water balance at each site including the water storage ability of the soils. A rainfall simulation model was also developed for the two sites, and simulated the rainfall patterns and amounts reasonably well at both sites. Simulated number of rainy days in each month of the rainy season for both the short and long term agreed well with observations at Accra ( R 2 = 0.98) and Tamale ( R 2 = 0.95), provided the appropriate short- or long-term model parameters were used. This was also the situation for the simulated monthly rainfall totals (Accra R 2 = 0.97; Tamale R 2 = 0.97). It was, however, obvious that the model parameters derived from the long-term rainfall records could not be used to simulate the short-term rainfall fluctuations which were commonly observed, especially at Accra. In a quest to overcome this deficiency, we sought possible relations between the rainfall parameters and other global factors that affect or modify the local rainfall patterns at the sites. In particular, the study focused on the relationship between the rainfall parameters and the Southern Oscillation Index (SOI). A correlation coefficient as high as 0.78 was established between the Sol and some of the model parameters. Thus, it appears possible to use information from a series of the SOI values to modify the numerical values of the model parameters before their use in the simulation runs. Prospects of using such an approach to improve the simulation of rainfall are discussed.