Stochastic precipitation model for the region of southern Africa.

A daily Precipitation Model is developed for the Region of Southern Africa. The model considers the stochastic sequence of wet and dry days in a given month of any year through seasonally (monthly) varying parameters to represent the likely heterogeneous nature of rainfall in the region. Measured daily rainfall data at 760 stations that are distributed over eight countries in the Southern Africa Development Community (SADC) region have been used develop the model. The model is based on the Markov-Chain presentation of dependent wet and dry day sequences, and parametric distributions for modelling wet day rainfall magnitudes. For the later case, four types of distributions namely; Weibull/PWM, Gamma/MoM Exponential/MoM, and Lognormal/MoM were considered. Based on the descriptive and predictive ability tests, the Weibull/PWM and Gamma/MoM distributions are found to be more capable of reproducing the frequency structure (the underlying pdf) of the wet day rainfall with less bias, and then they are selected as the best procedures in modelling the daily precipitation in the region of southern Africa. The application & testing of the precipitation model was carried out based on the comparison of the basic statistics of model-simulated and that of the historical rainfall, in terms of simulating the wet-dry day sequences as well as magnitudes. Comparison between the observed mean annual and that of model simulated rainfall has also been done. The comparisons revealed the model's capability to reproduce the observed daily rainfall at several stations in the region. The spatial applicability of the model was further tested by creating 112° by 1/2° resolution GIS-based database of the model parameters over the region and comparing the basic statistics and mean annual rainfall of the generated series with that of available long-term rainfall from FAO database. In this respect, the result of the statistical tests and comparisons of several stations in the region reveal the potential application of the model to generate rainfall at any selected grid in the region of southern Africa.