STATISTICAL ANALYSIS OF DAILY RAINFALL OCCURRENCE IN SOME SELECTED CITIES IN NIGERIA

Abstract

In this research, daily rainfall data was obtained from NiMET which spans the period 1971 to 2013 for Gusau, Ikeja and Lokoja stations in Nigeria to investigate the trends in the daily rainfall cycle to assess the implication of the recovery on monsoon season length. The Standardized Precipitation Index (SPI) for 11 year annual running time scales was used to assess the consistency of the rainfall recovery with changes from the drought states toward wet or normal conditions and frequency of daily rainfall occurrences, to determine the distribution of best-fit for probability of daily rainfall occurrence, to estimate the return periods and exceedance and apply statistical method to predict probability of daily rainfall occurrence in the study areas. Mann-Kendall time series trend test was used in the analysis of trend in the seasonal daily rainfall cycle and daily rainfall indices and Sen’s slope values were determined with the corresponding probabilities and trend values. Best distribution fits were tested by using the goodness of fit test of Kolmogorov-Smirnov and chisquare test in maximum likelihood estimation method (MLE) and Kolmogorov was accepted with higher probability for the significance level for the daily rainfall data and Chi-Square test was used in comparing the frequencies for observed and the theoretical data series. Hence, the return periods and exceedance was carried out using the Weibull plotting method matched with the lognormal probability distribution fit in the Microsoft excel package and probability of daily rainfall occurrence prediction was carried out using the model from the return period line fitting. For the comparison/validation test, the observed and predicted daily rainfall values were used to measure the strength of their relationship using line of best-fit in a time series plot. From the different analysis carried out, result showed that there was no trend in the daily seasonal cycle and 11 year annual running mean but an oscillatory shift was observed. Weibull (3) model was accepted as the best-distribution fit for the daily rainfall distribution for the three (3) zones and that the correlation coefficient estimated from the validation test between observed and predicted daily rainfall values gave a near perfect relationship with minimum errors which showed that the Weibull plotting position matched with log-normal trend line comparatively measures well with the observed data for the three zones.