Statistical Estimation of Onset, Length and Cessation of Rainfall in Selected Locations in South West, Nigeria

Abstract

Without quantitative data, agro-meteorological planning, forecasting, and services cannot properly assist agricultural practitioners to optimally meet the ever-increasing demands for food and agricultural by-products. This study was carried out to compare several proposed estimation methods and to identify the best method for determining the rainfall variables in Southwest States of Nigeria. Rainfall variables (rainfall onset, rainfall period and rainfall cessation) were estimated from the data obtained from Nigeria Meteorological Training (NIMET) and using 3 different models of Cocheme and Franquin (CAF), Walter and Kowal and Knabe. The estimates of these 3 models were compared with the control (NIMET data) and the results revealed that mean rainfall period ranged between 213.938 and 257.875mm (for Osogbo and for Ikeja) while the variance ranged between 400.706 and 598.71 (for Osogbo and for Ado Ekiti). Generally the variance as well as the mean of the rainfall on-set was lesser than those of the rainfall period. Similarly the descriptive statistics reveal that Cocheme and Franquin’s estimates is closer to the control (The NIMET observed value) while M3 estimate is farther away. The correlation analysis of the relationship between the estimates and the control (the observed data from the NIMET) consistently returned positive/direct correlation values for all the bivariate pair and none of the bivariate correlation was negative. It was observed that M1 has the highest relationship (0.695) with the control hence could be adjudged the most similar. The general linear model (glm) analysis of rainfall on-set across the sites showed that there exists significant difference in the rainfall on-set across the sites because F (5;168 :0.05) = 43.02 obtained for rainfall on-set is statistically significant (p<0.05). Also, the glm analysis of both rainfall period and rainfall cessation across the sites indicated that the F(5; 168 :0.05)= 57.11 and 22.36 ( for rainfall period and rainfall cessation) were both significant. The results of the analysis of the consistency and efficiency of the estimator indicated that for all the estimators, V( ̂) approaches zero as n   and there exist insignificant exception to the cases where the variance tends to jump up again.