Analysis of the stochastic Characteristics and Modelling of Monthly Rainfall Time Series of Abeokuta, Nigeria

Abstract

This study attempts to analyse the stochastic characteristics of rainfall for Abeokuta, Nigeria and its probable simulation. to this end, rainfall time series data of Abeokuta was obtained from the Nigerian Meteorological Agency (NIMET) for a period of 28 years. the analysis of the stochastic characteristics entails the assessment of temporal trend and periodicity while modelling the time series was done by employing the seasonal multiplicative Autoregressive Moving Average (ARIMA) modelling technique. Results obtained indicate that there is high periodicity and degree of randomness, a multiplicative seasonally differenced ARIMA model was found appropriate for the simulation of the monthly rainfall regime; model choice was on the basis of Akaike Information Criterion (AIC) and Autocorrection functions. On the basis of the two criteria, ARIMA (0, 0.1) X (0, 1.1) X (0,1,1)12 and ARIMA (1, 0,1) X (1,1,1) respectively were adjudged probable candidate models for simulation studies. However considering the fact that rainfall phenomenon exhibits high spatio-temporal variability, the seasonal persistence can only be explained relatively by the autoregressive component rather than solely the moving average component; thus, the second model is preferable. Despite this though, for effective generalisation of simulation results, Artificial Neural Network and Wavelet models are recommended and in this regard too, conditional probability of rainfall occurrence should be considered.