6. Modelling and Prediction of Rainfall in the North-Central Region of Nigeria Using ARIMA and NNETAR Model

Modelling and predicting rainfall in research are essential because the inferences from the study will assist decision-makers, planners, and policymakers in mitigating the effects of drought or flooding in the environment. This chapter aims to fit time series models to rainfall data from seven states in the Nigerian north-central region. The data used for this research was obtained from NIMET (Jan 1989–Dec 2019). The rainfall data set was modelled and predicted using the conventional seasonal Autoregressive Integrated Moving Average (ARIMA) and Neural Network Times Series Autoregressive (NNETAR) models. The time plot sequence shows the time series data is stationary, and the Augmented Dick Fuller (ADF) test did not suggest otherwise. Furthermore, the Hegy and Canova-Hansen tests indicate seasonality in the data with order 1. When the ARIMA and NNETAR models were applied to the rainfall data set, the analysis revealed that the NNETAR model outperformed the ARIMA model in modelling and predicting the Ilorin, Jos, Lafia, Lokoja, and Minna rainfall data sets. In contrast, the ARIMA model outperformed the NNETAR model for predicting rainfall in Abuja and Makurdi. The fitted models were used to predict monthly rainfall in the north-central region for the next five years. The forecast suggests an expected increase in rainfall in Lafia, Abuja, and Minna. At the same time, an expected decrease in rainfall in Ilorin, Lokoja, Jos, and Makurdi states in the north-central region of Nigeria.