Time series Modeling and Forecasting of Expected Monthly Rainfall in Some Regions of Northern Nigeria Amid Security Challenges

Nigeria’s crop production is predominantly rainfed. However, some key agricultural regions in the country are tangling with issues of insecurity and insurgency. Given this, it has become prominent for agricultural stakeholders in the country to seek for alternative regions to boost and rehabilitate agricultural practices. One option to help the country’s agricultural stakeholders produce alternatives is to model and anticipate rainfall in the country’s stable regions. In this study, we put forth two modeling approaches that can take into consideration trends and seasonality in Nigeria’s rainfall data. The proposed modeling approaches combine the classical-classical model with the classical-machine learning model. The machine learning model combines the artificial neural network, the polynomial, and the Fourier series fittings, whereas the classical-classical modeling approach combines the polynomial and the Fourier series fittings. The models are used to analyze the monthly rainfall data in a few chosen stable regions of Nigeria. Multi-step ahead forecasting using the results of the classical-machine learning model for all the chosen regions showed that starting in the threshold year, Kano, Zaria, Bida, and Nguru regions can expect an increase in rainfall over the next five years (2022–2026), while Yelwa is predicted to experience a decrease in rainfall for the first four years and an increase in rainfall only in the last year (2026).