APPLICATION OF SARIMA MODEL IN FORECASTING MEAN MONTHLY RAINFALL IN NORTHWESTERN NIGERIA

Rainfall forecasting plays a vital role in rain-fed agriculture, water resources management, and climate risk mitigation, particularly in semi-arid regions of sub-Saharan Africa. This study applies Seasonal Autoregressive Integrated Moving Average (SARIMA) model to analyse mean monthly rainfall data from seven locations in Northwestern Nigeria—Gusau, Hadejia, Kafanchan, Kano, Katsina, Samaru, and Sokoto to generate reliable forecasts over a 36-month horizon. Model selection was based on log-likelihood values, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and residual diagnostic tests including the Ljung–Box and Jarque–Bera statistics. The results revealed strong annual seasonality across all locations, with spatial variability in optimal model structure and performance. While some locations exhibited satisfactory residual diagnostics, others showed remaining autocorrelation and non-normality, reflecting the inherent variability of rainfall series. Despite these limitations, the SARIMA models provide useful short- to medium-term forecasts that can support irrigation planning and water resources decision-making in Northwestern Nigeria