Forecasting monthly rainfall using autoregressive integrated moving average model (ARIMA): A case study of Fada N’Gourma station in Burkina Faso

Abstract

Climate related hazards are challenging vulnerable communities and decision makers at any mitigation and adaptation planning stage. Accurate knowledge on projected climate variables such as rainfall is crucial in setting efficient adaptation strategies. The present study seeks to determine an optimum model to predict rainfall patterns within Fada N’Gourma. To this end, the autoregressive integrated moving average model (ARIMA) were fit to the monthly rainfall record for Fada N’Gourma meteorological stations spanning from 1981 to 2021. Then, the Box-Jenkins method has been applied under R programming language to identify the appropriate ARIMA (𝑝, 𝑑, 𝑞) ∗ (𝑃, 𝐷, 𝑄) model that fits the rainfall records. The stationarity of the dataset has been checked based on Augmented Dicky fuller test. The best model used to predict the next ten-year rainfall was selected based on Akaike information criterion (AIC) and Bayesian information criterion (BIC). The efficiency of the model was evaluated by the root mean square errors (RMSE) and the mean squared error (MSE). The results demonstrate that the ARIMA model (5,0,0)(2,1,0)[12] is an appropriate forecasting tool to predict the monthly rainfall across Fada N’Gourma. Base on this model, rainfall forecast for 10 years was then achieved. The Mann-Kendall trend test for the projected rainfall shows a z = 0.89 and a value of Sen's slope up to 0.88 depicting an increasing trend of the annual rainfall within Fada Gourma by 2030.