Advanced time series modelling to address seasonal variability and extreme rainfall in Ghana

Abstract

Rainfall forecasting is essential for environmental planning, disaster preparedness, and resource management. Accurate predictions enable better water resource management, agricultural planning, and climate risk mitigation. However, forecasting extreme rainfall events remains a challenge due to their inherent unpredictability, nonlinearity, and sensitivity to seasonal and climatic variations. This study evaluates the performance of various forecasting models—Facebook Prophet, Seasonal Autoregressive Integrated Moving Average (SARIMA), Exponential Smoothing State Space (ETS), Trigonometric Box‒Cox Transform, Autoregressive Moving Average, Trend and Seasonal (TBATS), and Hybrid Autoregressive Integrated Moving Average + Exponential Smoothing State Space (ARIMA + ETS). Monthly rainfall data from Ghana’s western region, spanning 84 months (2017–2023), were used for analysis. Model performance was assessed using Akaike information criterion (AIC), Bayesian information criterion (BIC), mean squared error (MSE), mean absolute error (MAE), R-squared, adjusted R-squared, and Theil’s U statistic. Among the models tested, the Facebook Prophet model demonstrated superior performance, achieving the lowest AIC, BIC, MSE, and MAE, Theil’s U statistic values. Additionally, its high R-squared and adjusted R-squared values indicate a strong model fit, effectively capturing rainfall trends and seasonality. Prophet’s ability to manage outliers further enhanced its forecasting accuracy, emphasizing the significance of outlier handling in time series predictions. This study highlights the importance of models that generalize well and account for nonlinearity and seasonality in rainfall forecasting. The findings provide insights into improving prediction accuracy in climate-sensitive regions. Enhanced rainfall forecasts can strengthen climate resilience, aiding policymakers in water resource management, disaster preparedness, and sustainable agriculture amid climate variability.