A time series forecasting model for rainfall in Kasese district, Uganda using the SARIMA approach

Abstract

The study developed a time series forecasting model aimed at predicting monthly rainfall in Kasese district, Uganda. Rainfall patterns play a critical role in agriculture, water management, and disaster preparedness in the region. Using data from January 1960 to December 2023 obtained from the Uganda National Meteorological Authority (UNMA), the study employed a Seasonal Autoregressive Integrated Moving Average (SARIMA) model to analyze and predict rainfall trends. Preliminary analysis revealed that the data was non-stationary, as confirmed by the Augmented Dickey-Fuller (ADF) test (test statistic = -8.56, p-value = 0.01), and showed significant autocorrelation based on the ACF and PACF plots. The analysis showed that the months of March–April–May (MAM) and September–October–November (SON), with a mean monthly rainfall of approximately 118.03 mm (95% confidence interval: 111.32–124.74 mm) receive the highest amount of rainfall. In contrast, the driest months—January, February, June, and July—had a mean monthly rainfall of about 47.53 mm (95% confidence interval: 43.02–52.04 mm). A SARIMA (3, 1, 1) (1, 0, 0) [12] model was selected based on the Akaike Information Criterion (AIC = 7948.98) and the Bayesian Information Criterion (BIC = 7976.55). A 12 – month seasonal model was used to capture monthly rainfall variations throughout the year, despite Uganda’s two main rainy seasons. The model demonstrated good predictive accuracy, achieving a MAE of 42.5962, RMSE of 54.972, and MASE of 0.8376. Residual analysis confirmed that the model adequately captured the seasonal and trend components without significant autocorrelation. The study concluded that the SARIMA model provided reliable short-term forecasts of monthly rainfall in Kasese, supporting agricultural planning and disaster risk reduction. The research recommended adopting of the model by local authorities, for short – term forecasts to improve agricultural planning, and also hold workshops and training sessions to educate the people on how to use and interpret the model forecasts and predict rainfall trends.