Daily Rainfall Prediction in Tangerang City Using an ARIMA Model Enhanced with a Kalman Filter

Source abstract

Rainfall forecasting plays an important role in agriculture, water resource management, transportation, and hydrometeorological disaster mitigation. In Indonesia, where rainfall patterns are highly variable due to its tropical climate, accurate forecasting remains a significant challenge. This study aims to improve daily rainfall prediction in Tangerang City by integrating the Autoregressive Integrated Moving Average (ARIMA) model with the Kalman Filter. Daily rainfall data collected from the Soekarno–Hatta Climatological Station, Indonesian Agency for Meteorology, Climatology, and Geophysics (BMKG), covering the period from January 2019 to December 2024, were analyzed. The analysis involved stationarity testing using the Augmented Dickey–Fuller (ADF) test and Box–Cox transformation, ARIMA model identification, residual diagnostic testing using the Ljung–Box test, and parameter refinement using the Kalman Filter. Forecasting performance was evaluated using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Scaled Error (MASE). Among the candidate models, ARIMA (2,0,2) achieved the best performance and satisfied all diagnostic assumptions, making it the selected baseline forecasting model. After parameter refinement using the Kalman Filter, forecasting accuracy improved considerably, reducing RMSE from 18.6395 to 7.374, MAE from 12.2425 to 4.243, and MASE from 0.7895 to 0.587. These findings demonstrate that the Kalman Filter effectively reduces prediction errors while improving forecast stability. Therefore, the proposed ARIMA–Kalman framework provides a reliable approach for daily rainfall forecasting and can support water resource management, infrastructure planning, and hydrometeorological disaster mitigation in Tangerang City.