Forecasting of Rainfall Using Arima Model

This research paper presents a long-term rainfall forecasting model for the Magadh region of Bihar (covering Nawada, Patna, Gaya, and Nalanda) using the Seasonal ARIMA (SARIMA) approach based on the Box-Jenkins methodology. Historical monthly rainfall data from 1986 to 2024 were utilized to develop the model, with data from 2020 to 2024 used for validation. The model development followed our key step: identification, estimation, diagnostic checking, and forecasting. Model validation was carried out through residual analysis and comparison with actual data, ensuring accuracy through statistical metrics including MAE, RMSE, MAPE, R2, and Nash efficiency. The SARIMA (1,1,1) (1,0,0) model was found to be the most effective for forecasting rainfall, offering predictions with 95% confidence intervals for the next 10 years. The analysis was conducted using SPSS 25 software, demonstrating its utility in handling time series data. Additionally, the research paper proposes a set of functional requirements for rainfall forecasting software, supported by mockups to guide user interaction. The research highlights the complexity of rainfall prediction and the importance of robust statistical tools in hydrological and meteorological forecasting.