Predicting daily water level with exogenous variables ARIMAX-MS-GARCH model

Human activities have intensified climate change, increasing the frequency and severity of floods, posing challenges for accurate flood prediction. This study introduces a multistate MS-generalized autoregressive conditional heteroskedasticity (GARCH) model with structural breakpoints to improve daily water level forecasting. The model is compared with autoregressive integrated moving average (ARIMA), ARIMA-GARCH, and ARIMA-MS-GARCH to assess its predictive capabilities. The key findings are as follows: (1) structural breakpoints significantly impact water level fluctuations; (2) volatility patterns vary across different states and are influenced by rainfall and water levels; (3) traditional models fail to account for these nonlinear changes; and (4) the autoregressive integrated moving average with exogenous variables (X) (ARIMAX)-MS-GARCH model effectively integrates structural breaks and external variables, enhancing prediction accuracy. These results highlight the necessity of incorporating structural changes and external factors in hydrological modeling, offering valuable insights for improved flood forecasting and early warning systems.