A Hybrid Grey Model to Forecast the Annual Maximum Daily Rainfall

This study proposes a hybrid grey model for forecasting annual maximum daily rainfall in order to determine long-term hydrological system trends. The proposed model uses an integral form of background value to improve accuracy, and applies two residual operators, the Fourier series and the exponential smoothing technique, to correct periodic and stochastic errors. The annual maximum daily rainfall measured by 5 stations around Taiwan are used to validation the proposed model. The performance of the proposed hybrid grey model is compared with those of the autoregressive integrated moving average (ARIMA) and artificial neural network (ANN) models. By evaluation of different indicators, it is shown that the proposed model outperforms both compared models. With more precise information, the proposed model will allow government officials and civil engineering-related industries to better prepare for heavy rainfall, averting potential disasters.