Hydrological time series model based on conditional heteroskedasticity analysis and its application

The conditional heteroskedasticity phenomenon is often neglected on the research of hydrological time series. Based on this problem, this paper establishes the hydrological time series model for analyzing and forecasting. First of all, the trend term and cycle term are obtained from the original time series by Census X12, which is given a conditional heteroskedasticity model. For the residual term, the Markov forecast model based on the BX data conversion is selected. The three models, were coupled, and the algorithm procedure was programmed. According to this hydrological time series model based on the conditional heteroskedasticity is put forward, which is used for analyzing and forecast. The model is applied to the monthly series data during from 1975 to 1999, of by a Henan Province's Zhanyushan hydrological station in Huaihe River Basin. The result shows that the average deviation of the model in this paper is 17.45%, which is significantly higher than the conventional models used in the hydrological time series analysis, such as the ARMA model, the ARCH model, etc.