Author(s): W. Wang; J. Ma; Pieter H. A. J. M. Van Gelder; J. K. Vrijling

Linked Author(s): Jun Ma, Weihao Wang

Keywords: Daily flow; Stochastic process; Long memory; Time series model; Forecast

Abstract: The daily flow processes (DFPs) of the headwaters of the Yellow River at Tangnaihai and of the middle reaches at Tongguan are decomposed into trend component, periodic component and stochastic noise component first, and the trend, periodicity, autocorrelation and long memory properties are analyzed correspondingly. The analysis shows that: There is no obvious trend in the average annual flow process at Tannaihai, but there is downward trend at Tongguan; The DFP at Tangnaihai is is much smoother than that at Tongguan because of different situations of rain distribution within the year and different intensity of human intervention; Strong seasonality present in autocorrelation functions (ACFs) at both Tangnaihai and Tongguan, and autocorrelation coefficients at Tangnaihai are generally much larger than at Tongguan; DFPs exhibit strong long-memory properties at both Tangnaihai and Tongguan, while DFP at Tangnaihai shows stronger long-memory properties. Autoregressive integrated moving-average (ARIMA) models, autoregressive fractionally integrated moving average (ARFIMA) model and periodic autoregressive (PAR) model are fitted to DFP at Tongguan and Tangnaihai and preliminary forecasts are made. The forecasts show that the predictability is highly correlated with autoregressive coefficients.

DOI:

Year: 2002