Research of sediment flux from Yellow River to the sea based on wavelet analysis and wavelet neural Network

Research on sediment flux from river to the sea is a frontier topic in Earth science worldwide. This paper analyzes the characteristics of annual sediment flux from Yellow River to the sea from 1950 to 2007, and presents a feasible approach to predict annual sediment flux. Methods used in this research include wavelet analysis and neural network. The results show that: (1) annual sediment flux variations from Yellow River to the sea have decline tendencies with an average decreasing rate of -20.09 million t/a between 1950 and 2007, and four catastrophe points detected are 1959, 1969, 1985 and 1997; (2) annual sediment flux series has the principal periods of 3-4 years, 8-11 years and 24 years, and the period of 3-4 years was influenced by human activities significantly; (3) prediction of annual sediment flux with the help of wavelet neural network (WNN) model is acceptable, which gives a better result than RBFN (Radial Basis Function Network) model.