LS-SVM forecast model of precipitation and runoff based on EMD

Through analyzing problem of Wavelet Transform, the annual precipitation series from 1956 to 2000 of the two sub-water resources regions of upper Lanzhou are decomposed into multiple time scale series with EMD method. The results show that the precipitation series have periods that about 3, 4~8, 11 and 22 years. A LS-SVM forecast model of precipitation and runoff based on EMD is established and then compared with the corresponding multivariable regression forecast model. It shows that, as a new and original signal processing method, Empirical Mode Decomposition, EMD, can be used as a tool to decompose hydrological time series into exact multiple time scale sub-series for finding their local change rule, and then to supply input variables with high quality and multiple levels to enhance model quality remarkably.