Research of a boundary prolongation method in runoff forecast based on wavelet transform

In order to improve the boundary effect of wavelet decomposition in runoff forecast, a new analogical boundary extension mode is proposed based on the similarity measurement. The extension mode can characterize the self-correlation of the runoff sequence excellently with the local similarities on the wavelet multi-scale spaces. The case study of the runoff forecast indicates that the analogical boundary extension mode introduced in this paper is very effective for the runoff sequence with obvious tendency, especially combined used with multi-scale decomposition of wavelet. The boundary extension errors are low and the forecast precision is improved. The initial results appear encouraging in favour of the analogical boundary prolongation method.