Abstract: On the basis of expounding the characteristics of different wavelet basis functions, this paper analyses the selection problem of wavelet basis and its parameters for precursory water level data processing and cases for several water level observation items are compared. It is concluded that the db6, bior2.6, rbio2.6, sym4 and sym6 wavelet basis are applicable to detect abrupt changes of precursory water level data, with the double orthogonal wavelet basis bior2.6 having the best result. On the selection of wavelet decomposing layer, the first and the second layers are suitable to get the abrupt change signals, while the fifth and the sixth layers are suitable to extract the slow variation signals, and the fourth to the sixth layers are applicable to recover the original trend process from the high-frequency information.

Key words: Precursory water level data, Wavelet basis, Parameter selection