Problems in wavelet analysis of hydrologic series and some suggestions on improvement

Applying the wavelet theory and methods to investigate the hydrologic processes such as precipitation and runoff is a hot field. However, several aspects in research are usually ignored: the effect of admissible condition of wavelet functions and the disturbance of noises for the detection of periods, the effect of the length of a hydrologic time-series on the final result, and the choice between the anomaly and the original time series for wavelet analysis. In this paper, these issues are fully discussed. Precipitation data from Lanzhou Precipitation Station are taken for case study. The result indicates that in the wavelet analysis of hydrologic series, denoise methods should be used to eliminate the influence of noises. The MexHat wavelet function satisfies the admissible condition, which ensures that the periodic properties of hydrologic processes can be well represented by using the MexHat wavelet for decomposition. The affected range of hydrologic series which should be discarded before analysis is given. It is also suggested that the anomaly series should be used to highlight the actual undulation of the hydrologic series.