Period identification in hydrologic time series using empirical mode decomposition and maximum entropy spectral analysis

Identification of periods is a key issue in hydrologic time series analysis. It is also a difficult task in practice when analyzing hydrologic series with complicated stochastic characteristics. In this paper, a new method of period identification is proposed in which empirical mode decomposition (EMD) and maximum entropy spectral analysis (MESA) are used in combination. The EMD method is capable of adaptively decomposing a series into a set of components called intrinsic mode functions (IMFs). By comparing the IMFs with the spread function of white noise with proper confidence level, different components of original series can be identified. These components may correspond to noise or true IMFs under different temporal scales. The EMD method can distinguish the type. The actual periods of original hydrologic series can be identified by analyzing each of the true IMFs using MESA. Analyses of both synthetic and observed series data indicated better performance of the proposed EMD–MESA method to identify periods. Compared with the conventional MESA method which is widely used presently, the EMD–MESA method can effectively avoid the influence of noise and trend on period identification, and it can accurately identify periods even in the case of series with multiple-peaked spectra. Therefore, EMD–MESA not only can improve the period identifying capability of MESA, but also can improve overall period identification by being able to distinguish noise, period, and trend.