Comparison Of Different Analysis Methods For The Periodicity Of Hydrological Time Series

It is a difficult question to identify and determinate the cryptic period of hydrological time series, including simple cycle, compound cycle and potential cycle, and its quality of analysis will influence the precision of prediction of the hydrology characteristic of the coming year or month directly in its cycle. So, how to carry on the cycle analysis of the hydrological time series effectively have already become a subject for research concerned by hydrology scholar extremely. This thesis discusses mainly several common analysis methods of hydrological time series. We compare them from two aspects of survey sample calculation and Monte-Carlo experiments, in order to understand the precision difference of the several methods farther.In this thesis, three kinds of calculation methods are considered and compared: simple partial wave method, Fourier analysis method and maximum entropy spectral analysis method. These three kinds of methods are often used to calculate the cryptic period of hydrological time series. Collect and survey the materials including the annual runoff, annual rainfall and annual flood season rainfall, etc. There are 57 arrays together in moist and arid regions of the south and the North. The calculation results that are received from the different method are not selfsame. The results of Fourier analysis method and maximum entropy spectral analysis method are comparatively close, but the result of simple partial wave method is very different with them, and the difference is big sometimes.In order to compare farther the validity of these three kinds of methods from theoretical way, the hydrological time series with fixed period are generated by Monte-Carlo method and their periodic components are calculated with these three kinds of methods respectively. The test results indicate: the result of simple partial wave method is affected greatly by the length of the sample; the results of two other methods are close to the population value, but their precision reduces when the stochastic components increase. The stability of maximum entropy spectral analysis method is better slightly according to the theoretical analysis and the results of observed data.