Study Of Analysis Methods For The Period Of Hydrological Time Series

Generally, there are periodic components in hydrological time series, which actually are only period on probability definition. It's still a big problem difficult to be solved to identify and determine the cryptic period in hydrological time series. So, the problem how to analyze the period effectively has already become a subject highly concerned by hydrologist. This thesis discusses period analysis methods of hydrological time series, and compares them from theory, measured sample calculation and Monte-Carlo experiments.In the period analysis of hydrological time series, there are five methods: simple partial wave method, Fourier analysis method, power spectrum analysis method, maximum entropy spectral analysis method and wavelet analysis method.Based on the theoretic analysis, this thesis calculates the cryptic period of measured data by the above five methods and compares their results. From results, the conclusion is that different methods lead to different period calculation results. As a whole, the results of Fourier analysis method, maximum entropy spectral analysis method and wavelet analysis method are comparatively close. Additionally there exist period regional regular in some hydrological time series. And the period regional regular of annual rainfall volume is more obvious than that of annual runoff volume.This thesis adopts Monte-Carlo experiments to analyze these methods in order to study them better. The results indicate that these methods can be used to analyze the period of hydrological time series. The results of Fourier analysis method, maximum entropy spectral analysis method and wavelet analysis method are all more accurate, and they can identify the periods well and can be thought as the primary methods. Simple partial wave method may identify some false periods. Power spectrum analysis method is more likely to be influenced by the human fector. The statistic results also show that the length of series will have a certain impact on the result. The longer the series is, the more close between the calculation period and the true period of the series.