Trend Identification Of Hydrological Sequence Based On Non-stationary And Wavelet Analysis

Water is the source of life and the foundation of civilization, so it’s an importantbirthplace of human civilization in the river area of the world. Correct understanding ofthe complex characteristics of the hydrological series and the trend is the basis of theanalysis related to hydrology and water resources, but also an important reference forhydrological forecasting, engineering, construction and water scheduling.This article is based on complex systems analysis with non-stationary theory, and itestablishes the theory of the GEV model to explore the relationship betweennon-stationary hydrological variables using the monthly maximum reservoir water leveltime series of the ShuangTaBao reservoir (1972-2001a total of30years) and streamflowtime series of the ChangmaKou observatory based on; Then the framework of the theoryof identification based on non-stationary sequence and wavelet analysis method is foundedwith86years (1919to2004)of Huayuankou runoff data, the obtained results is tested bypractice and comparative analysis verifying the correctness of the theory established. Themajor work is summarized as follows:Focused on the relevant theory and fast algorithms of wavelet analysis, we discussedthe time-frequency analysis theory, based on which wavelet decomposition andreconstruction was done with the original observation station runoff sequence.Furthermore, we extracted wavelet low frequency coefficient corresponding to the variousscales.The revised GEV (Generalized Extreme Value) model for access to the researchobject inherent complex features was first introduced and discussed, then the environmentand algorithms and platform of the theory was analyzed, based on which high reservoirwater level of the ShuangTaBao reservoir and its upstream ChangmaKou observatory runoff time series were studied and analyzed. By introducing the concepts ofnon-stationary measurement and evaluation criteria and the approximation algorithm, weevaluate the nonstationarity of the residuals trend after the extraction by wavelet analysis.Besides, the information structure of the residuals was extracted, and the informationentropy as a characterization of the complexity of the system was calculated.In order to test the validity of the theory on analyzing the sequence complexity andextracting hydrological runoff trend, the obtained results was compared with thetraditional trend extraction methods (Moving Average Method), which showed that theestablished theory to obtain hydrological runoff sequence of low-frequency trend wassuperior, and indicated that it can be effectively applied to engineering practice in theidentification of complex systems analysis and evolution trends.