中文摘要：

水文时序的多时间尺度分析对深入认识流域水文系统的演变特征及中长期水文预报有重要意义,但目前常用的短时Fourier分析,小波变换以及H ilbert-Huang变换等方法均有各自的不足之处.文中尝试运用现代时频分析工具S变换对于桥水库1956—2000年的年入库流量时序进行了时频分析,得到了该水文序列频谱在时-频平面上的三维分布图,它清晰地显示了该时间序列在不同时段内的多时间尺度演变特征.然后通过分析频谱图的拓扑结构,总结了于桥水库年径流量演变过程的关键特征.结果表明,S变换克服了小波变换或Huang变换（EMD）会因分解层数确定的主观性而可能产生的混频现象,这使它在水文时间序列多时间尺度分析及其特征提取方面有着广泛的应用前景.

英文摘要：

Multiple time scale analysis of hydrological time series has important significance for acquiring deeply understanding hydrological system evolution characteristics and mid-long term hydrological forecasting.But the common methods,such as Fourier transform,wavelet analysis and Hilbert-Huang transform,have disadvantages of their own.Stockwell-transform（S-transform）,a tool of modern time-frequency analysis method,was introduced and applied to the case study of annual inflow series of Yuqiao Reservoir for the years 1956—2000.And therefore,the 3-dimension spectrogram of the series was drawn,which clearly displayed the multiple time scale evolution characteristics of the time series in the different periods.Furthermore,through analyzing the spectrogram＇s topological structure,the key characteristics of the Yuqiao Reservoir＇s annual inflow evolution process was obtained.The result shows that S-transform solves the frequency mixing problems which could be induced by the arbitrary number of decomposition layers of wavelet transform and Huang transform（EMD）,which indicates that S-transform will have an extensive application prospect in the multiple time scale analysis and feature extraction of hydrological time series.