Detection of Nonstationarity in Hydrologic Time Series
提出一种基于谱特征和指数移动平均模型的方法，用于检测水文时间序列中由人为或自然因素引起的变化，并通过合成数据和实测数据验证了该方法的准确性。
Detection of changes in hydrologic time series due to intervention by man or natural causes is an important problem. Although intervention analysis has been used in the recent past to analyze nonstationary hydrologic time series, the necessity to specify a model of change and an initial time at which the time series has started to change are obvious disadvantages of intervention analysis. An alternative to intervention analysis is a method which is based on spectral characteristics and an exponential moving average model. The basic objective of the research discussed in the present paper is to test this alternative method. The model is tested by using synthetic uncorrelated and correlated data with step and gradual changes as well as by using real hydrologic time series. The sensitivity of the model to different parameters is also explored. The alternative model is found to be quite accurate in detecting changes in hydrologic time series.