基于离散小波分解的水文随机过程平稳性检验方法

检验环境变化影响下水文过程是否保持平稳性是开展水文分析与计算、水文模拟预报等的重要前提,本研究提出了水文随机过程平稳性检验方法,首先利用离散小波变换分离水文序列中的确定成分与随机成分,然后选择合适方程描述随机成分,利用Akaike information criterion和Bayesian information criterion计算最优时间滞后阶数,再采用单位根检验方法判别随机成分是否具有平稳性.人工生成序列和实测水文序列分析结果均显示,周期和趋势对随机成分的平稳性检验有很大影响,随着序列信噪比增大,检验结果的准确性变差.相比常规方法直接对原序列进行分析处理的做法,所提方法可首先准确分离确定成分并克服上述因素的干扰,因此较常规方法更加有效,进而可以得到更为合理的水文序列平稳性检验结果.

A method for testing the stationarity of stochastic hydrological process based on discrete wavelet transform

It is an important task to evaluate the nonstationarity of hydrological process, which is the basis of hydrological analysis and calculation, hydrological simulation and forecast, and many other hydrological works under the influence of environmental change. In this study, the method for testing the stationarity of stochastic components in hydrological time series is proposed. It is firstly to separate deterministic and stochastic components of hydrological series by discrete wavelet transform. Then, choose a appropriate equation to describe the stochastic components, and calculate suitable lag orders using both the Akaike information criterion (AIC) and Bayesian information criterion (BIC) information criterion. Finally, use the unit root test method to test the stationarity of stochastic components. Results of both synthetic and observed series analysis show that periodicity and trend of series have great influence on the stationary test of random components. With the increase of signal-to-noise ratio, the accuracy of the results is worse. Compared with other tranditional methods dealing with series derectly, the proposed method can firstly accurately separate and remove those deterministic components and overcome the interference of the above factors, so it is more effective than traditional KPSS and PP methods, based on the former we can obtain more accurate stationary test results of hydrological series.