Abstract
A methodology for extracting random components and estimating the relationship between coherent components of hydrologic time series has been presented. The technique has been verified using synthetic data and validated using measured data. The results show that random component statistics are accurately estimated, and the correct relation between pairs of stationary and nonstationary time series are estimated for frequency components whose amplitude significantly exceeds the standard deviation of the random component. Applying the method to measured, nonstationary hydrologic data demonstrated that the technique was able to accurately predict the measured fluctuations in 9 of the 12 years considered. Significant portions of the spectrum of the measured series were found to have characteristics consistent with the existence of a random component. A primary application of the results of this study is to fill in missing data at one location based on measured data at an adjacent location.
