Nonstationarities in river flow time series

Hydrologic time series often exhibit signs of nonstationarity. These non-station- arities may be due to natural or anthropogenic causes and may manifest as quasiperiodicities, trends, changes in mean and variance and in spectral characteristics. The nonstationarities may be analyzed by using, among other methods, (a) spectral analysis, (b) time-frequency analysis, (c) tests based on evolutionary spectra for nonstationarity in time series, and (d) time scale analysis based on wavelets. In this paper the monthly and annual river flow data from the Midwestern United States are analyzed by using the above-mentioned tests. The results are presented and compared. Perhaps the most important conclusion from this study is that there is a general agreement in the results about nonstationarity of the series. All the river flow sequences analyzed in the study exhibit nonstationarity characteristics. There is also evidence about changes in periodicities with time in these series. These changes indicate that the series may be nonlinear as well.