Nonlinear Dependence in Hydrologic Time Series

Studies investigating the chaotic nature of hydrologic processes commonly employ the correlation dimension method. Although this method is generally reliable, it may also have some limitations, influenced by a host of factors. One such factor is the delay time for reconstruction of the phase-space to represent the underlying dynamics. There is also an increasing realization that fixing the delay time window rather than just the delay time itself may be more appropriate. An attempt is made in this study to test the C-C method towards fixing the delay time window and assessing the nonlinear and chaotic dynamic characteristics of hydrologic time series. Three data sets are studied: daily flow from St. Johns River, Florida; biweekly volume from the Great Salk Lake; and daily rainfall from Seoul. The results from this method are also compared with those obtained using the autocorrelation function method.