Multiple time scales analysis of runoff series based on the Chaos Theory

摘要

The significance of accepting runoff processes as nonlinear has been gaining considerable in recent times. However, it is hard to explore the types of nonlinearity acting underlying the runoff processes and the intensity of the nonlinearity at different timescales. Daily runoff time series observed at the Pingshan hydrometric station are used for this study. An attempt is made to identify the existence of chaos and the intensity of nonlinear behavior at three characteristic time scales (one day, 1/3 month, and one month). Six nonlinear dynamic methods are used: (1) phase space reconstruction and the delay time is estimated using average mutual information; (2) the sufficient embedding dimension is estimated using the false nearest neighbor algorithm; (3) correlation dimension method; (4) Lyapunov exponent method; (5) 0-1 test algorithm for chaos; and (6) the multi-step Volterra adaptive method. A comparison of results reveals the presence of low-dimensional chaos in the runoff dynamics at the various time scales and the time scales composes only a limit fraction of the intensity of nonlinear behavior. The reasonably good predictions indicate the efficiency of the nonlinear prediction method for predicting the runoff series.