Searching for Evidence of Nonlinear Determinism in Streamflow Data
Richard Eyckholt
Department. of Physics
Colorado State University

Rainfall data shows evidence of low-dimensional chaos, but streamflow data  does not, which is surprising for streamflows that come from rainfall.  However, there are problems with the analysis of streamflow data. One  problem is that the data sets are so small that finding the correlation dimension is not an easy matter, and even finding an appropriate delay time
can be problematic.  Thus, we developed the C-C method for dealing with such  small data sets.  This method does a good job of finding an appropriate delay time, as well as an appropriate delay time window.  I will discuss this
method and show that the use of the delay time window, rather than the delay  time, makes it easier to find the correlation dimension for small data sets.  While this method leads to improved analyses of rainfall data, streamflow  data still fail to  how evidence of low-dimensional chaos.   The problem may lie in the aggregation and sampling processes involved in streamflow data.  I will show how such aggregation and sampling can obscure the nonlinear determinism in a low-dimensional chaotic system.