Next: Bootstrap Analysis
Up: Nonhydrostatic Effects in Long
Previous: Comparison of the variabilities
To examine the dependence on initial conditions, we repeat run holding
everything else fixed but the initial conditions.
The new initial condition was obtained from the old initial condition
by generating a random number at each location in the interior of the
domain, ,and changing the depth at that point by .This is clearly a very large change in the initial conditions with the
changes being introduced uniformly over all length scales ranging from the
grid spacing to the size of the domain.
The domain averaged potential energy time series for the two runs are shown
in Fig. 3. It is clear that there are large time oscillations or trends in the
time series of the run obtained from the perturbed initial condition.
This trend has been isolated using Singular Spectrum Analysis and is shown in
that same figure.
Physically, these modes may be understood as remnants of the domain scale
perturbation introduced in the initial conditions.
While the higher
wave number modes have readjusted to the dynamics of the system or so to speak
have been forgotten in the course of a 110 year evolution, it seems that
the evolution would have to continued for much longer times before these
domain scale perturbations are forgotten.
Thus, it is justifiable to remove this trend from the time series before
subjecting it to further analysis.
A bootstrap analysis (see later) shows that the overall statistics of the
two runs are the same.
Furthermore, the spectra of the two runs are largely similar.
Next: Bootstrap Analysis
Up: Nonhydrostatic Effects in Long
Previous: Comparison of the variabilities
Balasubramany (Balu) Nadiga
1/8/1998