[Mitgcm-support] Re: timelabs of velocity

[mitgcm-support at dev.mitgcm.org]

The Euler step should be wildly inaccurate
for trajectory calculation.

The RK method should work if you used the
same time level of (u,v) for both steps of
the RK method. Think of the flow (u,v)
being frozen and the RK method (2nd, 3rd,
4th,...) is just a way of improving the accuracy
of a trajectory given the frozen flow. That
the flow is evolving during the time-step is
probably not as important (for stablility or
accuracy) - I'm guessing from intuition rather
than experience that this will work.
There's no doubt the answer will be different
if you have turbulent flow but it would be
a step in the right direction.

There's no obvious place to put it, except somewhere
after correction_step().

In my development branch of the code I've undone the
pipelining and moved the correction step
to the end of forwardstep so that these things
will be easier in the future. The release should
have it all in there. If you code RK2/4 as suggested
above, it would be a small effort add more time-levels
at a later date.

A.