[MITgcm-devel] c51 (fwd)

[Stephanie Dutkiewicz]

Hi -
I'm including a message from Samar - I gave him a setup and pickup
file from a run I did with c51h_pre. He tried to pickup and run it with
c48 - and it crashed.
He then downloaded c51h_pre and it ran fine (but with some differences
occuring).

I am also having a similar problem - I spun up a model with c51h_pre,
and then have been trying to continue the run with tracers, but using
c46. And it crashes almost immediately. 

Have there been changes that make the model not backward compatible?

steph

---------- Forwarded message ----------
Date: Mon, 10 Nov 2003 18:19:44 -0500 (EST)
From: samar khatiwala <spk at ldeo.columbia.edu>
To: Stephanie Dutkiewicz <stephd at ocean.mit.edu>
Subject: c51

Hi Stephanie

I am running into an inexplicable problem. When I run the new
configuration with c48, the model blows up after sever years. This is
just a straight run (with no passive tracers). It seems to
happen suddenly. I did not want to spend time on figuring out the problem
so I checked out c51h_pre from CVS, recompiled and ran again. This time
everything is fine. Nothing blows up. But if I compare the T/S from this
output (after 20 years) with the output generated from running the
executable you send me, the two are not the same. The differences are
generally quite small, but in a few isolated locations (near boundaries)
they are of order 1 degree. I will live this for the moment, but its a bit
worrisome that the results are not reproducible. Perhaps you have a simple
explaination.

samar