[MITgcm-support] weird behavior when checkpointing with pkg/ctrl

[Raphael Dussin]

Hi,

I cannot reproduce my results between runs with and without checkpointing
when using the pkg/ctrl. I'm not expecting bitwise reproducibility but I am
obtaining large discrepancies after
checkpointing, as if the ctrl were not applied anymore.

Method: 1 job of 360 days, no checkpoint
              2 jobs of 180 days, 1 checkpoint

The ctrl on the initial condition were applied in a previous 1 timestep
job, and are not applied here.
More details and plots in the pdf.

Has anybody experienced similar issues? Any hints to fix this?

Thanks in advance for your input!

Regards,
Raphael Dussin (@Lamont/Columbia)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20190313/b71a000b/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: checkpoint_ctrl.pdf
Type: application/pdf
Size: 435031 bytes
Desc: not available
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20190313/b71a000b/attachment-0001.pdf>