[MITgcm-support] rhsMax

[Jean-Michel Campin]

Hi Uli,

> This is an easy one to answer for sure. I was just wondering if someone
> could give me a brief explanation what the 'rhsMax' stands for that is
> printed out in the STDOUT.* files, and what it tells me about the
> progress of the run.

rhsMax is the maximum value of the Right Hand Side term in the
eliptic equation that is solved implicitly using a Conjugate gradient
method (cg2d). It does not give you much information regarding
how the run behave. The same thing with Sum(rhs), which is the sum of the 
RHS, and is printed on the same line. 
What is clear is that any time the model blows up at any isolated 
point (a NaN or a very big value), 
since Sum(rhs) & rhsMax are computed over the whole domain, 
they will catch this unphysical value and print a NaN or a very
big number.

> I have had a few runs recently that blew up resulting in lots of NaNs in
> place of rhsMax, the model continued running though producing a lot of
> nonsense data. I am assuming it has to do with the open boundaries ...

The model don't always stop after finding a NaN. There are some compilers
that have an option to stop when a NaN is found. Otherwise, if
you are doing a short run and/or are not afraid of generating a large
standard-ouput file, monitorFreq=1 will stop the model if 
an unrealistic temperature is found.

Jean-Michel