[MITgcm-cvs] MITgcm/pkg/seaice CVS Commit
Gael Forget
gforget at faulks.csail.mit.edu
Fri Oct 8 17:17:41 EDT 2010
Update of /u/gcmpack/MITgcm/pkg/seaice
In directory faulks.csail.mit.edu:/u/u2/gforget/MITgcm/pkg/seaice
Modified Files:
seaice_growth.F
Log Message:
Merging seaice_growth codes -- part 2 -- actual changes.
Summary of what has happened so far:
- the code got cleaned-up (revisions 1.71 to 1.75) while keeping results unchanged.
- the lab_sea forward verification experiments were revised, because they appeared
to be too fragile to provide a reliable safety net. Typically, minor changes that
should only affect truncation errors (such as the present check-in) would
lead to complete failures (loosing ~ all digits) of lab_sea tests even on Faulks
with g77. The revised tests seem to better discriminate between issues (see below).
Hopefully the upcoming modifications to seaice_growth.F will further stabilize the tests...
Present check-in:
switch the unit of QbyIce (i.e. heat that may be extracted by ice from
the ocean) from ice meters to W/m2, consistent with the other heat stocks.
This change should only affect truncation errors.
Result on Faulks, with g77:
Y Y Y Y>11<16 16 16 16 16 16 16 16 12 16 13 13 16 16 11 14 . . . . . . . . FAIL lab_sea
Y Y Y Y>11<16 16 16 16 16 16 16 16 16 16 13 16 16 16 13 16 . . . . . . . . FAIL lab_sea.fd
Y Y Y Y> 2< 8 9 9 8 13 11 10 8 5 6 2 4 6 6 3 4 . . . . . . . . FAIL lab_sea.hb87
Y Y Y Y>16<16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 . . . . . . . . pass lab_sea.salt_plume
Comment: except for lab_sea.hb87 (that uses evp) truncation errors lead to the loss
of a few digits. Fair enough. I proceed with the commit and will update the results.
Other results: adjoint lab_sea went fine, seaice_obcs went fine,
offline_exf_seaice went bad (so I update its results too).
More information about the MITgcm-cvs
mailing list