[MITgcm-support] gmredi and adjoint

[Lars Czeschel]

Hi all,

as often suggested I usually turnoff gmredi in
autodiff_inadmode_set_ad.F when using the adjoint.

However is there any configuration/setup which allows
gmredi in an adjoint calculation?
There are several options in GMREDI_OPTIONS.h like
#define GMREDI_WITH_STABLE_ADJOINT
do they generally work?

The basic idea (although not very new) is to use kappa
as a control parameter like it is done with eddy stresses
in Ferreira eta al.,2005. As I understand it correctly the
main problem is the tapering for large slopes. Are there any
objections against using kappa_gm and/or kappa_redi as a
control parameter if large slopes would be prevented
in an idealized setup?

Thanks,
Lars

-----------------------------
Lars Czeschel
Theoretical Oceanography
KlimaCampus / IFM
Bundesstr.53
20146 Hamburg