[Mitgcm-support] for Patrick re: ecco_c43_e13

[mitgcm-support at dev.mitgcm.org]

hi Patrick:

I've been having fun implementing the new ecco release. The forward
model output is not identical to checkpoint 34, but I guess this is to
be expected. The differences are quite small. The cost function
evaluation is reasonable, except objf_ssh_mean is 0. This is because
"wp" is set to zero in cost_weights if #undef EGM_GEOID... . So there is
a slight conflict here. Maybe "wtp" can be used in cost_ssh_mean.F
instead, although this weight has been reserved for the time-varying ssh
misfit. Also, I like the gradient check package. The gradient check is
very good in most places, but very bad in a few select points. The
points are sometimes in the interior, sometimes on the boundary. No idea
why this is the case. The gradient check is also bad at these points if
KPP is turned off. So I have a running version of the 2 degree and 1/6
degree "natl_box" configuration for c43_e13. By the way, are you happy
with the output file format in addummy_in_stepping? So many files!! 

So I'm impressed with the new release and I will try to spread the
gospel at Scripps. 

Jake