[Mitgcm-support] patrick: cost_ssh issues.

[mitgcm-support at dev.mitgcm.org]

Hi Patrick: Lots of questions and suggestions for the SSH cost routines.
Bottom line is that they work, but they could use some fine tuning I
think.

1. Any idea why the TP and ERS anomaly fields only have positive values?

2. In cost_weights, I think #ifdef ALLOW_EGM96_ERROR_COV should be
replaced with #ifdef ALLOW_SSH_MEAN_COST_CONTRIBUTION because it doesn't
read the covariance matrix at all; it only reads the diagonal of the
covariance matrix which is used (sometimes) in the ssh mean cost
evaluation.

Also, after reading the geoid_errfile, there should be a check for -9999
values because they can sneak through. 

In cost weights, I have no idea why wsst and wsss should be downweighted
by 10. I would have expected an upweighting. 

3. cost_initvaria : change ALLOW_SSH_MEAN_COST_CONTRIBUTION to
ALLOW_SSH_COST_CONTRIBUTION because tpmeanmask is needed in the ssh
anomaly evaluation too.

4. cost_ssh_mean: at the bottom, a comment explains that wp includes
cosphi weighting already. Thats not true as far as I can tell.

Jake