[MITgcm-support] Re: regarding SSH cost

Patrick Heimbach heimbach at MIT.EDU
Wed Feb 27 16:38:15 EST 2008


Hi Suneet,

I need to guess, but since you don't prescribe
a priori errors for temp, salt, ssh,
the cost function for these terms cannot be computed
(rather than setting a default value we assume there's no data,
since we don't want to provide a default).
So you need to provide these.
Two parts are relevant:
For temp, salt, you can provide vertical profiles (horizontally  
constant)
in file data.err (see e.g. in verification/natl_box_adjoint/input/ 
data.err
on how this could look like).
Alternatively, you can prescribe 3-dim. fields
(and if you don't quite know what to put, fill them with same number  
everywhere).
It's useful though to get good estimates so as to get a good
preconditioning of the problem.
For SSH, again, you need to provide an error file.

We can provide you the fields that we have
(for T, S they're described in Forget & Wunsch, 2007;
for SSH in Ponte et al., 2007).

Cheers
-Patrick



On Feb 27, 2008, at 2:58 PM, Suneet Dwivedi wrote:

> Hi Patrick,
>
> I am trying to build and run the adjoint assimilation experiment for
> the southern ocean. The problem is that the SSH cost is not
> contributing to the total cost (for optimization and minimization)
> from different variables. Same is the case with Temperature and
> Salinity costs though the model is able to compute the individual
> costs from other variables like SST, aqh, uvwind, atemp, precip and
> swflux and the total cost is nicely decreasing by an order of 4-5 in
> just 20 steps.
>
> I am using Jason1 and ERS2 daily data for SSH and levitus94 monthly
> data for temp and sal in data.ecco file and set corresponding
> multipliers e.g. mult_h,mult_tp, mult_ers, mult_temp and mult_salt
> equal to 1.
>
> The only thing that I am not using for now are corresponding error
> files for temp, sal and SSH because rightnow I don't have any idea
> about how these data files are prepared. Is this making difference??
> Also I don't know the relavance and prepation technique for mean SSH
> file (topexmeanfile) so i have not included it as well in my data.ecco
> file.
>
> May I request you to please suggest me what may be the problem with
> the model set up? Whether this problem is related to bad data supplied
> / there is some other error? Also can you please inform me about how
> the *_errfile are generated?
>
> Eagerly waiting for your response,
>
> With best regards,
>
> Suneet

---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach





More information about the MITgcm-support mailing list