[MITgcm-support] Temperature drift in tutorial ( tutorial_global_oce_latlon: Global Ocean Simulation at 4◦ Resolution).

[Martin Losch]

Dear Krishnakumar,

thanks for your suggestions.

The manual usually lags behind the code development, but I agree that the tutorials should be kind of up-to-date. One way of getting this accomplished is that you contribute by updating the tex source (obtain them via: cvs ${youroptions} co manual). You can send you changes (somewhere in s_examples/global_oce_latlon/) to this list, and I'll be happy to check this into the repository. Is diffKhT/S=0 is the only detail requiring and update?

In general you are allowed to have both gmredi turned on and specify a horizontal diffusion. I agree that it does not make much sense, especially since gmredi has it's own minimal horizontal diffusion, but there might be cases when you want to do this for test purposes. The MITgcm is not a black-box that you can use without applying some of your own experience/training (and I have the feeling that you are doing this right now)

You cannot expect to get the same results as Danabasoglu (2004), because the forcing is not the same and the model is different (different code/MOM, different grid, different parameters, etc.). On the contrary, I think that the correspondence between your curve and the ones by Danabasoglu is most striking. So everything seems to be in order from my point of view.

Martin

On Apr 11, 2011, at 9:11 PM, Krishnakumar Rajagopalan wrote:

> Hi All,
> I have an update on the temperature drift in tutorial. 
> In the write up of the Global Ocean Simulation at 4 deg resolution (in the .pdf manual), 
> the horizontal diffusion coefficient for salt and theta, diffKh  is set at  1.E3. 
> But, the tutorial is configured to use the GMREDI package, thus diffKh has to be zero. Note that the \input\data  of this tutorial
> is correct with diffKh=0 in the input file.
> I think it would be good if the manual (.pdf document) is  revised. I followed 
> the manual and did many runs with diffKh=1e3 and GMREDI package, only realizing
> the mistake after reading  the writeup on \tutorial_cfc_offline where it is spelled out that
> when GMREDI package is used, diffKh must be zero.
> 
> I think when we set diffKh to some value, and also use GMREDI package, the diffusive flux,
> df gets calculated once outside the GMREDI_XTRANSPORT subroutine, before being incremented
> within GMREDI_XTRANSPORT. A configuration check to tell users to ensure diffKh=0 when
> using GMREDI scheme would be great.
>  
>  The attached figure (a) shows two cases where GMREDI package is used. In case 1, diffkH
> for salt and theta is set as 1e3. In case 2 diffKh=0. Thus case 2 is the correct setting.
> The results show that the overshoot of theta is larger in case 1.
> Danabasoglu (Ocean Modelling 7 (2004); 323-341; Fig1) showed that the mean theta
> increases to about 4.25 degrees when the tracer and momentum time step are different and stays constant
> thereafter. However, in my case (case 2), the mean theta is seen to reduce after reacing a maximum of about
> 4.15 deg. Any ideas on why this is happening is most welcome. ( I am yet to run cases where the
> tracer and momentum time steps are equal)
>  
> 
> Thanks in advance
>  
> Krishnakumar
>