[MITgcm-devel] thsice + seaice

Tue May 6 03:00:03 EDT 2008

Hi Jean-Michel,
with a running, not-exploding model I still get a lot of "errors"  
like this one
> BBerr - Bug: IceT(1) > Tmlt 239  32  1  1 -5.376776E-02 -5.400000E-02
with IceT(1) getting as high as -1.e-2, but never positive. I wonder  
if this is anything to worry about, or whether is this an occasional  
non-convergence of the solver (thsice_solve4temp).

BTW, these multiple write statements in thsice_solve4temp make it  
impossible to vectorize this routine (the other issue are the called  
routines that have to inlines). Is there a way to get around these  
write statements, like enclosing them in #ifdef ALLOW_DBUG_THSICE,  
since they don't do anything but print out warnings. Or is this  
package still so unstable that we cannot afford the absense of  
continuous warnings?

Martin

On 6 May 2008, at 01:51, Jean-Michel Campin wrote:

> Dimitris & Martin,
>
> Good that it's working for both of you.
> It seems that there is something strange in the initialisation.
> Don't have time to look at it now, but will at some point.
>
> An other thing:
> who is supposed to update the output.txt of seaice_obcs ?
> Otherwise, (since April 26) if something (wrong) happens to this  
> set-up,
> we will not be able to catch it.
>
> Cheers,
> Jean-Michel
>
> On Mon, May 05, 2008 at 04:03:46PM -0700, Dimitris Menemenlis wrote:
>> Martin, now that I have solved my diagnostics problem, I finally also
>> have a thsice + seaice setup running on the Arctic grid.  To get
>> around the "BBerr - Bug: IceT(1) > Tmlt 239  32  1  1 -5.376776E-02
>> -5.400000E-02" errors, I initialized with sea ice thickness only,
>> i.e., no area and no snow.  Maybe snow plus thickness would have
>> worked also.  I did not try that combination.  But area plus  
>> thickness
>> initialization did not work at all for me.  D.
>>
>> Dimitris Menemenlis <menemenlis at sbcglobal.net>
>> 5056 Oakwood Ave, La Canada, CA 91011-2450
>> tel/fax: 818-790-6735;  cell: 818-625-6498
>>
>> On May 2, 2008, at 6:23 AM, Martin Losch wrote:
>>
>>> FYI,
>>> I have put a quick and rough comparison of the seaice and thsice
>>> THERMODYNAMICS here:
>>> <http://mitgcm.org/~mlosch/cmpice.tar>
>>> This is a spinup-run that I have done here at AWI on the "NAOSIM"
>>> 1/4th degree grid, shown are monthly means effective thickness,
>>> concentration, "salt flux", and a histogram of ice thickness
>>> differences after 10 years of integration. It's immediately clear
>>> that thsice produces more ice and in particular in summer, the ice
>>> concentration is higher (which is good, because in general, the
>>> summer ice concentration is underestimated by a lot of models that
>>> get the winter concentration right, see AOMIP paper C04S11 in JGR by
>>> Johnson et al, 2007, doi:10.1029/2006JC003690). Also the ice extend
>>> is larger in thsice, and may require some tuning. I don't understand
>>> why the "salt flux" is so different, maybe I am not plotting the
>>> same thing. Also, with thsice, there are problem near the open
>>> boundary in the Bering Strait/Sea, which I don't want to investigate
>>> now.
>>> In general, the thsice seems to work nicely (although I has a
>>> significant portion of code that does not vectorize and thus is a
>>> pain in the ... for me, I'll look into that sometime), and I hope
>>> that Dimitris will get similar results. However there are always
>>> plenty of
>>> BBerr - Bug: IceT(1) > Tmlt 239  32  1  1 -5.376776E-02  
>>> -5.400000E-02
>>> which I choose to ignore.
>>>
>>> One problem, which I have not yet solved (and I guess that Dimitris
>>> is also struggling with), is that the model crashes after 1 timestep
>>> when I specify initial conditions like this:
>>>> thSIceFract_InitFile='ice_conc.bin',
>>>> thSIceThick_InitFile='ice_thick.bin',
>>> Maybe I am overlooking something here? I faintly remember that I
>>> need to specify not only thickness and concentration, but also some
>>> initial values for entalphy and snow? Is that true?
>>>
>>> Martin
>>>
>>> On 29 Apr 2008, at 16:36, Martin Losch wrote:
>>>
>>>> Hi there,
>>>>
>>>> I just had a chat with K+K (Frank Kauker/Michael Karcher) who work
>>>> on the Arctic with NAOSIM. They have made a lot of progress in
>>>> generating the adjoint to this model, so that they can look at real
>>>> science questions. Frank show me an extended abstract where they
>>>> investigate the probably causes of the Sept, 2007 ice minimum with
>>>> adjoint sensitivities back to March 2007; as far as I could see,
>>>> it's very nice work, in particular, because it addresses a real
>>>> question. I see this as an incentive to once again push our efforts
>>>> (and encourage Patrick to write up his adjoint stuff).
>>>> Unfortunately, we have been held up by various problems, which I am
>>>> trying to summarize:
>>>> General: I have updated some figures (JFMuv_*) to reflect the
>>>> latest runs, I also cleaned up the figs directory on skylla
>>>> (removed old stuff, that we no longer use)
>>>> - Section 1: Needs a lot of work, maybe Chris can have a look at  
>>>> it?
>>>> - Section 2: Is basically done, except for small bits and details
>>>> on THSICE, the latter I hope Jean-Michel can provide. I suggest
>>>> that we use this section to collect all information on the model
>>>> and then cut later.
>>>> - Section 3: forward integrations: Dimitris and I have finally
>>>> converged on a set of experiments to analyse, most of which have
>>>> been done (by D.) but in particular some of the EVP runs are not
>>>> there, yet. I can't continue with section 3.3 unless they are done.
>>>> - Section 3: I don't know what the status of section 3.1 (bi-polar
>>>> sea ice from global integration) is. Since I have now a global
>>>> integration, too, I could use data from that integration to plot
>>>> ice fields. Dimitris, what's your status here? Do you want me to do
>>>> this (I have no idea what to plot here, seasonal averages?)?
>>>> - Section 4: Patrick, my last email exchange with you was almost a
>>>> month ago, when you wanted to rerun some experiments after
>>>> adjusting SEAICE_gamma_t or availHeatFrac, in order to compare to
>>>> Ian Femty's code/results, what's the status here.
>>>> - Section 5: (Summary/Conclusions), nothing new so far.
>>>>
>>>> what can we do to push this manuscript?
>>>>
>>>> Martin
>>>>
>>>
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel