[MITgcm-devel] partial cell under ice-shelf

[Martin Losch]

Hi Jean-Michel,

do you want to just hard-wire uniformFreeSurfLev = usingZcoords.AND..NOT.useShelfice? This will break some set-ups: mine, Veronique's(?), Michael Schodlok's, Madeline Miller's, some people at BAS are using the shelfice package.
I am not all against it, but maybe we should tell them first? 

Martin

On Apr 25, 2012, at 4:06 PM, Jean-Michel Campin wrote:

> Hi Martin,
> 
> Sorry for the delay (I have been distracted by avalanche risk near 
> seaice edges) and thanks for running these tests.
> I looked at the plots you sent, and given what you wrote,
> it should be OK to have these partial cell turned on in calc_phi_hyd.F
> from my perspective:
> a) smoother eta is better for Non-Lin free-Surf.
> b) the fact that this does not change too much T & S might
>  be related to the fact the modification only affect
>  the hydrostatic pressure at the top (which can adjust via etaN 
>  to an equivalent state) but with linear freeSurf, the partial cell 
>  below the ice-shelf did not changed (were already there before).
> So, if it's OK with you, we could make the changes.
> Do you want to check-in the updated gendata.m + input file ?
> I can check-in the switch in set_parms.F first if you prefer.
> 
> There might be also an interesting simplification in calc_phi_hyd.F:
> the specific code (within #ifdef ALLOW_SHELFICE) is probably
> unnecessary now (if uniformFreeSurfLev=F). I will have to check this.
> 
> Cheers,
> Jean-Michel
> 
> On Thu, Apr 19, 2012 at 09:22:07AM +0200, Martin Losch wrote:
>> Hi Jean-Michel,
>> 
>> took me a while, these experiments are so old that with new code nothing works (o:
>> 
>> I now ran the isomip experiment again, basically the run of figure 4 (with CD-scheme and boundary layer scheme, so that it is fairly smooth), once with the original phi0surf-file (exp10_test01) and with your new code (exp10_test02), uniformFreeSurfLev = usingZCoords .AND. .NOT.useShelfIce, and a new phi0surf file that takes into accound the last hFacC*drF*gravity*drho/rhoConst: for that I replaced near the end of verification/isomip/input/gendata.m 
>> phi0surf(ix,iy)=phiHydF(k)*rhoConst;
>> with
>> kp1=min(k+1,nz);
>> drloc=1-g.HFacC(ix,iy,k)
>> dphi=phiHydF(kp1)-phiHydF9k);
>> phi0surf(ix,iy)=(phiHydF(k)+drloc*dphi)*rhoConst;
>> 
>> I am attaching various plots that show that after 4000 days of integration the runs are nearly indistinguishable (writeBinaryPrec=32, so the differences are near the level of working precision). THETA and SALT are scaringly similar. Not quire sure how that is possible. The only real difference you can see is in cmp_hfac_etan.png where the new formulation gives a much smoother ETAN. That's expected, as in the old formulation, I let ETAN do the job of adjusting the pressure, because back then I did not want to change calc_phi_hyd too much.
>> 
>> I'd say, your version looks much better. However, in order to have such "beautiful" ETAN, the exact hFacC is required, and you need to make good assumptions about the weight of the ice shelf (in isomip that simple, as the hydrography is homogeneous in space, but in a realistic scenario, this will not be possible). 
>> Note that there is no effect on the melt rate patterns, which remain as stripy as before (again, not surprising, because ETAN does not have anything to do with that). 
>> 
>> In the verification experiment, I was yet not able to get a good agreement with the reference, not clear why, maybe I screwed up somewhere.
>> 
>> Martin
>> 
>> On Apr 11, 2012, at 5:48 PM, Jean-Michel Campin wrote:
>> 
>>> Martin,
>>> 
>>> For now, it would be with Linear Free-Surf only, to uncomment the line 68 
>>> in set_params.F:
>>> c     uniformFreeSurfLev = usingZCoords .AND. .NOT.useShelfIce
>>> and (after changing the phi0surf file), to see if this works and
>>> if results make sense. But there is no rush, when you find time.
>>> 
>>> Cheers,
>>> Jean-Michel
>>> 
>>> On Wed, Apr 11, 2012 at 04:46:47PM +0200, Martin Losch wrote:
>>>> Hi Jean-Michel,
>>>> 
>>>> not quite sure what I am supposed to test:
>>>> On Apr 11, 2012, at 4:24 PM, Jean-Michel Campin wrote:
>>>> 
>>>>> Now regarding a real test with isomip set-up, I have the impression
>>>>> that some of the input files would need to be changed (in particular
>>>>> phi0surf.exp1.jmd95z ) to also account for partial cell 
>>>>> in order to really make it consistent with the new (uniformFreeSurfLev=F)
>>>>> partial cell code in calc_phi_hyd.F (this is what you described in the
>>>>> JGR paper). I can try to test something, but I am not sure I will
>>>>> do it right. So, when you have some time, if you can check and test this
>>>>> piece of code, it would be great. In the mean time, will keep
>>>>> set_params.F as it is now, so that nothing changed.
>>>> 
>>>> set uniformFreeSurfLev=F and reruns some of the isomip experiments with modified phi0surf file? For the case of a linear free surface? non-linear free surface? 
>>>> 
>>>> I'll look into your new code tomorrow (or the day after tomorrow, I have an all-day meeting tomorrow).
>>>> 
>>>> 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
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel