[MITgcm-support] Supercooled waters in ice shelf cavities

Jean-Michel Campin jmc at mit.edu
Fri Aug 22 23:31:17 EDT 2025 

Hi Dimitris,

Regarding pCellMix_select, if you just want to have something just below the ice-shelves
and not change the ocean bottom viscosity and diffusivity, you could try to use
 pCellMix_select=20,
like in verification/isomip/input.htd/data

Cheers,
Jean-Michel

On Fri, Aug 22, 2025 at 03:28:07PM -0700, Dimitris Menemenlis wrote:
> Hi Martin, thank you for quick and helpful response.
> 
> To save time, I could just have blindly followed your advice since I came to the exact same conclusions after building a small test case to reproduce the super cold water problem and experimenting: https://github.com/MITgcm-contrib/llc_hires/tree/master/trillium/isomip
> 
> To summarize, in case useful for others:
> 
> The supercooled waters happen for only a limited range of cavity water temperatures.  In the above example, they occur for initial temperature -1 deg C but they do not occur for initial temperature -2 deg C or 0 deg C.
> 
> The presence or absence of pkg/frazil does not make any difference to the accumulation of supercold waters under ice shelfs nor to the transport of the supercooled water.  So as Martin suggests, I will remove it.
> 
> The key problem is the use of "SHELFICEboundaryLayer = .TRUE.,???.  With this flag set, the waters become supercooled, without it, there is no supercooling.
> 
> The first time I tried using "pCellMix_select = 22,??? in llc1080 configuration, the model crashed.  But that simulation had some issues in the bathymetry and ice topography files, which have now been resolved.  So I plan to give "pCellMix_select = 22,??? another try.
> 
> Cheers, Dimitris
> 
> 
> > On Aug 15, 2025, at 1:08???AM, Martin Losch <Martin.Losch at awi.de> wrote:
> > 
> > Hi Dimtris,
> > 
> > but that???s a different issue then, isn't it? So this problem is not restricted to individual points in isolated holes.
> > 
> > Why do you want to have the somewhat untested frazil-ice package in this run? From your runs, it appears as if this pkg is moving too much negative heat to the surface in general, doesn???t it? It???s not clear to me, why there should be so much supercooled water (away from the ice shelf ocean interface) that it becomes a problem and needs to be dealt with in something like pkg/frazil.
> > 
> > Rather than making the configuration more complicated by adding some asymmetric fixes (which I don???t know how consistently to do in either shelfice not seaice package), I would simplify and remove the frazil ice package.
> > 
> > You use  SHELFICEboundaryLayer = .TRUE., together with
> >  pCellMix_select = 2,
> >  pCellMix_viscAr = 173*4.e-4,
> >  pCellMix_diffKr = 173*2.e-4,
> > 
> > The parameter ???2??? means enhanced mixing near the bottom, doesn???t it? if you put ???22??? you???ll also have it near the surace/iceshelf, and you could get rid of the boundaryLayer fix. I would probably avoid SHELFICEboundaryLayer = .TRUE., if possible.
> > 
> > Martin
> > 
> >> On 14. Aug 2025, at 17:56, Dimitris Menemenlis <dmenemenlis at gmail.com> wrote:
> >> 
> >> Hi Martin, would there be a more elegant (and less manual-labor intensive) way to speed this heat transfer even more.  Removing grid cells one by one might be practical at llc1080 but it will become incredibly costly if we were to attempt same strategy at llc8640.  This particular point is extreme, -100 deg K (not C), but there are many more locations where the temperature gets extremely cold, i.e., below the freezing point, including in the figures that I sent yesterday, the dark blue dots along the upper wall of the cavity, especially for the North/South section.
> >> 
> >> Here is another example of something weird (fun?) happening at the edge of the a cavity.  The attached are vertical sections of temperature and salinity.  Sea ice is also plotted on top of open ocean as white. Note the supercooled grid cell just at the open-water edge of the cavity, which causes a negative upward flow of heat (due to pkg/frazil) on adjacent open water cell, which gradually grows sea ice, ~20 m thick in ~20 days, and rejects salt which accumulates at bottom of domain, the orange grid cell below the 20-m sea ice cell.
> >> 
> >> I guess this is happening because we lack capability to properly deal with formation of frazil sea ice.  So question is, can we make the exchange coefficient for shelfice asymmetric, i.e., respond faster to hypercooled water than to water above freezing point, as it can be configured for pkg/seaice?
> >> 
> >> Dimitris
> >> 
> >> 
> >> <TemperatureSection.jpeg>
> >> 
> >> 
> >> <SalinitySection.jpeg>
> >> 
> >> 
> >> 
> >> 
> >>> On Aug 14, 2025, at 6:53???AM, Martin Losch <Martin.Losch at awi.de> wrote:
> >>> 
> >>> Hi Kayhan and Dimitris,
> >>> 
> >>> with SHELFICEuseGammaFrict = .TRUE., your uStar is limited from below by 1e-3 (ie. 1mm/sec, this is actually hardwired in the code), which will give you small but non-zero exchange coefficients, order 0.0027 (which is not so small after all), if I am not mistaken.
> >>> 
> >>> So with T = -100degC, the (freezing) heat flux at the interface should be also record breaking even though there???s no flow.
> >>> 
> >>> I would try to get rid of this grid cell, and any other hole like this in the ice-shelf topography (fill it).
> >>> 
> >>> Martin
> >>> 
> >>>> On 14. Aug 2025, at 14:57, Dimitris Menemenlis <dmenemenlis at gmail.com> wrote:
> >>>> 
> >>>> Dear Martin, Yoshi, and Jean-Michel, in working towards the llc8690 set-up, Kayhan Momeni and I ran into following problem, illustrated in the two attached figures.  The figures show North/South and East/West vertical sections of Theta in one of the cavities, beige is land and blue is ice shelf.  The grid cell circled in red (at top west end of the East/West section and middle top of the North/south section) reaches a temperature of close to -100 K ??? I think that must be a world record!
> >>>> 
> >>>> The model configuration that we use is documented here: https://github.com/MITgcm-contrib/llc_hires/tree/master/trillium/llc_1080
> >>>> Importantly we have enabled pkg/shellfice and pkg/frazil
> >>>> 
> >>>> What I think is happening is that this location does not see any circulation because it is surrounded by land or ice shelf so negative heat anomaly from neighboring deeper cavity cells, mediated through pkg/frazil, rise to the surface and get stuck in that ???inverted well" grid cell, which does not interact with the ice because u* is zero.
> >>>> 
> >>>> Have you ever ran across this situation and how did you solve (or suggest that Kayhan and I solve)?
> >>>> 
> >>>> Cheers, Kayhan and Dimitris
> >>>> 
> >>>> 
> >>>> <EastWestVerticalSection.jpeg>
> >>>> 
> >>>> <NorthSouthVerticalSection.jpeg>
> >>>> _______________________________________________
> >>>> MITgcm-support mailing list
> >>>> MITgcm-support at mitgcm.org
> >>>> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
> >>> 
> >>> _______________________________________________
> >>> MITgcm-support mailing list
> >>> MITgcm-support at mitgcm.org
> >>> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
> >> 
> >> _______________________________________________
> >> MITgcm-support mailing list
> >> MITgcm-support at mitgcm.org
> >> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
> > 
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
> 

> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support

More information about the MITgcm-support mailing list