[MITgcm-support] Strange output fields/discontinuities

Thu Feb 13 04:28:04 EST 2020

Dear Martin,

Thanks to you I've slowly fixed several issues. I've recalculated the
pressure below the ice shelf (now there is no more a discontinuity at front
in Eta); I'm employing now the "default" climate forcing set (A2, B3) in
exf and I'm not getting anymore the too warm water in the middle. The funny
values in climate at the boundary sometimes appear, sometimes not, and I
don't think it is a real issue.

I still have some problems with the EVP method in sea ice..

I've run the model with both the LSR and EVP solver but only one ice
thickness category, however, (SEAICE_multDim=1). Whenever I use the 7
categories suggested, the LSR works fine but the EVP crashes almost
immediately (MOM_IMPLICIT_R error when solving the 3-Diag problem, or start
finding NaNs).
I've tried EVP* and adaptive EVP, and from the papers by Kimmritz et al. it
seems that the alpha and beta parameters alone control the convergence of
the EVP scheme, while the number of pseudo-time steps only increase the
accuracy. I've set them in the range from 1 to 1000 but still the model
crashes after a few iterations. Trying less categories or changing the
pseudo-timestep is no help, as also trying unsetting initial conditions for
SI velocity or OBCS options for SI at the boundaries.

I've noticed that only if I unset in data.seaice the SEAICEnEVPstarSteps
for aEVP the models runs ok but very slowly, and a funny value of
SEAICEnEVPstarSteps=123456789 is assigned (!)
With the values found on the documentation for the aEVP it crashes
immediately (setting  SEAICEaEVPcoeff =0.5, SEAICEuseEVPrev = .TRUE.,
SEAICEnEVPstarSteps=500) and also increasing SEAICEnEVPstarStep.

I've also noticed that sometimes the parameters which are not set in the
namelist are given strange values like:

SEAICEaEVPcoeff   = /* adaptive EVP parameter*/
                1.234567000000000E+05
 SEAICEaEVPcStar   = /* adaptive EVP parameter*/
                 1.234567000000000E+05
 SEAICEaEVPalphaMin= /* adaptive EVP parameter*/
                 1.234567000000000E+05

Do you have any suggestions as to make the EVP scheme work? May it be
related with the resolution of my grid (5 km) or the fact that the method
is not implemented on a cartesian grid?

Thanks,

Enrico

Il giorno gio 9 gen 2020 alle ore 17:23 Martin Losch <Martin.Losch at awi.de>
ha scritto:

> Hi Enrico,
>
> this looks to me as if there are probably more than one problem:
>
> - Eta will always be non-zero ***underneath*** ice shelves. That is
> expected. You also need to have this to adjust for potentially (actually
> always) inaccurate loading of the ice shelf … which maybe already explains
> your large positive and negative values: If your
>   SHELFICEloadAnomalyFile = ‘pload_iceShelf_5km_hfac0.2Dr20_extended'
> is not consistent with your initial conditions (T/S) within the ice
> cavities, then you can get very large adjustment processes. In your case it
> looks to me that your loadAnomaly is rather arbitrary and that the model
> needs extra Eta within the cavity and that leads to a reduction of sea ice
> outside of the cavity (the default model is not set up to adjust this type
> of volume change via the open boundaries, so it needs do it via sea level
> change, see PR#251 on github).
> Check out the isomip example in verification/isomip to understand how eta
> works with the shelfice pkg.
>
> non-zero eta over land is also possible with a linear free surface, but
> should be constant in space and time (the Poission equation for eta can be
> solved up to one constant, in some configurations this constant can be
> different from zero, probably because of obcs or similar). I think that’s
> the case for you results.
> Your mean Eta (dynstat_eta_mean) is also ~-1.7m after one day (similar to
> the value over land). I assume it has to do with the obcs.
>
> - obcs: there are funny values along the boundaries in the surface
> forcing, probably not problematic because they are not used, but for
> debugging, it would turn off the obcs pkg.
>
> - your surface forcing definitely looks funny (i.e. what exf makes of it).
> You surface temperatures reach 3degC and more. Is that expected? I usually
> use the sea ice model with the “default set” (A2, B3). Did you try that?
> The net heat flux (forcing_qnet_mean) is order 200W/m^2 downwards. Is that
> expected? It would explain the warming! Did you get the heat flux
> conventions right (positive = upward for latent and sensible heat)?
>
> - I think that your sea ice is simply melting because of your forcing in
> the middle of the domain, that would also explain the excess of fresh water
> in the middle of your domain and the strong downward freshwater flux
> underneath the remaining. Where does the warm water come from?
>
> Martin
>
> > On 9. Jan 2020, at 14:45, Pochini, Enrico <epochini at inogs.it> wrote:
> >
> > Dear all,
> >
> > I'm running a regional model of the Ross Sea with atmospheric forcing,
> sea ice and ice shelf
> > I'm observing strange things in some output field, and in particular in
> some of the fields calculated by the EXF package.
> >
> > In the EXF module I'm giving in input pressure, winds, precipitation,
> evaporation, heat fluxes and downward lw and sw radiative fluxes, and I've
> set the #define to have the configuration  A2+B5 as indicated in
> EXF_OPTIONS.h. I'm also using the SEAICE package with mostly default
> parameters, OBCS with/without sponge layer.
> >
> > I'm observing big negative values in Eta (and Eta not being zero on land
> or ice shelf), and big discontinuities at some boundary in the FU,FV,
> EmPmR, Qnet fields. I didn't use the interpolation option, as I remapped
> the fields offline and they don't have discontinuities in the original
> files. Some seaice-related fields have also discontinuities at domain
> edges, and seaice is missing in some region (maybe because of low thickness
> that was cut off).
> >
> > I attach the configuration and the output files; in results there is a
> STDOUT, and files with suffix "extra" contain surface climate fields as
> calculated by the model. The ERA5 file is the first timestep of the climate
> input forcing, the glorys_2000 file (reanalysis data) is the first timestep
> of a monthly average that I used to make initial conditions and set
> conditions at OB.
> >
> > In netcdf results, 2016 is a week output, 8064 28 day output, starting
> date 2000-01-01
> >  code.tar.gz
> >  ERA5_ross_3hrs_2000_regrid_sample.nc.tar.gz
> >  glorys_2000_monthly_final_jan.nc.tar.gz
> >  input_nmlst.tar.gz
> >  results.tar.gz
> > What do you think is causing the discontinuities in my fields?
> >
> > Enrico
> >
> > _______________________________________________
> > 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20200213/8382c5c7/attachment-0001.html>