[MITgcm-support] more KPP blues, maybe related?

Martin Losch mlosch at awi-bremerhaven.de
Sun Oct 26 10:35:44 EST 2003


Dimitris,

the first problem I am having may also be related to the fact that I am 
close to numerical instabiblity with my set up: I get these 
checkerboard patterns also without KPP, but not as much, although the 
model doesn't blow up. So I played with the diffusion/viscosity 
coefficients a little, and increasing them reduces the problem with KPP 
as well, but I also tried your smoothing suggestions which reduced the 
checkerboard, too. So thanks for your suggestions, it's good to know 
that I am not alone (o:

The second problem that I am having (deep convection/mixing in the 
southern ocean in austral summer, January, in global_ocean.90x40x15), 
may be related to my wrong choice of short wave heating. From the NCEP 
suite of data, downloaded from the LDEO ingrid server, I use for qnet:
qnet = net solar+net long wave+latent+sensible (heat flux)
This is the ncep_qnet.bin in the verification experiment 
global_ocean.90x40x15.
For SHORT_WAVE_HEATING input "Qsw" I use downward solar flux (with 
oposite sign) which may be wrong (now I am almost sure it's wrong). I 
then substract my "Qsw" from qnet to get qres=qnet-qsw and put in data
surfQfile='qres',
surfQswFile='downward solar flux',
Where is my misunderstanding? Should I use net solar heat flux as short 
wave heating input? Please forgive my ignorance ...

Martin

On Friday, October 24, 2003, at 05:55  PM, Dimitris Menemenlis wrote:

> Martin, the checkerboard KPP pattern is something that I have also
> struggled with in the past.  In my case it appeared in tropics, where 
> we
> had the highest resolution.  I had traced it back to a positive 
> feedback
> in KPP physics (large stratification reduces mixing coefficient in turn
> enhancing stratification, and vice-versa) that amplified numerical 
> noise
> in the model.  That effect seemed to be stronger than mechanisms in the
> model (horizontal diffusion, advection, etc.) that tend to smear out 
> 2DX
> noise.
>
> To get rid of this, I have included a bunch of grid-scale filter in
> pkg/kpp.  The filters operate on the KPP fields not the dynamic
> variables.  In my experience these filters got rid of checkerboard
> pattern completely without changing large scale averages of the 
> DYNVARS.
>
> The filters are turned on in the KPP_OPTIONS.h file.
> The default is:
>
> C o When set, smooth zonal shear meridionally and
> C   meridional shear zonally with 121 filters
> #define KPP_SMOOTH_SHSQ
> #undef KPP_SMOOTH_DVSQ
>
> C o When set, smooth dbloc KPP variable horizontally
> #define KPP_SMOOTH_DBLOC
>
> C o When set, smooth all KPP density variables horizontally
> #undef KPP_SMOOTH_DENS
> #ifdef KPP_SMOOTH_DENS
> #  define KPP_SMOOTH_DBLOC
> #endif
>
> C o When set, smooth vertical viscosity horizontally
> #undef KPP_SMOOTH_VISC
>
> C o When set, smooth vertical diffusivity horizontally
> #undef KPP_SMOOTH_DIFF
>
> C o Switch on vertical smoothing, the number of sweeps is set in the
> namelist
> #undef ALLOW_KPP_VERTICALLY_SMOOTH
>
>
> which is good enough to get rid of checkerboard pattern in the tropics.
> I have less experience with checkerboard pattern at high latitudes.  
> But
> you can experiment with
>
> #define KPP_SMOOTH_DVSQ
>
> In general the SHSQ term operates in tropics while the DVSQ term
> operates at high latitudes.  If KPP_SMOOTH_DVSQ is not sufficient to 
> get
> rid of your problem, you can try adding
>
> #define KPP_SMOOTH_DENS
>
> and possibly also
>
> #define KPP_SMOOTH_DIFF
>
> On your second question I am running global_ocean.90x40x15 for 5000
> years with KPP turned on and I have not noticed any problems, but maybe
> it's because I have not looked closely.
>
> D.
>
> -- 
> Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
> Jet Propulsion Lab, California Insitute of Technology
> MS 300-323, 4800 Oak Grove Dr, Pasadena CA 91109-8099
> tel: 818-354-1656;  fax: 818-393-6720
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list