[MITgcm-support] KPP and MLDs questions

[Baylor Fox-Kemper]

Hi Galen,
> Stephanie, Dave, Val and I are trying to get our mixed layer depths  
> correct
> in our new 0.5x0.5 North Atlantic setup. Right now they seem too  
> deep in
> comparison to WOA01 data.

Raf and I have been working on restratification of the ML after a  
mixing event.  We have a parameterization of sub-mesoscale eddies  
that shoals the MLD where horizontal density gradients are present.   
This ought to improve MLD in comparison to data (at least, it does in  
another GCM).

However,  the parameterization is not yet in the MITgcm (although we  
plan to implement it soon).

> At the same time, we are trying to understand how the KPP HBL and  
> the MLD
> relate to each other. It seems that the KPP HBL has the potential  
> for a lot
> more variation with time over a month, but the MLD is going to be a  
> more
> integrated measure of vertical homogenization. This suggests the  
> KPP HBL
> monthly mean should be shallower than the MLD, which is indeed what  
> we find.
> Do you agree with this interpretation?

This is a sensible interpretation, so long as the ML is not rapidly  
restratifying (in which case, you'd expect HBL > <MLD>).  Since there  
are few mechanisms to restratify the ML implemented in the model,  
this is probably true.

> Since we are interested in the biogeochemistry (for example, from  
> how deep
> were nutrients mixed up to the surface) I wonder if a diagnostic of  
> maximum
> KPP HBL over a month might be a better estimate of the depth of actual
> mixing. Does this sound sensible?

This would be the correct interpretation for the mixing depth of  
tracers other than density (or temp).  Just because the density and  
salinity are mixed to the MLD doesn't mean that other tracers (e.g.,  
biological ones) will be mixed.  KPP will mix those tracers as well,  
but since they have different boundary conditions and different  
initial conditions, the results may differ.

> Finally, any suggestions on where to start tuning KPP to make the  
> MLDs a bit
> shallower?

Not to be overly self-promoting, but I think the real solution is to  
use our sub-mesoscale eddy parameterization when it gets implemented...

In the meantime,
1) Ricr (in kpp_readparms.F)  can be made larger which will make the  
Richardson number cutoff higher (i.e. deeper mixing depth).
2) You can increase the rapidity of mixing during convection  difmcon  
= momentum  difscon = salt (and other tracers)  diftcon =  
temperature.  This may not do much, since you may have already  
annihilated whatever gradients are present with the 0.1 default  
settings.

Cheers,
   -Baylor