[MITgcm-devel] serious bug in rStar code

[Jean-Michel Campin]

Hi Dimitri,

On Sun, Jun 19, 2005 at 06:41:33PM -0700, Dimitris Menemenlis wrote:
> Jean-Michel, what do you think would be the consequence of the 
> rStar/hFac bug?  One thing that I (and others at JPL) have noticed is 
> that sea-surface height variability in shallow coastal regions seems a 
> bit excessive.  Could the rStar/hFac bug be in part responsible?

It will definitively have some effect on shallow regions. But I don't
precisely know if it will reduce SSH variability or not 
(where eta < 0 => reduces it , where eta > 0 => increases it, I guess).

> Second hypothesis and question regards formulation of bottom drag 
> coefficient in shallow regions, specifically of "bottomDragQuadratic". 
> If you look at experimental literature for shallow regions, the typical 
> value "bottomDragQuadratic=.002" is obtained relative to depth-averaged 
> horizotnal velocity.  The way that bottomDragQuadratic is implemented in 
> the model, it seems to be relative to bottom level velocity only (i.e., 
> mom_calc_ke is applied to uFld and vFld of bottom level).  This would 
> give a drag coefficient that depends in a weird way on choice of model 
> vertical levels and which in general would tend to be too weak relative 
> to experimental values.  Am I missing something?
>
> Dimitris

My understanding of bottomDrag coefficient is that it's fitted
(given the vertical resolution near the bottom, weather you resolve
the bottom boundary layer or not, the viscosity near the bottom ...)
to give reasonable results. One way to deal with shallow regions (and
small dz near the bottom) could be to increase the vertical viscosity
(implemented implicitly) near the bottom in those regions (as it should 
be when you start to resolve the bottom boundary layer). 
But you can think also of using an average velocity (from the 2 bottom
layers for example) if the bottom dz is very small, in order to compute
the bottom stress, with the idea that it will make the drag less 
dependent on the vertical resolution. One thing that looks strange to me 
is not to apply the bottom stress only in the bottom level (because
it is supposed to go there). And finally, it's worth to check that there
is no weird behavior (stability Pb) when the stress is computed
from and applied to different levels.

How does KPP deal with bottom friction ? Does it increase bottom
viscosity as it should ?

Jean-Michel