[MITgcm-devel] update on radiation part of bulk code

[Jean-Michel Campin]

Hi, 

It's just a comment on the way to adjust fluxes:
It seems to me that the ocean emissivity & ocean albedo are
relatively well known value (I mean, narrow uncertainty).
The reason for keeping ocean emissivity = 1 is given,
e.g. in Large & Yeager, 04.
In contrast, the uncertainty in SW downward and LW downward 
(and specially the global mean) is much larger than 
for albedo. 
Ideally, It would be more logical to adjust the 2 radiative
downward fluxes (and keep the albedo close to its physical
value, or at least, not too far), which means, to do what 
Martin is doing.

Now, since NCEP is using a far too high ocean albedo, this
does not simplify the process.

And finally, changing the albedo or changing the downward SW 
heat flux might just end-up to be equivalent. But seems easier
to justify and present it as a flux adjustment/correction.

Jean-Michel

On Fri, Dec 22, 2006 at 03:52:11PM +0100, Martin Losch wrote:
> Thanks for the suggestion. I'll read your paper again to remind me of  
> the details.
> 
> I am not sure about the linearity: I have removed the mean heat flux  
> from lwdown and the solution is not that much different! I still get  
> an average flux of 8-9W/m^2 into the ocean *in spite* of the  
> corrections (I can check with monitor_exf that it is really applied  
> and it is). If find this all very strange ...
> 
> Martin
> 
> On 22 Dec 2006, at 15:38, Dimitris Menemenlis wrote:
> 
> >>Patrick: we need optimized parameters for exf. When are you  
> >>finally going to do it? (o:
> >
> >This is also what I was going to suggest for Patrick.  Would it be  
> >hard to add
> >emissivity and albedo as time-mean, spatially invariant control  
> >parameters?  But
> >that would not necessarily solve your problem Martin, unless you  
> >used same model
> >configuration, optimized surface boundary conditions, and  
> >integration period.
> >
> >>One could also try to change the ocean emissivity, but I expect to  
> >>be a much
> >>harder tuning exercise.
> >
> >Martin, if you have one baseline integration and one or more  
> >sensitivity experiments and if you assume that model response is  
> >linear, it's really not very hard to do this calibration: http:// 
> >ecco2.org/manuscripts/2005/green.pdf
> >_______________________________________________
> >MITgcm-devel mailing list
> >MITgcm-devel at mitgcm.org
> >http://mitgcm.org/mailman/listinfo/mitgcm-devel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel