[MITgcm-devel] [Fwd: Re: ECCO numerics]

[Dimitris Menemenlis]

All, is anyone aware of bug mentioned below?
Is it already fixed in pkg/exf/exf_bulk_largeyeager04.F
Also, I have been using Large & Pond 1981 for all the CS510 integrations.
Should I switch to Large & Yeager 2004?
I seem to recollect that someone had verified that the
two formulations were identical?

Dimitris


-------- Original Message --------
Subject: Re: ECCO numerics
Date: Tue, 07 Aug 2007 10:23:39 -0400
From: Stephen Griffies <Stephen.Griffies at noaa.gov>
To: Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
CC: Alistair Adcroft <Alistair.Adcroft at noaa.gov>

Hi,

Thanks for the feedback. Very interesting...

One (pesky) thing to note is that just a few weeks ago, a French student
discovered a (perhaps minor) bug in the bulk formulae code used for the
CORE forcing.  This bug affects the ce and ch heat exchange
coefficients.  I do not know if the bug would affect your CORE runs
much, but something to consider for future runs is an update to this
code (assuming you downloaded it from GFDL).

Go to

http://data1.gfdl.noaa.gov/nomads/forms/mom4/CORE/code.html

and update the code in ncar_ocean_fluxes.f90

Steve



Dimitris Menemenlis wrote:
> Steve, thank you for helpful feedback.
> 
> Yesterday I looked at the first overall comparisons of 18-km 
> cubed-sphere integrations with and without GM-Redi versus global 
> temperature and salinity profiles from XBT, CTD, TAO moorings, and ARGO 
> floats and your reservations regarding use of GM-Redi at this resolution 
> are justified.  Overall three 1992-2002 integrations with GM-Redi do 
> slightly worse than a baseline integration without GM-Redi.  I still 
> need to examine the solutions in more detail though, especially in Mode 
> water formation region, to figure out what happened.
> 
> You are also right about tunable parameters for GM-Redi: there are way 
> too many, especially when you add tapering so I was a bit at a loss what 
> to choose.
> 
> For 10-m wind forcing we have tried NCEP reanalysis, Large and Yeager's 
> CORE, which you provided (thanks!), ERA40 followed by ECMWF analysis 
> after August 2002, and some winds estimated from a coarse-resolution 
> adjoint-method optimization.
> 
> Another motivation for turning on GM-Redi is that OS7MP solution, even 
> though spectacularly better in places, e.g., the Arctic Ocean, creates 
> regions that are marginally stable and where grid-scale noise appears.  
> An integration forced by ECMWF analysis and without GM-Redi blew up in 
> June 2004 (for a reason that I have not yet diagnosed) and I am 
> reluctant to lower time step or to add horizontal diffusivity, hence the 
> experiments with GM-Redi.
> 
> Dimitris
> 



-- 
Dimitris Menemenlis <menemenlis at jpl.nasa.gov>
Jet Propulsion Lab, California Institute of Technology
MS 300-323, 4800 Oak Grove Dr, Pasadena CA 91109-8099
tel: 818-354-1656; cell: 818-625-6498; fax: 818-393-6720