[MITgcm-support] Re: AIM_v23
David Ferreira
dfer at mit.edu
Wed Nov 14 10:46:15 EST 2007
Hi Christophe,
I transferred your e-mail to the mitgcm-support list, as others
could contribute to/benefit from the discussion (you can register
on http://mitgcm.org/mailman/listinfo/mitgcm-support)
Back to your question: I don't think your problem is related
to AIM/SPEEDY itself. What limits the time-step (and the
performance of the model) is the dynamics. ECHAM is a spectral
model and can achieve large time-steps while the MITgcm is a
grid-point model which, in the particular experiment you are
looking at (aim.5l_LatLon), is probably limited by the
convergence of the meridians at the poles.
That said, you might be able to increase the time-steps by using
two things:
1) the latest JMC numerical improvements (notably the implicit
gravity wave). I put in attachment an example of a "data" file
that activates these options.
2) the cube-sphere grid (an example is in the verification
experiment aim.5l_cs).
Using these 2 elements, a 16 minutes time-steps is possible
with a C32 cube-sphere grid (C32 is about 2.8 degree resolution
not far from the guaussian grid of T42 I think).
Cheers,
david
Christophe (Kristof) Sturm wrote:
> I've just started setting up the MITgcm (with AIM as atmosphere
> alone, later coupled to the ocean and its biogeochemistry), with the aim
> to implement the dynamical vegetation model LPJ to it and perform C4MIP
> type experiments. Kerim recommended me to ask you advice concerning
> AIMv23, since he has not much experience with it. Would you be willing
> to give me some guidance ?
> Currently, my problem is as follows: I was able to download, compile
> and run the aim.5l_LatLon control experiment (on a single CPU of my
> desktop Linux PC, with the ifort compiler). I extended the run period to
> one year, and it took over 4h to complete the simulation. Shouldn't it
> be somewhat faster ? This is as fast/slow as ECHAM T42, and
> substantially slower than the original SPEEDY. I was further surprised
> that the default model time-step (deltaT) was set to 450s, i.e. ~7 min,
> where ECHAM T42 has an internal ~30min time step. Is this setting
> accurate, if not would this explain why AIM is surprisingly slow ?
> Thanks for your help, Christophe
>
> In case you wouldn't have the papers describing the LPJ and SLAVE
> models, I send you the corresponding references (in BibTeX format).
>
> @article{sitch:2003,
> author = {Sitch,S. and Smith,B. and Prentice,I. C. and
> Arneth,A. and Bondeau,A. and Cramer,W. and
> Kaplan,J. O. and Levis,S. and Lucht,W. and
> Sykes,M. T. and Thonicke,K. and Venevsky,S.},
> title = {Evaluation of ecosystem dynamics, plant geography
> and terrestrial carbon cycling in the {LPJ} dynamic
> global vegetation model},
> journal = {Global Change Biology},
> volume = 9,
> number = 2,
> pages = {161-185},
> year = 2003,
> adsurl = {http://www.blackwell-synergy.com/doi/abs/
> <http://www.blackwell-synergy.com/doi/abs/>
> 10.1046/j.1365-2486.2003.00569.x},
> doi = {10.1046/j.1365-2486.2003.00569.x}
> }
>
> @article{friedlingstein:1995a,
> author = {{Friedlingstein}, P. and {Fung}, I. and {Holland},
> E. and {John}, J. and {Brasseur}, G. and {Erickson},
> D. and {Schimel}, D.},
> title = "{On the contribution of $CO_{2}$ fertilization to the
> missing biospheric sink}",
> journal = {Global Biogeochemical Cycles},
> year = 1995,
> volume = 9,
> pages = {541-556},
> doi = {10.1029/95GB02381},
> adsurl = {http://adsabs.harvard.edu/cgi-bin/nph-bib_query?
> <http://adsabs.harvard.edu/cgi-bin/nph-bib_query?>
> bibcode=1995GBioC...9..541F&db_key=PHY},
> adsnote = {Provided by the Smithsonian/NASA Astrophysics Data
> System}
> }
>
> _______________________________________________________
> Christophe (Kristof) Sturm <kristof.sturm at bjerknes.uib.no>
> http://www.uib.no/People/kst064/
> Bjerknes Centre for Climate Research - CARBOOCEAN
> University of Bergen, Allégetan 55, 5007 Bergen, Norway
> ph.: +47 5558 8699, fax: +47 5558 9883
>
>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: data
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20071114/b5d9c07f/attachment.el>
More information about the MITgcm-support
mailing list