NEMS NMMB topic

All other topics about postprocessing model data (GrADS and other software), about other numerical weather prediction software (including WRF-NMM and WRF-ARW discussion unrelated to UEMS/WRF EMS), and general meteorology talk go in this forum.
meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

NEMS NMMB topic

Post by meteoadriatic » Tue Feb 02, 2016 5:01 pm

Hello,

Anybody interested in sharing experience in building NMMB model? I'm stuck little bit, need some help. Just reply here that you're in if you want to participate.

norulz
Posts: 68
Joined: Thu Mar 12, 2015 1:43 pm

Re: Building NMMB

Post by norulz » Fri Feb 12, 2016 9:34 pm

Hi,

I'm keen to try out the NMMB as soon as possible.
Do you know if it can be built for a stand-alone machine?
The user documentation states:
The NEMS-NMMB model can only be built for shared- and distributed-memory
(OpenMP and MPI) environments. All required libraries are presently available on the
few supercomputers currently supporting NEMS installation. NEMS-NMMB is not
currently supported for installation on other platforms.
Is it just an issue of having a compiler that supports openMP or is there also a need to install openMP libraries on the development machine?

David

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: Building NMMB

Post by meteoadriatic » Fri Feb 12, 2016 11:17 pm

MPICH or similar MPI implementation is, like for WRF, required to distribute tasks among CPU cores or/and multiple cluster nodes.

You need Intel compiler, gcc and PGI seems impossible to use at least for now until someone adapts whole bunch of library code for those compilers. Even with Intel you will have troubles. Required libs are same as for WRF: netcdf, mpich (or similar MPI), png, zlib, jasper, but then in addition to that also whole set of NEMS libs (nemsio, bacio, sp, ......) and ESMF. When you manage to install all that and configure build environment (without any documentation to help you), then you can compile NMM and NPS. If all goes OK, you can go on to try it, again without much help like you can find for WRF. Not to mention that any error you get during whole process, will not be very self-explanatory. At the end, you need WRF installation to compile UPP and finally, script out whole thing to run automatically. Good luck :mrgreen:

Gippox
Posts: 59
Joined: Sat Mar 02, 2013 1:16 am

Re: Building NMMB

Post by Gippox » Sat Feb 13, 2016 7:15 pm

Hello Ivan, I was able to compile NMM-B following this guide, although a little 'time ago ...for Intel/MPICH
https://sites.google.com/site/communitynems/
I was able to run it, but not to make the post production. If you want I can give you the binaries.

Gippox

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: Building NMMB

Post by meteoadriatic » Sat Feb 13, 2016 9:08 pm

Thx,

Not needed I did it finally, by trial and error approach, no guides. UPP works OK with it (not EMSUPP of course).

Gippox
Posts: 59
Joined: Sat Mar 02, 2013 1:16 am

Re: Building NMMB

Post by Gippox » Sun Feb 14, 2016 9:23 am

Great ! :)
At the time i tried to visualize the files by the nmmbvis.gs script , but the maps produced showed the parameters only until the zero meridian .... :?: I will try using UPP . I had noticed that the speed processing of the core is remarkable... :o
On Meteociel i saw was implemented operationally http://www.meteociel.fr/modeles/nemsnmmb.php

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: Building NMMB

Post by meteoadriatic » Sun Feb 14, 2016 10:34 am

Yes it looks faster than old NMM, and that was one of goals creating B grid. Model can even output ctl file with every history file, so you can call GrADS directly without postprocessing, but then important derivable fields like T2m, mlcape, .... will be missing of course. In UPPV3 there is run_unipost script that is adapted for NMMB, you can set it up in few minutes to work.

Take a look also here, these guys run it on global scale:
http://www.hidmet.gov.rs/eng/prognoza/nmmb.php

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: NEMS NMMB topic

Post by meteoadriatic » Sun Feb 14, 2016 5:06 pm


norulz
Posts: 68
Joined: Thu Mar 12, 2015 1:43 pm

Re: NEMS NMMB topic

Post by norulz » Sun Feb 14, 2016 8:55 pm

Hi Ivan,

I'm stuck with the following message:

Code: Select all

conf/configure.nems:3: /home/david/NMMB/19677_libs/esmf_3_1_0rp5/lib/libO/Linux.gfortran.64.mpich2.default/esmf.mk: No such file or directory
svn: E155007: '/home/david/NMMB/19677/src' is not a working copy
gmake: *** No rule to make target `/home/david/NMMB/19677_libs/esmf_3_1_0rp5/lib/libO/Linux.gfortran.64.mpich2.default/esmf.mk'. Stop.
I'm using gfortran. Can you suggest a solution?

Thanks,
David

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: NEMS NMMB topic

Post by meteoadriatic » Sun Feb 14, 2016 9:39 pm

Well, do you have this file? /home/david/NMMB/19677_libs/esmf_3_1_0rp5/lib/libO/Linux.gfortran.64.mpich2.default/esmf.mk

Also I believe that ESMF 3.1 won't work. If it doesn't produce emsf.mod file you are screwed, it is mandatory to have that module as far as I know, at least for NEMS v0.9. ESMF 6.2 works OK for me.

Post Reply

Who is online

Users browsing this forum: Baidu [Spider] and 1 guest