Search found 91 matches

by alfe
Wed Dec 31, 2014 6:19 pm
Forum: EMS Hacking
Topic: Using GFS 0.25° grid dataset
Replies: 54
Views: 46460

Re: Using GFS 0.25° grid dataset

Hi meteoadriatic, I have downloaded the following file for model initialization : /pub/data/nccf/com/gfs/para/gfs.2014123112/gfs.t12z.pgrb2.0p25.f000 from server ftp.ncep.noaa.gov. But metgrid crashes. This is the error copied from the crash log : ERROR: The mandatory field TT was not found in any i...
by alfe
Tue Dec 30, 2014 11:10 pm
Forum: EMS Hacking
Topic: Using GFS 0.25° grid dataset
Replies: 54
Views: 46460

Re: Using GFS 0.25° grid dataset

Hello,
For me it doesn't work.
The gfs-para file for initialization has only 180 MB, not 230 MB, and temperature at 2m is missing so metgrid crashes... :shock:
I have downloaded it at NCEP ftp location.
by alfe
Sat Mar 01, 2014 9:16 am
Forum: EMS v3.4 support
Topic: ems_run execution time extremely variable
Replies: 28
Views: 9362

Re: ems_run execution time extremely variable

but it seems to me that numad -K 1 -u 800 works for me.
Unfortunately, I don't get better results with above command line...
Maybe the harware configuration plays an important role here, so it seems to me that there are only individual cases...
by alfe
Fri Feb 21, 2014 11:31 pm
Forum: EMS v3.4 support
Topic: ems_run execution time extremely variable
Replies: 28
Views: 9362

Re: ems_run execution time extremely variable

Well,
I have to say that the only good configuration (for me :| ), is 15/16 cores without numad. Otherwise it is chaotic : +50% runtime, sleeping cores... :cry:

I would prefer to play with NUMTILES... in the ncpu conf file.
by alfe
Thu Feb 20, 2014 6:51 pm
Forum: EMS v3.4 support
Topic: ems_run execution time extremely variable
Replies: 28
Views: 9362

Re: ems_run execution time extremely variable

I see...
I run with 15/16 cores because ems doc tells to do so. 1 core must be left otherwise it slows down the simulation. I will try with all the cores and numad active. There is a fine tuning of numad but it is very technical... :?
by alfe
Wed Feb 19, 2014 9:49 pm
Forum: EMS v3.4 support
Topic: ems_run execution time extremely variable
Replies: 28
Views: 9362

Re: ems_run execution time extremely variable

Hello guys, I have tried with or without numad, but I found that numad slows down the run. Anyway, on my system it was deactivated by default. So I just step back to the "standard" service configuration by switching it off. I have also seen that at the beginning of the run (without numad), sometimes...
by alfe
Mon Jan 13, 2014 9:48 am
Forum: General NWP, Postprocessing, Meteorology
Topic: Unrealistic Snow cover on ground for GFS
Replies: 8
Views: 3900

Re: Unrealistic Snow cover on ground for GFS

Hello alfe, I am using 45 vertical levels I was thinking about something quite odd. I have recently tried with 71 levels (DX = 3km), but I got very cold local temperatures, like a cold droplet inside a warmer field. Then, I have tried with 61 levels and it was much better.... But if you say that yo...
by alfe
Sat Jan 11, 2014 4:48 pm
Forum: General NWP, Postprocessing, Meteorology
Topic: Unrealistic Snow cover on ground for GFS
Replies: 8
Views: 3900

Re: Unrealistic Snow cover on ground for GFS

Hello meteosciez,
Interesting comparison. How many vertical levels are you using ?
by alfe
Sat Nov 30, 2013 10:09 am
Forum: General NWP, Postprocessing, Meteorology
Topic: Unrealistic Snow cover on ground for GFS
Replies: 8
Views: 3900

Unrealistic Snow cover on ground for GFS

Hello guys, I am referring to a post of this section (http://www.wrfems.info/viewtopic.php?f=6&t=295#p3052)concerning what is explained in the title of my present topic . I propose to open a specific discussion on this post, to avoid interferences with another subject. This is an important matter an...