[Wrf-users] Different results for 2 SGI clusters (WRFv3.1)

Mikhail Titov TitovM at ap.aurecongroup.com
Mon May 17 20:37:26 MDT 2010

Dear WRF Help and Users,

We bought new SGI cluster and I transferred all WRF versions and jobs from 
old SGI cluster on a new one.

After installation of WRF on new SGI cluster I ran several trial jobs for 
WRFv2.2, WWRFv3.0 and WRFv3.1
to be sure that WRF output is completely consistent with WRF output from 
the old SGI cluster.

Configuration of both clusters is the same and I used 1cpu and multi-cpu 
WRF trials to check that 
both clusters produce the same results. 

I've got exactly the same results for WRFv2.2 and WRFv3.0 and was very 
surprised to receive very
close (correlation coefficient about 0.99) but still different results for 
WRFv3.1 (for 1 cpu and for multi-cpu runs)
on a new cluster via the old one.

WRFv3.1 has the same configuration for old and new SGIs, the same 
'namlist.input' and LANDUSE and VEGPAR tables.
Also the input terrestrial files (height and landuse) and input synoptic 
files are the same.
But for WRFv3.1 I still have a bit different results on old and new SGI 

Has anyone got such a problem before and why does it happen with WRFv3.1 
I have no logical explanation to this difference between 2 SGIs with 
exactly the same software (WRFv3.1 only).

Any help is highly appreciable.

Thanks a lot,
Dr. Mikhail Titov | Energy  | Aurecon 
Ph: +64 3 366 0821 ext.9231 DDI +64 367 32 31 | Fax: +64 3 379 6955 | Mob: 
+64 21 106 5563 
Email: TitovM at ap.aurecongroup.com 
PO Box 1061, 122 Gloucester Street | Christchurch 8140 | New Zealand 
Please consider your environment before printing this e-mail.

Disclaimer - http://www.aurecongroup.com/apac/disclaimer/ 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/wrf-users/attachments/20100518/396494a7/attachment.html 

More information about the Wrf-users mailing list