[RegCNET] Restart Run - 'NaN'

Moetasim mashfaq at purdue.edu
Mon Feb 27 22:00:09 CET 2006


In theory, there should not be any difference between two runs, a continuous run 
and run restarted using RegCN SAV file. So, your restart run should not give 
problems for that month if run without restart goes fine in those dates. 
Anyway, the science of machine arcitecture is sometimes beyond logic..This might 
be beacause of 0.001% difference between the two runs, but again, even this 
difference is not logical.. 
My suggestion is that you should run the problem dates with smaller time step.. 
and it will solve your problem. Try dt=100 and if even this doesn't work, try 
dt=80.. once you are through these dates, restart RegCM with normal time step.

Moet   
Quoting Srivatsan Raghavan <s.raghavan at ucl.ac.uk>:

> 
> 
> [Please ignore the earlier one..that was incomplete - Apologies]
> 
> Hi all,
> 
> I have been trying to do a restart run fro DJF 1991-92. The problem I have
> is
> the 'NaN error ..
> 
> The  set of idates for the first run are :
> 
> 'false'
> idate0 = 1991120100
> idate1 = 1991120100
> idate2 = 1992010100          SAV  written :  SAV.1992010100
> 
> The set of idates for Restart I
> 
> 'true'
> idate0 = 1991120100
> idate1 = 1992010100
> idate2 = 1992020100          SAV written : SAV.1992020100
> 
> The set of idates for Restart II where I have problems:
> 
> 'true'
> idate0 = 1991120100
> idate1 = 1992020100
> idate2 = 1992022800
> 
> -----------------------
> 1. dx  from Domain is 40km
> 2. dt = 120 ; abatm = 360
> -----------------------
> 1. I have checked for the input file for ICBC1992020100 ..It is complete and
> fine and I am able to display all variables including 'ps'
> 
> 2. Since the other restart worked fine, I guess the model did NOT crash
> because
> of the 'dt' and 'abatm' values. I think they are Ok.. {??}
> 
> The run for Feb 1992 is fine and there is NO problem when it is NOT a
> restart.
> The 'NaN' comes only during restart and the run hangs only on the dates -
> 1992020218 - 1992020300..
> 
> I have had the same problem occuring in the exact 2 and 3rd day in that
> month,
> despite running it several times.
> 
> 
> Any suggestions,
> 
> Thanks
> 
> Sri Raghavan
> Dept. of Geography
> UCL
> _______________________________________________
> RegCNET mailing list
> RegCNET at lists.ictp.it
> https://lists.ictp.it/mailman/listinfo/regcnet
> 




More information about the RegCNET mailing list