[RegCNET] Time step problem or something else ?

Rauscher Sara A srausche at ictp.it
Mon Jan 8 10:10:07 CET 2007


Hi Dipil,

During a long run this can sometimes happen. Usually I restart (before
the time of the crash) using a shorter time step in regcm.in. When the
simulation moves past the time of the crash, I then stop the simulation
and restart with the original time step. This is a way for dealing with
one or two crashes in a long simulation.

If you are experiencing lots of crashes, this is NOT an appropriate way
to handle it - you should follow Jagadish's suggestions regarding
checking the time step, location and boundaries of the domain, model
resolution, and the physical parameterizations etc. and do further tests
before beginning a long simulation.

Let me know if you require further information.

cheers,
sara

jagadish karmacharya wrote:
> Dear Dipil,
>
> You shouldn't change parameters at the middle of an experiments. My
> suggestion was to make the changes if you are starting new experiments.
>
> Have you checked the ICBC file for that date/month in grads?
>
> One thing you can do is to run new identical simulation begining from
> the  1st date of the month where you are getting problem  just to
> check if the problem persist or not.  You may also try  to restart the
> experiment with different convection scheme just for the purpose of
> checking. Obviously these will not be part of you long simulation.
>
> All the best
>
> Jagadish
>
> */Dilip Ganguly <dganguly at prl.res.in>/* wrote:
>
>     Dear Jagadish,
>
>     As you are suggesting that changing domain boundary, convective
>     scheme,
>     increasing resolution etc. might help in avoiding this kind of
>     situation.
>     But my problem is that I am already in the middle of a long
>     simulation and
>     this problem due to convective instability is occuring towards the
>     end of
>     my simulation. At this stage how can I introduce changes which you
>     want me
>     to try ? Is there any other way to get out of this problem at this
>     stage ?
>
>     Regards,
>
>     Dilip.
>
>
>     > Dear Dilip,
>     >
>     > You have tried all the way down to dt=1 sec but as you have seen
>     > decreasing time step doesn't always solve the problem. I generally
>     > decrease dt upto 1.5*resolution in km (instead of dt=3*res) to
>     check if
>     > if overcome the problem.
>     > When the problem persists with decreased timestep as well it is
>     difficult
>     > to figure out a solution.
>     >
>     > Changing domain boundary, convective scheme, increasing
>     resolution may
>     > help but if one has already completed long simulation then it
>     become a
>     > problem. One possibility is that your ICBC or driving data file
>     for that
>     > month may be corrupt but its less likely.
>     >
>     > Can anyone give some suggestion?
>     >
>     > Regards
>     > Jagadish
>     >
>     >
>     > Dilip Ganguly wrote: Dear all,
>     >
>     > I am facing problem in running RegCM3. The program hangs and
>     does not seem
>     > to work after a particular time and I am getting messages on the
>     screen
>     > like the following.
>     >
>     > BATS variables written at 2004120121 180.000000000000
>     > at day = 396.8762, ktau = 210200 : 1st, 2nd time deriv of ps =
>     > 0.10915E-02 0.49528E-05, no. of points w/convection = 81
>     > at day = 396.8849, ktau = 210250 : 1st, 2nd time deriv of ps =
>     > 0.97549E-03 0.41641E-05, no. of points w/convection = 71
>     > at day = 396.8936, ktau = 210300 : 1st, 2nd time deriv of ps =
>     > NaN NaN , no. of points w/convection = 0
>     >
>     >
>     -------------------------------------------------------------------------
>     >
>     > It do not know if it is due to improper choice of time steps. I have
>     > restarted the program several times with lower and lower values
>     of time
>     > step dt but every time the program hangs around a certain period
>     of time
>     > and not able to cross this period. I have tried with numerous
>     combinations
>     > of gradually decreasing time steps. Some are given below.
>     >
>     > radfrq= 30 30 30 30 15 15 15
>     > abemh = 18 18 18 18 9 9 9
>     > abatm = 300 120 90 45 30 9 3
>     > dt = 100 60 30 15 10 3 1
>     >
>     > But none of them worked. Everytime it shows the same problem
>     possibly due
>     > to convective instability as told by Jagadish. I am using a 90 km
>     > horizontal resolution in my simulation.
>     >
>     > Can anyone suggest a way to solve this problem ?
>     >
>     > Thanking you in advance.
>     >
>     > Dilip.
>     >
>     >
>     ---------------------------------------------------------------------
>     > DILIP GANGULY
>     > Postdoctoral Fellow
>     > Space And Atmospheric Sciences Division
>     > Physical Research Laboratory
>     > Ahmedabad 380009
>     > Gujarat, India
>     > Phone: +91 79 26314558 (Laboratory)
>     > +91 79 26300505 (Hostel)
>     > Fax: +91 79 26314659
>     > Alternate Email: ganguly.dilip at gmail.com
>     > Visit me at http://www.prl.res.in/~dganguly
>     >
>     ---------------------------------------------------------------------
>     >
>     >
>     > _______________________________________________
>     > RegCNET mailing list
>     > RegCNET at lists.ictp.it
>     > https://lists.ictp.it/mailman/listinfo/regcnet
>     >
>     >
>     > __________________________________________________
>     > Do You Yahoo!?
>     > Tired of spam? Yahoo! Mail has the best spam protection around
>     > http://mail.yahoo.com
>
>
>     -- 
>     Dilip Ganguly
>     Postdoctoral Fellow
>     Space and Atmospheric Sciences Division
>     Physical Research Laboratory
>     Navrangpura
>     Ahmedabad-380009
>     Gujarat, India
>     Phone:+91 79 26314558, Fax: +91 79 26314659
>     Alternate Email: ganguly.dilip at gmail.com
>     visit me at: http://www.prl.res.in/~dganguly
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> RegCNET mailing list
> RegCNET at lists.ictp.it
> https://lists.ictp.it/mailman/listinfo/regcnet
>   

-- 

Dr. Sara A. Rauscher
Earth System Physics Section 
The Abdus Salam International Centre for Theoretical Physics
Strada Costiera 11 
34014 Trieste ITALY
ph: +39 040 2240 225 fax: +39 040 2240 449
email: srausche at ictp.it





More information about the RegCNET mailing list