WRFV3.6.1 - Adaptive Time Step no longer effective?

Compiling NMM/ARW code for EMS system, upgrading WRF cores on your own, changing EMS scripts to suit your particular needs, and other modifications to original EMS distribution goes in this forum. These are officially unsupported actions.
oib
Posts: 117
Joined: Mon Jan 31, 2011 3:29 pm

WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by oib » Mon Aug 25, 2014 2:49 pm

I have compiled different version of WRF 3.4.1, 3.5, 3.5.1, 3.6 to the newer 3.6.1 and I'm comparing the various versions with regard to the processing speed.
And I'm also comparing the above with WRF-EMS 3.4.1.

I have compiled the various WRF versions using last Intel compilers, same libraries and MPI, same configure.wps, configure.wrf options (in particular Linux x86_64 i486 i586 i686, [Xeon (SNB with AVX mods)] ifort compiler with icc (dmpar)).

I'm running a test domain, ARW core, adaptive time step is enabled with these entries in namelist.input:

Code: Select all

time_step = 30,
time_step_fract_num = 0,
time_step_fract_den = 10,
time_step_dfi = 60,
use_adaptive_time_step = T,
adaptation_domain = 1,
step_to_output_time = T,
target_cfl  = 1.2,
target_hcfl = 0.84,
max_step_increase_pct = 5,
starting_time_step = 20,
(this is the domain I actually run on WRF-EMS)

Domain resolution is 4 km.

A strange thing happens: WRFV3.6.1 behaves differently from the others with regard to the setting of the time step.

Looking into rsl.out.0000 file I see that the version 3.6.1 always sets a very small time step (average is 15 s).
Instead older versions often use "smarter" (=greater) values (often I see 60 s).

The consequence is that the execution of wrf.exe version 3.6.1 requires 4 times the time it takes for the older versions :cry:

I do not know why this adaptive time step has suffered "loss of intelligence" in version 3.6.1. :?

Does anyone can help?

smartie
Posts: 94
Joined: Sat May 21, 2011 7:34 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by smartie » Tue Aug 26, 2014 6:59 pm

I was going to upgrade from 3.6.0 to 3.6.1.
Have you tried setting default values in the namelist?

Code: Select all

 max_step_increase_pct      = 5, 51
 starting_time_step         = -1, -1
 max_time_step              = -1, -1
 min_time_step              = -1, -1

oib
Posts: 117
Joined: Mon Jan 31, 2011 3:29 pm

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by oib » Wed Aug 27, 2014 8:23 am

smartie wrote:I was going to upgrade from 3.6.0 to 3.6.1.
Have you tried setting default values in the namelist?

Code: Select all

 max_step_increase_pct      = 5, 51
 starting_time_step         = -1, -1
 max_time_step              = -1, -1
 min_time_step              = -1, -1
Hi smartie,
I have tried with the default values.
Unfortunately, the time step is always set to very small values​​, with the result that wrf.exe V3.6.1 needs 2-3x the time taken by earlier versions :?
Last edited by oib on Wed Aug 27, 2014 9:02 am, edited 1 time in total.

oib
Posts: 117
Joined: Mon Jan 31, 2011 3:29 pm

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by oib » Wed Aug 27, 2014 8:58 am

A chart made with a simple script that ingest rsl.out.0000 files, helps to clarify the situation ...

Image

smartie
Posts: 94
Joined: Sat May 21, 2011 7:34 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by smartie » Wed Aug 27, 2014 11:23 am

The 3.4.1 trace (red) looks wrong here as there seems to be very little variation (in relative terms). However, I agree the I'd expect the 3.6.1 to be more economical than is shown by the blue trace. I'd report this as a possible bug in 3.6.1 to wrf_help, if it isn't a bug maybe they can suggest an explanation.
Unfortunately I don't have time ATM to update and confirm the problem, but will do so in due course.

meteoadriatic
Posts: 1512
Joined: Wed Aug 19, 2009 10:05 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by meteoadriatic » Wed Aug 27, 2014 1:04 pm

To me it doesn't look wrong, because there is a maximum time step limit that if reached, it will not go over it (in the picture it is 60 seconds). Model will shorten that time step if finds CFL less stable.

However, about 3.6.1 it is wierd. I have no explanation. Didn't tried it myself yet.

oib
Posts: 117
Joined: Mon Jan 31, 2011 3:29 pm

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by oib » Wed Aug 27, 2014 2:22 pm

meteoadriatic wrote:To me it doesn't look wrong, because there is a maximum time step limit that if reached, it will not go over it (in the picture it is 60 seconds). Model will shorten that time step if finds CFL less stable.

However, about 3.6.1 it is wierd. I have no explanation. Didn't tried it myself yet.
I'm agree with you meteoadriatic, the flat 3.6.1 line is due to the upper limit imposed (60 s) in the namelist.input to the time step.
I have tried incrementing the vertical CFL limit value from 1.2 (default) to 2.2 and now 3.6.1 run with a lower time step (around 30 s). Don't know if this could be a solution...
Will report to wrf_help.
I have asked also on the official forum, but my question has not been yet published...
W wrf-ems forum :D

meteoadriatic
Posts: 1512
Joined: Wed Aug 19, 2009 10:05 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by meteoadriatic » Wed Aug 27, 2014 3:32 pm

Try to increase CFL limit more and see how it runs. If it still runs stable... watch also the output in that configuration, if you see noise patterns in for example geopotential heights in ares with high winds then it really runs near CFL limits before crashing.

smartie
Posts: 94
Joined: Sat May 21, 2011 7:34 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by smartie » Wed Aug 27, 2014 7:20 pm

Ah -blindingly obvious now!
:oops:
I suppose you could look for a code change by diffing solve_em for 3.6.0 vs 3.6.1
But I suspect the problem may be across modules,else it would have been picked up. Perhaps something has been changed in one of the modules for your physics options which affects things?

smartie
Posts: 94
Joined: Sat May 21, 2011 7:34 am

Re: WRFV3.6.1 - Adaptive Time Step no longer effective?

Post by smartie » Fri Nov 07, 2014 10:05 am

Just got round to upgrading to 3.6.1.
Can't see any problems with the adaptive time step.
Pretty standard compile with iFort 13.x , netCDF 4 etc. Got a Fortran 2003 warning.

Did you find a solution?

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests