Adaptive timestepping problem

[leuthold]

I've been having an intermittent problem with the adaptive timestepping option stopping the WRF simulation way before it's suppose to finish. I get the SUCCESS COMPLETE WRF message only after a couple of hour of simulation time. What is interesting and must have something to do with it, is that the last dt is negative.
Timing for main (dt= 13.96): time 2014-02-07_14:59:46 on domain 2: 0.17649 elapsed seconds
Timing for main (dt= 13.96): time 2014-02-07_14:59:46 on domain 1: 0.59860 elapsed seconds
Timing for main (dt= 14.49): time 2014-02-07_15:00:00 on domain 2: 0.16762 elapsed seconds
Timing for main (dt= 14.49): time 2014-02-07_15:00:00 on domain 1: 0.58959 elapsed seconds
Timing for Writing wrfout_d01_2014-02-07_15:00:00 for domain 1: 2.17174 elapsed seconds
Timing for Writing wrfsolar_d01_2014-02-07_15:00: for domain 1: 8.48620 elapsed seconds
Timing for processing lateral boundary for domain 1: 3.06879 elapsed seconds
Timing for Writing wrfout_d02_2014-02-07_15:00:00 for domain 2: 2.27773 elapsed seconds
Timing for Writing wrfsolar_d02_2014-02-07_15:00: for domain 2: 7.97761 elapsed seconds
Timing for main (dt= -0.54): time 2014-02-07_15:00:00 on domain 1: 25.50731 elapsed seconds
d01 2014-02-07_15:00:00 wrf: SUCCESS COMPLETE WRF

Slightly changing one of the target_cfl options will usually fix the problem, but that is not the solution I'm looking for as that has to be done manually and WRF rerun. Bug or am I missing something? step_to_output_time and adjust_output_times are set to true.

thanks.

[Phillip]

What is interesting and must have something to do with it, is that the last dt is negative.

Setting step_to_output_time=true makes the model to go exactly to the time when the output is supposed to occur. This seems to include going backwards if the output time is missed.

But this should not make you simulation stop. Check carefully if your end dates and run times are set correctly. And check them both, i think if they don't match the model will stop at whichever time comes first. Post your namelist here if you are not sure what to do.

[leuthold]

What is interesting and must have something to do with it, is that the last dt is negative.

Setting step_to_output_time=true makes the model to go exactly to the time when the output is supposed to occur. This seems to include going backwards if the output time is missed.

But this should not make you simulation stop. Check carefully if your end dates and run times are set correctly. And check them both, i think if they don't match the model will stop at whichever time comes first. Post your namelist here if you are not sure what to do.

The end date and run_hours are set correctly. If I change target_cfl slightly (doesn't matter which way) I can get it to run. As of now, I've seemed to of solved the problem this way as I've increased it by .05. Thanks for the reply.
Mike

[MeteoAdriatic]

Experiencing same problem with 3.8.1.

Any clues up to now?

Bye