real.exe does not create wrflowinp_d02 when D2 start is delayed.

[ahahmann]

I am trying to run a simulation using sst_update=1 with two domains. There are no problems when starting the run in the usual way. But I have to drastically reduce the time step to get it through the first few minutes of the run. So, I thought I would try to delay the start of D2 by a few hours. When using a later time for D2 and input_from_file=.false. the wrflowinp_d02 file is not generated by real.exe. Thus WRF crashes when it is time to update the SSTs in D2.
Has anyone else encountered the same problem? Any possible solutions?
I enclose the relevant bit of the namelist.input.
Any help will be appreciated.
Thanks, Andrea

 

[Ming Chen]

Hi, Andrea,
Once you set input_from_file = false, the input data for D02 will be derived from the input data for D01. In this case REAL will not produce wrfinput_d02 and wrflowinp_d02. Pease try the following steps and let me know whether they work for your case:
(1) set input_from_file = true and let the two domains start at the same time. In this way you will be able to obtain wrfinput and wrflowinp for both domains.
(2) change input_from_file = false and specify different starting time for the two domains.
Then rerun wrf.exe.
This is just a possible workaround and hope it works.

 

[ahahmann]

Hi, Ming,

Thanks for the prompt answer.

Yes, that workaround works, except that I had to remove some initial times (until the D2 domain starts) from the wrflowinp_d02. That can be quickly done with, e.g., ncks.
Too bad there is not a more straightforward solution. It makes my launch script a lot more complicated.
Setting up the SST files is also a problem with restart runs with multiple domains and adaptive time steps. But that is another problem to solve.

Thanks again
Andrea

 

[Ming Chen]

Andrea,
Thanks for the update. I agree that such a solution is kind of clumsy. The model developers cannot consider all circumstances and users have to make changes based on their specific cases.