Questions about aborted calculations resulting from WRFDA's 4DVAR calculations

[stakahashi]

Hello to all WRF users!

I would like to build an environment that can calculate 4DVAR for WRFDA, but I am having trouble with errors.

After building v4.4 WRF and WRFDA cloned from github, I have confirmed up to the following 4DVAR tutorial works.
(WRFDA Practice)

In the tutorial, the generated wrfinput_d01 was distributed as a dataset, so I generated a new wrfinput_d01 using WRF ARW's real.exe and used it in 4DVAR, but the calculation was terminated by SIGABRT.

I would like to hear from those who have already built an environment that can calculate WRFDA.

Thank you in advance.

 

[sunhh]

You can see that many people in the forum are troubled by the use of 4DVAR. Unfortunately, I tell you that this problem of yours is very easy to appear in the 4DVAR of WRFDA. Through my many experiments, I found that if the physical suite is simple, it will have a higher probability of successful operation, and it is only for routine observation. Once the assimilation of radiation data is added, it will be very difficult. I asked for help many times in the forum, but I didn't get any help. Finally, I gave up the use of 4DAVR.

 

[stakahashi]

You can see that many people in the forum are troubled by the use of 4DVAR. Unfortunately, I tell you that this problem of yours is very easy to appear in the 4DVAR of WRFDA. Through my many experiments, I found that if the physical suite is simple, it will have a higher probability of successful operation, and it is only for routine observation. Once the assimilation of radiation data is added, it will be very difficult. I asked for help many times in the forum, but I didn't get any help. Finally, I gave up the use of 4DAVR.

Thanks for your response! I have been looking at the Forum, and I see that this is still a common and unsuccessful case.... I wonder about the future of WRF development as well.... I hope you will let me know if you find out anything else! I'll make a little more of an effort!

 

[sunhh]

I can tell you that if I want to run 4DVAR successfully, I need to make physics setting very very simple. For example, mp_physics=3,cu_physics=0. Even if 4dvar is successfully run, the experiment is meaningless to me becausethe simple physics settings are very poor for my case. There are other limitations, like hybrid_opt must be 0，use_theta_m must be 0 . Too many limitations make the use of 4DVAR very inconvenient, ineffective and not very meaningful.

 

[sunhh]

I can tell you that if I want to run 4DVAR successfully, I need to make physics setting very very simple. For example, mp_physics=3,cu_physics=0. Even if 4dvar is successfully run, the experiment is meaningless to me becausethe simple physics settings are very poor for my case. There are other limitations, like hybrid_opt must be 0，use_theta_m must be 0 . Too many limitations make the use of 4DVAR very inconvenient, ineffective and not very meaningful.

I mean the 4DVAR in WRFDA。

 

[stakahashi]

I mean the 4DVAR in WRFDA。

I appreciate your valuable comments. I understood that the model needs to be simplified, and that's what you meant by the settings. I will check in our configuration file and try to implement it. It is true that oversimplifying the model makes no sense from the standpoint of feasibility, which is a problem. It would only make sense to try out a sensitivity study for the time being.... If you notice anything else, I would be glad to hear from you!