WRFDA and RTTOV

Setting up the 3DVAR package within WRF.

Re: WRFDA and RTTOV

Postby hnlim » Mon Jul 12, 2010 10:32 am

Hi Thomas,

Now I remembered. Yes wrfhelp did tell me about the problem. That is why I had compiled the code serially only.
Yes the negative cost function is what I get too though after many more iterations.
I had tried emailing wrfhelp regarding this. They had asked me to test with AMSU data as well as using only water pixels, which I did and none of them had any problem. It happened only to AIRS.
After which I had no more replies from wrfhelp regarding this issue.
Did you get any reply from wrfhelp regarding this issue?
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: WRFDA and RTTOV

Postby thomas2 » Mon Jul 26, 2010 8:07 am

After some test and emails to NCAR, I think the error is found:

In "WRFDA/var/da/da_radiance/da_rttov_tl.inc " and "WRFDA/var/da/da_radiance/da_rttov_ad.inc" the statement ".or. coef%id_sensor == 3" needs to be added around line 212 in TL and line 228 in the ADJ.

Thomas
thomas2
 
Posts: 75
Joined: Mon Sep 01, 2008 4:56 am

Re: WRFDA and RTTOV

Postby hnlim » Mon Jul 26, 2010 11:32 am

Hi Thomas,

Thanks for the information. I will try it to see if it fixed the problem on my side.

Thanks
Agnes
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: WRFDA and RTTOV

Postby hnlim » Mon Aug 02, 2010 5:27 pm

Hi Thomas,

Could I just check with you, usually how many iterations do u need to reach a convergence when u ran with AIRS data?

Agnes
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: WRFDA and RTTOV

Postby thomas2 » Tue Aug 03, 2010 6:28 am

Hi,

in my test case the number of iterations is 71 compared to 14 without radiances (~1000 additional obs).
I use version 3.2.

Thomas
thomas2
 
Posts: 75
Joined: Mon Sep 01, 2008 4:56 am

Re: WRFDA and RTTOV

Postby hnlim » Tue Aug 03, 2010 10:24 am

Hi Thomas,

Did you use the be file that was provided or you own be file?

Agnes
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: WRFDA and RTTOV

Postby thomas2 » Tue Aug 03, 2010 10:48 am

Hi Agnes,

do you mean be.dat? I use my own.

Thomas
thomas2
 
Posts: 75
Joined: Mon Sep 01, 2008 4:56 am

Re: WRFDA and RTTOV

Postby hnlim » Tue Aug 03, 2010 10:56 am

Hi Thomas,

Yes/ apologies for the confusion.

I had asked because I felt that in my test case, its taking too many iterations to converge.
From the summary report printed during assimilation, in my local domain there were 429 observations, but I am not getting convergence even with 200 iterations.
I am trying to find out what is the fault.

Have you had such experience or be able to provide some suggestions? I did not change any of the convergence criterion.

Agnes
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

Re: WRFDA and RTTOV

Postby thomas2 » Tue Aug 03, 2010 11:09 am

The only thing I notice that it takes around 100 iterations when using radar data. I never tested it with the B matrix provided for the test case.

Maybe your background field is "not the best"?

That's the part of my namelist:
Code: Select all
RTMINIT_NSENSOR = 12
RTMINIT_PLATFORM = 1, 1, 1, 9,10, 1, 1, 1, 1,10, 9, 2
RTMINIT_SATID = 15,16,18, 2, 2,15,16,17,18, 2, 2,16
RTMINIT_SENSOR = 3, 3, 3, 3, 3, 4, 4, 4,15,15,11,10
thinning_mesh = 80.0,80.0,80.0,80.0,80.0,80.0,80.0,80.0,80.0,80.0,80.0,80.0,
thinning = .true.
qc_rad = .true.
write_iv_rad_ascii = .true.
write_oa_rad_ascii = .true.
rtm_option = 1
only_sea_rad = .true.
use_varbc = .true.


Thomas
thomas2
 
Posts: 75
Joined: Mon Sep 01, 2008 4:56 am

Re: WRFDA and RTTOV

Postby hnlim » Tue Aug 03, 2010 11:46 am

Hi Thomas,

I have only assimilated AIRS data.
Initially I used the gen_be utility to generate the be.dat using 1 month worth of WRF forecast but that too also had the same issue.
I too set the use_varbc to true and by doing so I actually see a significant increase of iterations required compared to the case when use_varbc is set to false.
From your namelist that you have attached, I see that I had used much less data because my thinning mesh is at 120km. Thus if your test case could reach convergence in less than a 100 iterations, I would expect mine will be so too since I have less observations.

Agnes
hnlim
 
Posts: 74
Joined: Wed Apr 16, 2008 11:51 am

PreviousNext

Return to Installation

Who is online

Users browsing this forum: No registered users and 1 guest