해결
http://mailman.ucar.edu/pipermail/wrf-users/2011/002486.html
아래와 같이 만들어 주면,
Ob too early 라는 메세지는 사라진다. In rsl.out.0000
그러나, n = 21 unknown ob of type FM-88 SATOB 는 여전히 남는다. 여튼 결과는 ws10m 가 별 변화가 없다.
Sounds like you have 2 issues:
1. OBSGRID.EXE is not writing output at the times you want
2. WRF-Chem is bombing when nudging
I posted a bit on problem 1 - I was misinterpreting the DOCs (or the
DOCs are misleading, depending on your viewpoint). See
http://mailman.ucar.edu/pipermail/wrf-users/2011/002424.html.
Basically, set &share::interval_seconds to match your input GRIB data
interval (must match the interval you ran METGRID.EXE with). This is
21600 if your GRIB files come every 6 hours.
Then make sure your little_r format files contain data every
&record7::intf4d (from your email, it looks sorta like you want to nudge
every 3 hours). The main point is that OBSGRID.EXE will look for a new
file every intf4d seconds, and will not look in a file that's already
been opened (from a previous interval).
I'm nearly certain that you must combine both your surface and upper-air
data into the same file, but that's pretty easy - you can just "cat"
them together.
I suggest you fix problem 1 first, so you have non-empty OBS_DOMAIN1??
files for the in-between hours, before addressing the bombing of WRF.
Bart Brashers
From: wrf-users-bounces at ucar.edu [mailto:wrf-users-bounces at ucar.edu] On
Behalf Of Chris Klich
Sent: Sunday, October 30, 2011 12:31 PM
Subject: [Wrf-users] Help with Obs Nudging Crash
Hi all, I have recently been trying to use obs-nudging with my WRF run.
I am currently running WRF-Chem version 3.2.1 due to a long-term project
and unable to upgrade to 3.3.
I currently have both upper air and surface data in little-r format
(separate however). I convert these files into the proper naming format
for OBSGRID, run obsgrid and move the files to the /run directory,
concatenate all the OBS_DOMAIN files into OBS_DOMAIN101, and then run
real.exe and wrf.exe. However, when OBSGRID runs, the output files are
every 3 hours, while my met_em files and observation files are every 6.
These 3 hour intermediate times (3, 9, 15, 21) are empty when the
OBS_DOMAIN and plotobs and qc files are outputted, while the normal
intervals of 0, 6, 12, 18 are several megabytes. Even in this case, I
had tried concatenating all files to OBS_DOMAIN101 and running WRF.
However, it seems the entire run finishes only when I use just upper air
data specifically. When I run using just surface obs or concatenating
surface to upper air, the run crashes, and it seems at a very random
time. When using just surface it crashed after about 12:49 into the
run, while running with the concatenated upper air/surface, it crashed
after about 3 days, 12:49. I can't seem to figure out what is causing
this. The end of my rsl.error.0000 file looks like:
Timing for main (dt=135.00): time 2008-05-20_12:45:13 on domain 1:
1.08750 elapsed seconds.
Timing for main (dt=135.00): time 2008-05-20_12:47:28 on domain 1:
1.05240 elapsed seconds.
OBS NUDGING: Reading new obs for time window TBACK = 12.125 TFORWD =
13.458
for grid = 1
OBS NUDGING: 1 previously read obs are now too old for the current
window a
nd have been removed.
****** CALL IN4DOB AT KTAU = 326 AND XTIME = 767.48: NSTA =
43 ****
**
++++++CALL ERROB AT KTAU = 326 AND INEST = 1: NSTA = 43 ++++++
OBS NUDGING FOR IN,J,KTAU,XTIME,IVAR,IPL: 1 10 326 767.48 3 3
rindx= 5.
3
OBS NUDGING FOR IN,J,KTAU,XTIME,IVAR,IPL: 1 10 326 767.48 4 4
rindx= 5.
3
OBS NUDGING FOR IN,J,KTAU,XTIME,IVAR,IPL: 1 10 326 767.48 1 1
rindx= 5.
3
OBS NUDGING FOR IN,J,KTAU,XTIME,IVAR,IPL: 1 10 326 767.48 2 2
rindx= 5.
3
Timing for main (dt=135.00): time 2008-05-20_12:49:43 on domain 1:
1.07960 elapsed seconds.
forrtl: error (78): process killed (SIGTERM)
Image PC Routine Line
Source
libc.so.6 00000033C34DD1E3 Unknown Unknown
Unknown
libopen-pal.so.0 00002B47160653DD Unknown Unknown
Unknown
libopen-pal.so.0 00002B471606213D Unknown Unknown
Unknown
libopen-pal.so.0 00002B471605555C Unknown Unknown
Unknown
libmpi.so.0 00002B4715B43AA8 Unknown Unknown
Unknown
libmpi.so.0 00002B4715B721AC Unknown Unknown
Unknown
wrf.exe 0000000001C64282 Unknown Unknown
Unknown
wrf.exe 0000000000F373F4 Unknown Unknown
Unknown
wrf.exe 00000000014F73B5 Unknown Unknown
Unknown
wrf.exe 0000000000E8EC4F Unknown Unknown
Unknown
wrf.exe 0000000000D8DB6D Unknown Unknown
Unknown
wrf.exe 000000000052ECDF Unknown Unknown
Unknown
wrf.exe 00000000004BC733 Unknown Unknown
Unknown
wrf.exe 00000000004BC6E7 Unknown Unknown
Unknown
wrf.exe 00000000004BC67C Unknown Unknown
Unknown
libc.so.6 00000033C341EC9D Unknown Unknown
Unknown
wrf.exe 00000000004BC579 Unknown Unknown
Unknown
In addition, when I use upper air data, I get thousands and thousands of
lines in my rsl.error.0000 file that read something like:
n= 9061 unknown ob of type FM-88 SATOB
Is this normal or is there an option I am missing where it will read
these correctly?
Any help for either of these problems would be greatly appreciated as I
need to figure this out ASAP.
I've also attached both namelist.input and namelist.oa for any help that
may provide.
Thank you.
--
Christopher Klich
Graduate Student
Florida State University
(908) 208-9743
Pasted from <http://mailman.ucar.edu/pipermail/wrf-users/2011/002486.html>