Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Segmentation Fault

Khin Myat

Hi ,
Now i am running the simulation with les-sfs option in the innermost domain. But the simulation stopped with segmentation fault error. Before that, I faced cfl error. When i reduced the time step from 6 to 1s and set up time_step_sound=8 in the innermost domain, the simulation stopped with segmentation fault error.
Is there any advice for this problem?
Any help or comments are appreciated.
I attached namelist and rsl error file here.

Thank you very much for your time in advance.



  • namelist.input
    11.9 KB · Views: 3
  • namelist.wps
    969 bytes · Views: 1
  • rsl.error.0000
    1.2 MB · Views: 3
  • rsl.error.0010.txt
    13.9 KB · Views: 2
Hi Khin,
Many apologies for the long delay in response. We have all been occupied with several other deadlines and have gotten behind with forum responses. Are you still experiencing issues with this? Since you have num_metgrid_levels = 34, I'm guessing you're using GFS input (probably 0.25 degree data). Is that correct? If so, you probably need an outer domain that has a coarser grid spacing than 3km. 0.25 degrees is about 27 km grid spacing, and going from that, straight to 3 km is about a 9:1 ratio, which is a bit large. We recommend that your first domain should have a ratio of about 3:1 or 5:1. Can you give that a try and see if it makes a difference?

Additionally, it looks like you're probably using some older WRF code. Can you update WRF to a newer version to see if that makes any difference? The most recently released version is V4.5.2.
Dear kwerner,
Thank you very much for your reply and advice.
Now I am using GFS(1 degree data every six hours). If so, if I run the simulation with 3km grid spacing, should I use higher resolution input data? And I will give it a try as your advice.
Now I am running the simulation with LES option in the innermost domain. Before this simulation, I run the simulation with Km-opt 2 and SFS option 1. At that time, the simulation can be run without error, When I change SFS option 2, I faced CFL error. So, I thought it may be due to changes of turbulence calculation.
So, does the problem come from input data resolution? Anyway , I will follow your advice and give it a try.
Thank you so much for helping me a lot.
Have a good day.

Best regard,
1) It is recommended to use the highest-available resolution input data you can. So if you want to use GFS, use the 0.25 degree data, as opposed to the FNL 1 degree data.
2) As I mentioned above, if you're interested in a 3km domain, even with 0.25 degree data, you need to have a parent domain around the 3km domain (so that the 3km domain will be the nest) - otherwise, the ratio between your 3km domain and the input data will be too large.
3) I'm not sure what SFS is. Can you tell me the specific namelist variable you are referring to?
Dear @kwerner,

I am so confused about two data sets NCAR RDA Dataset ds084.1 Data Access and I understood that ds084.1 data came from the Global Forecast System and ds083.3 came from NCEP GDAS FNL data. But there are many differences in the number of files in one day. For example, there are more than 300 files in one day in ds084.1. But there are only 12 files in ds083.3. If I would like to use ds084.1, should I download all data? Could you please explain the difference between the two data sets? Although I read, I could not understand. Which data sets do you mean in the previous comment?
"gfs.0p25.2019010100.f000.grib2" in that file name, what does it mean f000? I could not understand the file arrangement.

Thank you very much for your time in advance.
Have a good day.

Best regard,

f000 means that the data is at initial time of GFS, f003 means it is 3-hr forecast of GFS, and so on.
ds084.1 provides GFS forecast data, which could be 120-hr forecast at 3-hr interval output and the initial time can be at 00 UTC, 06 UTC, etc.
ds083.3 provide FNL data, which is global reanalysis dataset.
You are right that these two datasets are quyite different, although both can be used to drive WRF.
Dear @Ming Chen,

Thank you very much for your reply.
I understood a little more. But, I am still confused about the forecast timestep and forecast run. As you mentioned, in ds084.1, the forecast times step is 3hr for the first 240 hours and 12hr for the rest from 240 to 384. The forecast run is at each 6hr.
When we run the simulation, as the data provide 384hr data during a 6hr forecast run, which data will be used? Is the data of the first 24 hour?
When we set interval second in WPS, should I set forecast time step or forecast run?

Thank you very much in advance for your time.

Best regard,
If you intend to run a 24hr simulation, you can choose GFS over the 24hr of your study period. I would suggest you obtain gfs.f000 files, for example, gfs.f000.00UTC, gfs.f000.06UTC, gfs.f000.12UTC, etc. interval second in WPS is the interval of your frocing data. In the example I give above, the interval sedcond should be 21600 (6hr). Forecast time step is totally different to interval second, and it is the integration time step for WRF run.
I would suggest you read the WRF user guide here and gain basic knowledge for numerical modeling study. It is hard to explain such a complicated model system of WRF in the Forum.