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

Error in running high resolution (10m)

Naser

Naser

Member
Dear all

Hi

I am attempting to run WRF with a 5 km x 5 km domain and 10 m resolution. After producing the first hour of output, my model stopped with the attached error.

I would be grateful for any assistance in resolving this issue.

Yours sincerely,
Naser
 

Attachments

  • namelist.input.txt
    4.3 KB · Views: 13
  • met_em_ncdump.txt
    26.8 KB · Views: 1
  • rsl.error0000.txt
    3.7 KB · Views: 5
  • wrfexout.txt
    5.3 KB · Views: 1
Good morning,

You have a CFL violation. Take a look at this FAQ:
forum.mmm.ucar.edu

Segmentation Faults and CFL Errors

Segmentation faults can be difficult to track down. As there isn't usually a clear error message, it can take some trial and error to figure out the problem. 1) A segmentation fault is often the result of using too many or too few processors, or a bad decomposition. Take a look at this FAQ...
forum.mmm.ucar.edu forum.mmm.ucar.edu
 
William.Hatheway said:
Good morning,

You have a CFL violation. Take a look at this FAQ:
forum.mmm.ucar.edu

Segmentation Faults and CFL Errors

Segmentation faults can be difficult to track down. As there isn't usually a clear error message, it can take some trial and error to figure out the problem. 1) A segmentation fault is often the result of using too many or too few processors, or a bad decomposition. Take a look at this FAQ...
forum.mmm.ucar.edu forum.mmm.ucar.edu
Click to expand...
There is also several issues with your namelist.

see these websites to:


 
Dear William,

Thanks for your response. I have read what you recommended. I still could not find where my error was. Honestly, I want to run the model for a very fine resolution, about 50m or less, and I need one domain. I changed "time_step" several times, and sometime before making 00, it showed an error. Sometimes, after some hours, it shows an error. Sometimes, my job does not kill, but there is no progress to rsl.errors or making output. I am totally confused, and I have to fix this. Can you help me to solve this?

Thank you
Naser
 
For such a high resolution case with dx=10m, many options in namelist.input must be set different to those in typical mesocsale cases (i.e., dx > 3km).

I have a few concerns about your options:
(1) for dx=10m, maximum time_step should be 0.06s. You set it to 30, which is too large for this case.
(2) vertical resolution must increase in accordance with the high horizontal resolution. Note that dz <<dx.
(3) cumulus scheme should be turned off.
(4) without PBL option, you need other schemes to handle mixing process.

In addition, I am not sure what data you can use to force such a high resolution run. Resolutions of most of the analysis products are much coarser than 10m, which makes it hard to create initial and boundary data for this case.
 
Ming Chen said:
For such a high resolution case with dx=10m, many options in namelist.input must be set different to those in typical mesocsale cases (i.e., dx > 3km).

I have a few concerns about your options:
(1) for dx=10m, maximum time_step should be 0.06s. You set it to 30, which is too large for this case.
(2) vertical resolution must increase in accordance with the high horizontal resolution. Note that dz <<dx.
(3) cumulus scheme should be turned off.
(4) without PBL option, you need other schemes to handle mixing process.

In addition, I am not sure what data you can use to force such a high resolution run. Resolutions of most of the analysis products are much coarser than 10m, which makes it hard to create initial and boundary data for this case.
Click to expand...
Dear Chen,
  1. I tried running the simulation with a time step of 0.06 seconds, but after several hours, it stopped and showed a CFL error. I then reduced the time step to 0.04 seconds and later to 0.03 seconds, but the simulations still failed after running for some hours.
  2. As you can see in my namelist, my vertical level is set to 30. I am wondering how I can increase the vertical resolution. Should I increase the number of levels, or do I need to modify the eta_levels? Since my horizontal resolution is 10 meters, how should I set the vertical resolution in terms of meters or height parameters?
Additionally, I understand that for running WRF-LES, the bl_pbl_physics option must be set to "0." Do you think this could be causing issues with the simulation?

Lastly, I am using FNL (Final Analysis) reanalysis data, which has a horizontal resolution of approximately 1.0° × 1.0° (~100 km). Do you think this data is too coarse for such high-resolution simulations? What would you recommend for running the model with high resolutions? Also, what is generally the lowest resolution suitable for running WRF?

I look forward to your advice.

Best regards,
Naser
 
Naser,
There are a few critical issues for your case.
First of all, using quarter degree FNL as forcing data for 10-m resolution WRF is unreasonable. The large discrepancy between resolutions of FNL and WRF could lead to severe problems during WRF integration.
Second, you have to specify vertical eta levels to make sure that dz << dx, which means that the vertical resolution needs to be extremely high. WRF is designed for mesoscale simulation with vertical resolutions within the range of tens of meters to 1km. For your case with dx=10m and dz<<10m, I am suspicious that WRF can work as expected.
Basically I think you need to redesign your experiment. Please read WRF Technote to better understand how it works.
 
Ming Chen said:
Naser,
There are a few critical issues for your case.
First of all, using quarter degree FNL as forcing data for 10-m resolution WRF is unreasonable. The large discrepancy between resolutions of FNL and WRF could lead to severe problems during WRF integration.
Second, you have to specify vertical eta levels to make sure that dz << dx, which means that the vertical resolution needs to be extremely high. WRF is designed for mesoscale simulation with vertical resolutions within the range of tens of meters to 1km. For your case with dx=10m and dz<<10m, I am suspicious that WRF can work as expected.
Basically I think you need to redesign your experiment. Please read WRF Technote to better understand how it works.
Click to expand...
Dear Chen

Thank you for your clear and helpful explanation.

Best,
Naser
 
Dear Yuankw,

I encountered an issue running the model at a 10m resolution due to an error, likely related to the time step. However, I successfully ran the model at a 50m resolution and 5km with a p_top_requested value of 50000 Pa, using 60 vertical levels.

From my experience, running this type of model often requires trial and error. It can be quite tricky, so you may need to experiment with different options. Occasionally, slight adjustments to the domain size can make a difference.

I hope you find success with your setup!
 
You must log in or register to reply here.
Top