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

cfl error and unusual streaks in output wind speed (RESOLVED)

This post was from a previous version of the WRF&MPAS-A Support Forum. New replies have been disabled and if you have follow up questions related to this post, then please start a new thread from the forum home page.

mgolbazi

New member
View attachment namelist.inputstreaks_June1st_2018.PNGHi everyone!

I am trying to run WRF4.2 for 3 months of summer in 2018. My domain is 400*400 with a 4km resolution. Overall I have defined 56 vertical levels and I need 4 levels below 30 meters and one level at around 120 meters. I have used "dzstretch" options introduced in: https://forum.mmm.ucar.edu/phpBB3/viewtopic.php?f=40&t=10233

For input data, I am using the 12km NAM data from: https://www.ncdc.noaa.gov/data-access/model-data/model-datasets/north-american-mesoscale-forecast-system-nam

Problem: My runs keep crashing due to CFL error at random times and random gridcells. I understand that the CFL errors happen because the model becomes numerically unstable and a few solutions have been given in the forum like:

1. reducing the time step.
2. using the adaptive time step.
3. using "epssm" in the namelist.
4. using "smooth_cg_topo = .true." in the namelist.

Considering all this, I still get CFL errors randomly!

Another problem is, even though sometimes the runs do not crash for some simulation periods by using some of the solutions above, there are still streaks in my wind speed plots from the outputs. The streaks are so persistent. I have tried so many things and 3 kinds of input data (fnl, gfs, NAM) but they are still there. The streaks normally show up around times (14:00-23:00) in the first day of simulation and then (18:00 to 23:00) for the rest of the runs. I attached a plot to this post for your reference.

I doubted that the streaks are because of thin vertical layers close to the surface, so I ran a default simulation with the same number of vertical layers without thin layers close to the surface and streaks are still showing up in the output. Now, I am doubting that there is something wrong in my namelist that I cannot notice.

I also attached my namelist.input file for a successful run (no cfl errors) WITH streaks in the output.

I am stuck here and any suggestion would be very helpful!

Thank you!
 
Hi,
I apologize for the delay. It can sometimes take us a few business days to get to every post, as we receive numerous and sometimes an individual problem can take more than a day to resolve. We also do not work nights/weekends. Thank you for your patience.

I've been trying to recreate your issue but haven't been able to thus far. For the namelist that you sent,

1) will you send the files I would need to run wrf (i.e., wrfbdy, wrfinput, wrflowinp). I'd like to try to use your exact input.. If the files are too large to attach, please see the forum home page for information on sending large files.
2) Please let me know the time you start seeing the problem for this particular case.
3) I know your plot shows 'wind.' Is it the U field? V field? Please let me know exactly what you're plotting. If you're using an external post-processing tool, let me know if you see the problem in the raw output data - like in the U field on the wrfout* files. You can use ncview to check this.

Thanks
 
Sorry if I got impatient and for my own delay! It took some time to upload the files.

kwerner said:
will you send the files I would need to run wrf (i.e., wrfbdy, wrfinput, wrflowinp). I'd like to try to use your exact input.. If the files are too large to attach, please see the forum home page for information on sending large files.

I uploaded my input files to Nextcloud. Please find them here: https://nextcloud.mmm.ucar.edu/nextcloud/index.php/s/o8nYwN3nHMQr1Ov
The file name is : mgolbazi_4km_streaks.tar.gz

kwerner said:
2) Please let me know the time you start seeing the problem for this particular case.

In a 2 day run (June 1st to June 3rd), the streaks happen around times ~17:00-24:00, and ~44:00-48:00 (or ~18:00-23:00 the second day). So, I would say the last 6-7 hours of every simulation day.

kwerner said:
3) I know your plot shows 'wind.' Is it the U field? V field? Please let me know exactly what you're plotting. If you're using an external post-processing tool, let me know if you see the problem in the raw output data - like in the U field on the wrfout* files. You can use ncview to check this.

I am plotting the total wind speed (sqrt(u^2+v^2)) at the 9th model level (~121m). I use NCL to do so. I also included my NCL code for plotting the wind speed in the attached .tar folder. In the "U" field, there are also similar streaks in the bottom left side of the plot and there are white areas (I assume they are NANs) on the top left where streaks happen.

I also would like to share some things that I have found since my 1st post here in case if it helps at all:

1. I switched the "moist_adv_opt" from option 1 (positive-definite) to option 0 (simple). I learned from chapter 5 user's guide that "model filter "diff_opt = 1 and km_opt = 4 (a commonly-used real-data case mixing option) is not guaranteed to be positive-definite nor monotonic due to the variable eddy diffusivity, K.". So I think a positive definite moisture advection option was part of what was causing the CFL errors.

2. when I increase the thickness of the 1st model level from 5m to 10m the streaks are SLIGHTLY better but still there.

3. I ran 2 separate runs with a simple namelist for two different horizontal resolutions: 12km and 4km for the same area. These two runs used the same namelist, except for variables that are resolution-dependent e.g. the time step. The result was, there were no streaks for the 12 km run while the 4km run was all streaky! So I think it is a resolution-dependent problem.

Please let me know if I can help with anything.

Thank you!
 
Thank you for sending everything! I'll work on trying to recreate this. In the meantime, however, based on your test #3, can you try to add an outer more-coarse resolution around the 4km domain to see if that makes any difference? With 12km NAM input, I wouldn't think this would be necessary, but since your coarse simulation was streak-free, it seems like it could potentially help. Thanks!
 
Hi,
I ran your case and I am able to see what you are. Can you try to set diff_6th_opt = 2 in your namelist and see if that makes any difference?
 
Thank you!! After a month of trying and failing, I finally do not see the streaks anymore! This is great news for my PhD dissertation! : )

The plot below is how it looks now!

I also intend to try this with the recently released WRF-4.3 with the newly added Implicit-Explicit Vertical Advection (IEVA) scheme. I can update the post when I have the results.

Thanks again for the support,
Maryam
 

Attachments

  • plotwind_000025.png
    plotwind_000025.png
    158 KB · Views: 1,306
Maryam,
That is great news! I'm so happy to hear that. Was it the "diff_6th_opt = 2" change that fixed the problem, or something else?
 
Thank you!

Yes, I changed the "diff_6th_opt = 2" and I used it with "diff_6th_factor = 0.12" and it worked!
 
Top