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

Nesting issue

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.


New member

I am using WRF/WPS v4.0 to downscale ERA5 dataset. wrf.exe works fine when I just use two nested domains with spatial resolutions of 30km and 10km. However, when I tried to add a 2km domain, the model crashed because of cfl violation. I tried w_damping option and reducing time step, but it didn't help.

Attachments are my namelists and the wrfinput, wrflowinp and wrfbdy from real.exe. Could you please have a look of it?

Best regards,
Xun Wang


  • namelist.input
    5.3 KB · Views: 62
  • namelist.wps
    935 bytes · Views: 50
    232.7 MB · Views: 48
There are a few issues:
(1) Please change radt = 30, 10, 2, to radt = 30, 30, 30.
(2) Can you increase the vertical levels? For example, set e_vert=41?
(3) Why num_land_cat = 28?
(4) If the domain is over complex terrain area, please set a larger value of epsom (default is 0.1)
(5) To narrow down the issue, probably to turn off set_update.
Thanks so much for your reply and suggestion!
I will try (1) and (2). I have some further questions regarding your reply:
(3) num_land_cat = 28 is because I use USGS land cover data with lakes. But I noticed that USGS is no longer in the Geographical Input Data Mandatory Fields Downloads. Is there any advantages using MODIS land cover instead of USGS? My research region is high and central Asia
(4) What is epsom? I did not find it in the WRF user's guide
(5) Why dose sst_update impact the modle stability?

Best regards,
Modis is kind of 'new' data that is more representative of the present landuse situation. USGS is 'old' data that might be out-of-date.

It should be epssm ( I am sorry for the typo, which is changed automatically by the machine from epssm to epsom)

sst_update shouldn't affect the model stability. I suggest to turn it off only because this will help us to narrow down possible issues in the data and/or model.
Hi Ming,

Sorry for the late update. I changed the following of namelist.input following your suggestion:
(1) radt = 30, 10, 2, to radt = 30, 30, 30
(2) e_vert= 28 to e_vert=41
(3) epssm =0.5
(4) sst_update turned off
But the model still crashed.

CFL violation occurred at the peak of topography in the domain (~6920 m). I change the output interval to 1 min. At the 1st min the values (T2, U, V,W ...) near the peak become very large. At the 2nd min there are NaNs in the output, and the area of NaNs become larger. At the 4th min all the domain is filled with NaNs but the model keeps running.
Your model domain covers large topography gradient area including the Tibetan Plateau. This often causes trouble in the stability of numerical integration, especially when running in high resolution.
I would suggest you further increase epssm to large values like 0.9 or even 1.0. If this still doesn't work, then try to smooth the topography (using coarse resolution input data). If none of the above options works, then I guess you will have to adjust your domains locations.
Thanks for your reply and help. I changed the smooth_option for HGT_M in GEOGRID.TBL to 1-2-1 and it seems to work!