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

Low wind speeds

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
Good morning! I usually run WRF for a whole year to use it for an air dispersion model. When I represent the wind rose at a height of 10 m, I see that the speed that comes out is usually lower than the actual measured speed (around 1 m/s difference) and that did not happen to me before. I've upgraded to WRF 4.2.2 (before 3.4) and I've also upgraded to ds083.3 for gridded data (before ds083.2). I am attaching namelist.wps and namelist.inp in case you can help me. Thank you


  • namelist.input
    8 KB · Views: 46
  • namelist.wps
    947 bytes · Views: 37
Since WRVV4.0, we change the vertical coordinate of WRF from terrain following to hybrid, which may to a certain extent affect the results. However, I would expect a better results with this change.
Looking at your enamelist, I found the following settings:
e_we = 20, 36, 56,
e_sn = 20, 36, 56,
e_vert = 21, 21, 21,

For a triply-nested case, the domain size of all the three domains are too small and the vertical resolution is too coarse, which will severely affect the results.

Can you enlarge your model domain, for example, to run over 100 x 100 grids, and set the vertical levels o be at least 40 ?
Thanks for your answer. I am going to make the changes that you indicate and I test it. I'll tell you. Greetings
One question before making the modifications. To run WRF for the previous year (366 days in 2020) and an area of ​​75km * 75km, is it better to use nested domains or could I just use a single domain larger than this size?
It depends on the issues you intend to address. For example, if you focus on synoptic scale analysis, you can run over a single large domain. If you want to study small scale phenomena like convection, you may run nesting cases.
Please take a look at the website:
It provides various papers that describe the physics in WRF.
I ran more tests, with larger domains and higher number of vertical levels, but I still have the same problem. I simulated a week, from January 1st to 7th, 2020, and compared these results with those measured at a station. Attached are the files I used and the wind roses obtained, in case you could tell me what I am doing wrong. As seen in the wind roses, the directions achieved are reasonably good, but the speed of the WRF is much lower.
On the other hand, just as a remark, the global model database I'm using is the following:
NCEP GDAS/FNL 0.25 Degree Global Tropospheric Analyzes and Forecast Grids
I very much appreciate your help in advance. Greetings


  • namelist.input
    8.2 KB · Views: 21
  • namelist.wps
    949 bytes · Views: 21
  • Wind_rose_NOAA_station.pdf
    85.3 KB · Views: 24
  • Wind_rose_WRF.pdf
    85 KB · Views: 21

Your namelist.wps looks fine. For namelist.input, I have a few concerns:
(1) time_step=225 is too large for dx=9km. please reduce it to 45.
(2) eta levels are not set correctly. Below is a sample for your reference:
eta_levels                            = 1.00000 , 0.99307 , 0.98348 , 0.97105 , 0.95551 ,
                                         0.93651 , 0.91363 , 0.88644 , 0.85460 , 0.81855 ,
                                         0.77877 , 0.73579 , 0.69016 , 0.64246 , 0.59329 ,
                                         0.54573 , 0.50104 , 0.45908 , 0.41972 , 0.38281 ,
                                         0.34824 , 0.31589 , 0.28563 , 0.25735 , 0.23096 ,
                                         0.20635 , 0.18343 , 0.16209 , 0.14226 , 0.12384 ,
                                         0.10677 , 0.09095 , 0.07633 , 0.06282 , 0.05036 ,
                                         0.03889 , 0.02835 , 0.01868 , 0.00983 , 0.00000

In fact, it is probably better that you don't specify eta levels. In this case, REAL will produce eta levels for you.
(3) It is better to turn off analysis nudging (i.e., grid_fdda=0, 0, 0). This is because the resolution of your forcing data is too coarse compared to your WRF resolution (quarter degree versus 9-3-1km).
(4) For real-data case, km_opt should be 4.
Hello! It ran WRF again with the namelist.inp I uploaded and it has improved wind speed compared to what it got previously, however it still comes out a bit slower. Is there any other option I could try to improve my result? Thank you very much for your help, I'm making a lot of progress with it


  • namelist.input
    7.8 KB · Views: 32
  • Wind_rose_WRF_2.pdf
    85.1 KB · Views: 25

The only other option that may help is topo_wind. Please set

topo_wind = 1, 1 ,1

Let me know whether it can improve the results.