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

Stripes in U and V fields

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.

chunchih

Member
Hi WRF support,

Recently, I noted there are some unphysical stripes (where the velocity of a particular component is some magnitude at one point while it is the magnitude but the opposite direction at the neighboring grid point) in my 3D U and V fields when viewing a wrfout file using NCVIEW (see below. The domain has a 3-km resolution in x and y. The plotted velocity range is from -10 to 10 m/s for both components). At first, I thought these are related to the horizontal staggering of the grids since the clearest stripes in U are mostly north-south orientated and the stripes in V are mostly west-east orientated. These stripy features will average out nicely when I de-stagger the U and V fields along their respective i and j directions.

I did a few tests and found out that the clearest stripes usually occur at where the component speed is small (< 1 m/s). They seem to appear after midday and wane during the evening hours and disappear at higher altitudes. I thought these sort of make sense since the opposite velocities of the same magnitude will average out to be a small number, and convection resulting from daytime heating usually adds frictional drag to boundary-layer winds and wind speed usually increases with height.

I was just wondering if the strange-looking stripes in the staggered U and V fields play a role in the calculation of other fields or it is the de-staggered components that are used in the calculations? I did consider these might be numerical noise and could be removed using diff_6th_opt=2, but I wasn't sure if it is a good idea. I am running a real-case simulation using WRF 4.2.1. The namelist.input I used is attached below.

Any clarification would be greatly appreciated.

Thanks,
David
 

Attachments

  • namelist.input
    6.5 KB · Views: 47
  • namelist.wps
    1.9 KB · Views: 46
  • U_staggered_NCVIEW.png
    U_staggered_NCVIEW.png
    365 KB · Views: 652
  • V_staggered_NCVIEW.png
    V_staggered_NCVIEW.png
    356.4 KB · Views: 652
Hi David,
Yes, if you haven't already tried, can you set diff_6th_opt =2 to see if that removes the noise? It should be okay to use that option.
 
Hi Dr. Werner,

Yes, turning on diff_6th_opt=2 did remove the stripes in both staggered U and V. However, I was wondering if the stripes are a normal model behavior and I didn't do anything wrong in my simulation setup. In other words, the model wasn't producing wrong results per se if I don't use the optional diffusion filter. The stripes disappeared when I de-stagger the U and V fields, that's why I didn't think it was a major issue. But I wasn't too sure about this.

Thanks,
David
 
David,
If the stripes are resolved by the end of the simulation and you believe the results are reasonable, then it's probably okay to have them there. I would just advise to run a few small tests to see which way provides the better results.
 
Hi Dr. Werner,

Sorry, I just want to make sure that I understood your response correctly. When you say the stripes are 'resolved' by the end of my simulation do you mean that they disappear? Most of the stripes are gone (or fizzle out) about 3-4 hours before the end of my simulation.

In terms of test simulations, I have already conducted a few with diff_6th_opt=2 and different diff_6th_opt_factors (ranging from 0.02 to 0.12). Thus far, it seems that the simulation without the explicit diffusion produced the 'best' results. However, that is with respect to the phenomenon of interest in my project, not about the stripes in the wind fields. So I am torn on whether to use this option. The stripes do disappear when diff_6th_opt is turned on.

David
 
Hi David,
I apologize for the delay in response. I've been out of the office much over the last couple of weeks. I ran your inquiry by our physics expert, who noticed you use MYJ PBL, which can try to resolve PBL eddies at grid scale, and may be what you're seeing. YSU may have less of that, so that's another option. You could also try Shin-Hong, which is intermediate.

These stripes are probably due to surface heating and PBL rolls parallel to the shear, but the scale may be limited by the resolution, so they could be larger than real ones. It's not really wrong - just that it's poorly resolved. The mean behavior may be OK.

If you'd still like to look into this more, please let me know more specifically how the results worsen when using or not using diff_6th_opt.
 
Hi Dr. Werner,

That's OK.

We ran the same simulation with different diff_6th_opt_factors ranging from 0.005 to 0.12. Using diff_6th_opt = 2 sufficiently reduces the stripes with any use of diff_6th_opt_factor above 0.005, but for reasons that we still do not understand very well, it also affects the initiation/maintenance of the moist convection that we want to simulate, which is primarily driven by small-scale processes. Since the moist convection of interest is sufficiently far away from the stripes, we have decided not to use this explicit diffusion option after all. The stripes are not that apparent when I de-stagger the U and V fields for analysis anyway.

Thank you for your explanation and help!

David
 
Top