Should using an Urban Scheme create differences outside urban areas?

[struggling_phd]

Dear WRF community,

I am trying to do some research regarding urban scheme parameterizations within WRF. However, I am a little surprised with simulation outcomes. Here is my experiment:

1. Create a domain, boundary and initial conditions (wrfbdy , wrfinput) for some period of time
2. Make two copies of wrfbdy, wrfinput and namelist.input
3. Conduct two different simulations, one with urban scheme off (sf_urban_physics = 0) and one with BEM urban scheme turned on (sf_urban_physics = 3)
4. Observe contour differences at surface level for various meterological variables (T2, WSPD, RH, HFX, LH ...)
5. get worried why there are significant differences outside urban areas??

I attach a snapshot of winds and T2 after 43h of simulation. The middle panel is with urban physics turned on (sf_urban_physics = 3), the right panel is with urban physics turned off (sf_urban_physics = 0), and the leftmost panel is the difference between the middle and the rightmost. As can be seen, there are differences even over ocean.. Can someone please tell me if I am doing something wrong, or is this a regular behaviour?

I inspected the boundary file thoroughly, I wrote a simple code which went through all the 4D and 3D variables to make sure there were no difference, and indeed there are were differences. I double checked the namelists, and the only difference it he urban scheme activation...

Thanks!
Leo

 

[Ming Chen]

Leo,

Would you please send me your namelist.input and namelist.wps to take a look? Are there many urban grids in your model domain?

 

[struggling_phd]

Dear @Ming Chen ,

I am attaching my NoUrb , and Urb namelists, along with the wps namelist. Also I show you a plot of the land_use variable where I have isolated the urban grids (LU_INDEX = 13). In the left subplot the white pixels are urban, and on the right, the only thing that is not white are the urban grids. The resolution is 30km x 30km for these runs.

 

[Ming Chen]

Your namelist.input looks fine. Also there are not a large number of urban grids in your domain. I agree that the impact of urban scheme seems to be overestimated.

Did you run the two cases with the same number of processors?

If the number of processors and all other options except the urban scheme are the same, then I suppose the difference is the model response to urban effects. Sorry that due to limited human power and tight schedule, I don't have more options to further dig into this issue and figure out exactly what is going on inside the model.