Obvious urban areas and 'speckled' fields

[TyBuck]

Hi all,

I'm currently running a nested (25km > 5km > 1km) configuration centred over Lubbock, Texas, yet I'm finding some intriguing artefacts in my output fields.

The plots attached show a handful of variables of interest where some clear features are present:
1. The town of Lubbock itself (centre of the domain).
2. Odd speckle effects which create an almost static like appearance.

Can anyone explain what's going on here? It seems unrealistic to have a town so aggressively noticeable in all time steps, and the speckled field (particularly the OLR/U10/V10) seems very unrealistic.

Attached is the namelist.input for the described run.

Thanks for the help!
Ty

 

[Ming Chen]

Hi, Ty,
In your namelist.inoyt, you have the option of sf_surface_physics = 1,1,1, which is the thermal diffusion scheme that cannot well describe detailed soil physics process. Can you change to Noah or NoahMP and try again?

 

[TyBuck]

Hi Ming,

Thanks for the advice, I re-ran the simulation using

sf_surface_physics = 2, 2, 2,
sf_urban_physics = 2, 2, 2,

Unfortunately, I'm still seeing an obvious Lubbock and the speckled effects. I also tried a fixed colour scale to see if the effects were actually small and being automatically populated due to the automated colour scaling, but that didn't work either.

Any ideas? Neither effect seems realistic. This is the first high-resolution domain I've ran focussed on an urban area since upgrading to WRF4.4.1 and downloading a fresh high resolution GEOG dataset.

The effect is also replicable on other machines, so it doesn't seem like it's an isolated problem.

Thanks,
Ty

 

[TyBuck]

I think I've found the culprit. Analysing the LU_INDEX in the simulation's geo_em.d03 file returns a familiar looking pattern.

So, for some reason, the simulation is doing something odd with the landuse. Any ideas?

 

[TyBuck]

Digging into this a little more, I looked into near-surface variables outputted in wrfout files compared to interpolated near-surface variables using WRF-Python. Attached is an example, where the interpolated u10 clearly shows the town of Lubbock, but it looks smoothed and more realistic. The u10 directly from the output still has strange speckled effects.

Is there something odd going on with landuse and near-surface variable interactions?...

 

[Ming Chen]

I am thinking that the model behavior is reasonable since the result is consistent with the landuse type. However, I have no way to judge whether lu_index is reasonable for your D03. If not, this is an issue in MODIS static data.