TR_URB in WRF version 4.5.2 is strange

[N.Suzuki]

We use WRF version 4.5.2 for urban climate studies. Registory.EM_COMMON has been modified to output variables related to urban areas.

Some skin temperature results, such as TR_URB2D (urban roof skin temperature), TB_URB2D (urban wall skin temperature), and TG_URB2D (urban road skin temperature), have initial values based on ground surface temperature. These initial values are set by the module_sf_urban.F.

However, I found that the results are incorrect for TR_URB2D.

Specifically, only TR_URB refers to the skin temperature of the parent domain. This issue was observed during a Bangkok simulation. This symptom is seen only in the innermost domain, and it occurs regardless of whether there are two or three domains.

This issue persists across various simulation settings and regions.

I investigated the following source code modules to find the cause of the problem:

• module_sf_urban.F
• module_sf_noahmpdrv.F
• module_physics_init.F
• start_em

Unfortunately, I could not identify the root cause. This may potentially be a bug.

If you are an expert, please provide your opinion as we would like to hear your insights.

Best Regards,

 

[Ming Chen]

Would you please clarify what you mean by saying "only TS_URB refers to the skin temperature of the parent domain"?

Also, please explain why you think TR_URB2D is not correct. How about the variable TG_URB2D?

 

[N.Suzuki]

Thanks for the response. And sorry for the late reply.

The description of each variable is as follows:
TR_URB(TR_URB2D): Roof skin temperature
TG_URB(TG_URB2D): Ground skin temperature
TB_URB(TB_URB2D): Building skin temperature

These variables are utilized by the Urban Canopy Model (UCM) and are calculated within the urban grid. The initial values of these variables are determined by TSK. While initial values are also provided for non-urban grids, they are not computed by UCM and remain unchanged throughout the simulation.

I will share a new figure. The first panel displays the skin temperature with initial values (t=0):
1st row: Land use category map
2nd row: TR_URB
3rd row: TB_URB
4th row: TG_URB
Left: D01 , Center: D02 , Right: D03

The second panel illustrates the distribution of TR_URB and TG_URB during the simulation:
Left: After 11 hours , Center: After 17 hours , Right: After 23 hours

In addition, the simulation settings are as follows:
Max domain: 3
DX: 18km, 6km, 2km
Time step: 90s (D01)

The first figure indicates that only the distribution of TR_URB resembles the distribution of the urban grid set in D02.
In the second figure, there is no improvement in the distribution as the simulation progresses.

It is understood that TR_URB should ideally resemble TG_URB and TB_URB. This can be confirmed in module_sf_urban.F.
When simulations were conducted using the previous version (4.3.3), it was confirmed that TR_URB has a similar distribution to TB_URB and TG_URB has a similar distribution as well.

Additionally, we observed that TS_URB (Urban skin temperature) and TSK were influenced by the incorrect TR_URB values.

 

[Ming Chen]

Thank you for the detail description of this issue. I will talk to our expert and get back to you.

 

[cenlinhe]

I am still a little confused by your question. If I understand it correctly, the problem you referred to is that for the spatial distribution of TR_URB in D03, it is too similar to the TR_URB pattern in D02 without any high-resolution distribution signals, whereas for TB_URB and TG_URB, they both show high-resolution features in D03 compared to those in D02. You are asking why TR_URB does not show the high-resolution features in D03 compared to D02. Did I interpret your question correctly?
It seems like a downscaling issue from D02 to D03. Did you see an expected high-resolution pattern from D01 to D02 for TR_URB? If it works for TR_URB from D01 to D02, then it should also work for TR_URB from D02 to D03. Which urban physics option are you using? Are you using ndown.exe to run three domains in a one-way coupled mode or are you running it in a two-way coupled mode?