Advice on WRF setup for valley cold-air pools and inversions

[seti]

Dear WRF Support Team,


I am using WRF version 4.6.0 for wintertime simulations over complex mountainous terrain. One of the main problems is that the model does not reproduce nocturnal cold-air pools and valley inversions strongly enough. At several valley-floor stations, observed nighttime temperatures are several degrees colder than the simulated T2, while the model remains too warm during the night.

In my current setup, I use bl_pbl_physics = 2 (MYJ), sf_sfclay_physics = 2 (Eta Similarity), and sf_surface_physics = 4 (Noah-MP).

I would appreciate your advice on the following:

• Horizontal resolution: My current innermost domain is about 1 km. I am considering reducing it to around 300 m and using a 30-m DEM. Is ~300 m sufficient for resolving cold-air drainage and valley cold pools, or would finer resolution be preferable?
• Vertical resolution: What height would you recommend for the first mass level above ground? Would around 10–15 m AGL be appropriate? Also, approximately how many vertical levels should be placed below 500 m and below 1 km? Is about 60–70 total eta levels reasonable?
• PBL scheme: Which PBL and surface-layer schemes are most suitable for stable nocturnal boundary layers, strong inversions, and cold-air pooling? I am currently testing MYJ + Eta similarity, but would MYNN or another option be more appropriate?
• 300-m gray zone: At approximately 300 m grid spacing, should a conventional PBL scheme still be used, or should a different turbulence treatment be considered?
• Other settings: Are there important settings related to horizontal diffusion, terrain smoothing, slope radiation, topographic shading, land-surface physics, or snow that are particularly important for valley cold-air pools?
• Nesting ratio: Is it acceptable to use parent-grid ratios of 3, 3, 5, or is 3, 3, 3 generally preferable? Could a ratio of 5 for the innermost domain cause problems with interpolation, numerical stability, two-way nesting, or terrain-induced circulations?

My main goal is to improve nighttime cooling, inversion strength, and cold-air accumulation at valley-floor locations.


Thank you very much for your advice.

 

[seti]

@Ming Chen

 

[Ming Chen]

Hi Seti,
Please see my answers below:

Dear WRF Support Team,


I am using WRF version 4.6.0 for wintertime simulations over complex mountainous terrain. One of the main problems is that the model does not reproduce nocturnal cold-air pools and valley inversions strongly enough. At several valley-floor stations, observed nighttime temperatures are several degrees colder than the simulated T2, while the model remains too warm during the night.

In my current setup, I use bl_pbl_physics = 2 (MYJ), sf_sfclay_physics = 2 (Eta Similarity), and sf_surface_physics = 4 (Noah-MP).

I would appreciate your advice on the following:

• Horizontal resolution: My current innermost domain is about 1 km. I am considering reducing it to around 300 m and using a 30-m DEM. Is ~300 m sufficient for resolving cold-air drainage and valley cold pools, or would finer resolution be preferable?

It depends on the scale of the valley. How wide is it? WRF must have sufficient grids to resolve the feature of the valley.

• Vertical resolution: What height would you recommend for the first mass level above ground? Would around 10–15 m AGL be appropriate? Also, approximately how many vertical levels should be placed below 500 m and below 1 km? Is about 60–70 total eta levels reasonable?

Generally, higher vertical resolution is required for finer horizontal resolution. The 1st level above ground and number of levels below 500m should be adjusted based on the depth of the cold pool. Note that an extremely low AGL of 1st level may lead to numerical instability. I am not sure how many vertical levels are appropriate for the specific case of cold pool simulation, Please refer to the literature and hopefully you can find helpful information.

• PBL scheme: Which PBL and surface-layer schemes are most suitable for stable nocturnal boundary layers, strong inversions, and cold-air pooling? I am currently testing MYJ + Eta similarity, but would MYNN or another option be more appropriate?

I would recommend MYNN because it is well developed and widely used in NOAA real-time simulation.

• 300-m gray zone: At approximately 300 m grid spacing, should a conventional PBL scheme still be used, or should a different turbulence treatment be considered?

This is a zone in which assumptions used in PBL schemes are invalid. It is also too coarse for LES adoption. Please try the
SMS-3DTKE scale-adaptive LES/PBL scheme. Note that this scheme must be used with diff_opt = 2. PBL schemes must be turned off. And it works with

sf_sfclay_physics = 1, 5, 91

Other settings: Are there important settings related to horizontal diffusion, terrain smoothing, slope radiation, topographic shading, land-surface physics, or snow that are particularly important for valley cold-air pools?

• Nesting ratio: Is it acceptable to use parent-grid ratios of 3, 3, 5, or is 3, 3, 3 generally preferable? Could a ratio of 5 for the innermost domain cause problems with interpolation, numerical stability, two-way nesting, or terrain-induced circulations?

My main goal is to improve nighttime cooling, inversion strength, and cold-air accumulation at valley-floor locations.


Thank you very much for your advice.

nesting ratio 3 is commonly used, but the ratio of 5 also works. The boundary of the nested domain should avoid large topography area with steep slope.