setting issues for several WRF parameters

[sd16121]

Dear Colleagues,

I would like to know the following questions about some namelist options:

1. My dx of three domains are 4500, 1500, 500 m. Can I set the cu_physics as below? What are the differences between the two settings?
cu_physics = 2,2,0,
or
cu_physics = 2,2,2,

2. For surface_input_source, what are the differences between setting it as 1 and 3?

3. For stand_lon, how to set the value for it? If I set it to the same value as ref_lon (see below), does it work all the time?
ref_lat = 54.990311,
ref_lon = -1.629891,
truelat1 = 30.0,
truelat2 = 60.0,
stand_lon = -1.629891,

4. My dx of three domains are 4500, 1500, 500 m, and I set the history_interval = 60, 60, 60,
Can I set the time_step = 3? It should be domain 3's dx(0.5 km) × 6 = 3.
If my parent_grid_ratio = 1, 3, 3, can I set parent_time_step_ratio = 1, 1, 1,?

Many thanks for your time!

Best regards,
Stella

 

[kwerner]

Hi,

1. Yes, you can set it either way. The difference here is that if setting it to equal option 2 (BMJ), you're turning this on for all 3 domains. If you set it equal to 2, 2, 0, you're using the BMJ cumulus parameterization for domains 1 and 2, but turning it off for d03, which is a common practice; however, we recommend turning it off for all domains with grid distance <= 3km, so you'll probably want to turn it off for d02, as well (cu_physics = 2, 0, 0).

2. Setting 'surface_input_source' to either 1 or 3 are technically the same thing - or the same end result. When geogrid runs, it does a computation to determine the dominant landuse categories per grid cell. For historical reasons, real.exe would, by default, also do this exact same calculation again, by setting it to option 1. That tells real to recompute. A few code versions back, there was an improvement to the code that allowed us to skip that unnecessary redundant step of recalculating for the same answer in the real program, but it was necessary to give surface_input_source a new option number (3), which is now the new default. So essentially there is no difference. Leaving it as default (=3) just saves a small (negligible) amount of time.

3. I agree this one can be a little confusing. You only need to specify this when running a lambert, polar, or lat-lon projection. stand_lon is often set to the same value as ref_lon, which means your coarsest domain will be centered.You can read about it a bit more in chapter 3 of the WRF users guide.

4. You CAN set your time_step to be 3, but the rule is that the time_step should be set to no more than 6xDX (where DX is the parent domain). Setting it to 3 would make the model run very slowly. I would recommend setting it to something more like 27, unless it's becoming unstable, and then you would need to lower it. It is okay to set the parent_time_step_ratio to something different than the parent_grid_ratio, but it's not really necessary. Setting it to 1, 1, 1 would then make the time_step too large for domains 2 and 3.