ERA5

[seti]

Hi all,

I just have a basic question.
I need to use ERA5 data to run the WRF model with 3 nested domains. Considering the resolution of ERA5 data, which is 0.25 degrees, may I set my domains as 27, 9, and 3 Km?

Please kindly advise me in this regard.
Thanks.

 

[William.Hatheway]

@seti

Since ERA5 data is already 27km you can start at 9km.

 

[William.Hatheway]

@seti

Since ERA5 data is already 27km you can start at 9km.

At least that is what I do and it does well

 

[seti]

At least that is what I do and it does well

@William.Hatheway Thanks, but I think it could be start at higher values like 25 Km!

 

[kwerner]

Thanks, but I think it could be start at higher values like 25 Km!

There is no need to use a domain that coarse when the input data have a grid spacing of 0.25 degrees. The ratio between the input data and your parent domain should be somewhere between 3:1 and 5:1, and starting with 9km falls within that interval. And then it is perfectly fine to use a 3:1 ratio between the parent and child (i.e., 9km, 3km).

 

[seti]

There is no need to use a domain that coarse when the input data have a grid spacing of 0.25 degrees. The ratio between the input data and your parent domain should be somewhere between 3:1 and 5:1, and starting with 9km falls within that interval. And then it is perfectly fine to use a 3:1 ratio between the parent and child (i.e., 9km, 3km).

@kwerner thanks for your reply. I need to use a coarse domain (27 km) because I use chem_opt 202 which runs normally slow.
Therefore if I use a domain with 9 km then I should have a shorter time step, which runs too slowly. That is why I need to consider 27 km for the outer domain. Now, in this case, I wonder if I consider 27 km for the outer domain, using ERA5, would I face any problem? or doing this is technically wrong?

 

[kwerner]

I'm actually not sure what will happen if you use a domain that is more coarse than the input data. This is not something I've ever tested, but you could give it a try and see what happens.

Just so you know, although you set time_step for your outer domain (in this case, 27km), the variable parent_time_step_ratio (which is typically set to the same value as the parent_grid_ratio) determines the time_step for the child domains. Therefore, if your dx/dy = 27, 9, 3, and let's say, for example, that you set time_step = 135, then the model will still run a time step of the following per domain

d01 = 135
d02 = 45 (3 times to reach the d01 time step)
d03 = 15 (9 times to reach the d01 time step)

It would look something like this

d01: 0_________________________________________________________________135
d02: 0_____________________45
d03: 0_____15______30______45
d02: 45_____________________90
d03: 45______60______75_____90
d02: 90__________________135
d03: 90_____105____120__135
d01: 135_______________________________________________________270

...etc.

So including a larger parent domain is simply only adding additional time_steps you don't need, and ultimately will make the simulation longer.

 

[seti]

I'm actually not sure what will happen if you use a domain that is more coarse than the input data. This is not something I've ever tested, but you could give it a try and see what happens.

Just so you know, although you set time_step for your outer domain (in this case, 27km), the variable parent_time_step_ratio (which is typically set to the same value as the parent_grid_ratio) determines the time_step for the child domains. Therefore, if your dx/dy = 27, 9, 3, and let's say, for example, that you set time_step = 135, then the model will still run a time step of the following per domain

d01 = 135
d02 = 45 (3 times to reach the d01 time step)
d03 = 15 (9 times to reach the d01 time step)

It would look something like this

d01: 0_________________________________________________________________135
d02: 0_____________________45
d03: 0_____15______30______45
d02: 45_____________________90
d03: 45______60______75_____90
d02: 90__________________135
d03: 90_____105____120__135
d01: 135_______________________________________________________270

...etc.

So including a larger parent domain is simply only adding additional time_steps you don't need, and ultimately will make the simulation longer.

@kwerner So many thanks for your advice. But in one case I experienced something different. I considered 3 domains with 18, 9, and 3 km. I set thime_step to 72 seconds.
Then set:
chemdt = 0, 0, 0,
bioemdt = 1.2, 0.4, 0.13,
photdt = 1.2, 0.4, 0.13,

and I noticed that the model ran too slowly once using this setup (I had 10 days for simulation, and found out by this slow speed it will take 22 days to simulate my 10 days).

However, I found out that once I used 18, 9, and 3 km the output results were much better. So I need to run the model with this special resolution. But my problem is the speed of the simulation. What should I do to speed up the simulation, please?
Look forward to hearing from you.
Best wishes,

 

[kwerner]

@seti
Since your new question is asking something different than the original question in this thread, will you please create a new thread to ask about simulation speed? We like to keep threads separated by a single topic so they are easier to read in the future. Thanks!