Question about time_step, wrfout_d <domain>_<date> file, etc

[dakyoung]

Hi.

Using WRF, I'm trying to improve the wind speed data from 1.5km resolution to 500m resolution.
I have a few questions, can you give me an answer to this?
I will attach the namelist.wps, namelist.input that I wrote for your reference.


1. When high-resolution using nesting, is it increased in resolution by the relevant physical expression by the &physics setting of namelist.input? Or is it interpolated due to input data?
And for high resolution using WRF, can I use nesting? Please let me know if there is another way to increase the resolution.



2. I want to high-resolution data from 2022-05-05_00:00 at 1.5km to data from 2022-05-05_00:00:00:00 at 500m. (Input data is GFS, FNL etc.)
To be more specific, I would like to get high resolution wind speed data at 00:00:00 on May 05, 2022. At this time, can I put only one input data from 2022-05-05_00:00:00 to extract the output from 2022-05-05_00:00? If you don't recommend this way, can you suggest how many hours of input data to pull out the output as of 2022-05-05_00:00:00?



3. It's a question that follows number 2. The figure below is part of the single-domain WRF calculation process set to time_step=9, and it seems that the calculation is performed every 9 seconds. What calculations are being performed every 9 seconds? I guess the calculation is repeated every 9 seconds, resulting in the accumulated result value in the final wrfout_d01 file. However, as I said in number 2, I want to get the result value at a certain point in time, so isn't it unnecessary to calculate every 9 seconds as below? I'm not sure what exactly the process is for calculations performed every 9 seconds. I think the file size is getting too big as it's calculated every 9 seconds. Is there a way to get rid of this process if it's not necessary?




3. When setting the time_step within the &domains of the namelist_input, I understand that it should be set to less than 6 times the dx value(km). Is that right?
Can I set the time_step large regardless of dx? For example, would putting a value of time_step=100 or 1000 when dx is 1.5km cause problems with the calculation process?



4. When I ran a single domain WRF with the attached namelist.wps, namelist.input, the two files were printed out as below.
What does each of these two output files mean? I wonder if it represents the average value for an hour or the value of 05-05_00:00, 05-05_01:00:00, that is, the pinpoint value.




I was curious about a lot of things, so I asked this and that.
I think it would be nice if you could answer only what you know.

Thank you in advance!

 

[dakyoung]

I still haven't found a solution.
Is there anyone who can help me?

 

[kwerner]

Hi,
Apologies for the delay in response. I notice you've had several other posts, questioning some applications for running WRF. Have you already reviewed the WRF Users' Guide and other helpful resources? If not, I would recommend starting with those, as many of your questions can be answered by them.

• We have all of our tutorial presentation recordings available on YouTube. And the .pdf versions of those are available here.
• WRF Online Tutorial
• WRF Users’ Guide
• Compiling Tutorial
• WRF Technical Note
• Best Practice for Namelists (provides recommended settings for creating your domain and other namelist options):
  • WPS: WRF namelist.wps Best Prac
  • WRF: WRF Namelist.input Best Prac
• For other information on downloading, etc. see the WRF Users’ Home Page.

I'll try to answer your current questions the best I can, and then please read some of the above guidance to get a better idea of how the model works.

1. When high-resolution using nesting, is it increased in resolution by the relevant physical expression by the &physics setting of namelist.input? Or is it interpolated due to input data?
And for high resolution using WRF, can I use nesting? Please let me know if there is another way to increase the resolution.

I'm not sure I fully understand the question, but nesting is performed based on the settings given in the WPS namelist, which sets up the size and resolution of the domains, and then the input meteorological data is interpolated to that domain, based on the requested resolution. Yes, you can use nesting and it's probably necessary for a high-resolution nesting simulation. If you are interested in nesting down to 500 m, you will likely need to start with a much coarser resolution nest and then nest down to 500m in increments. The namelist setting 'parent_grid_ratio' determines the ratio of resolution between each domain. We recommend odd ratios (either 3:1 or 5:1). This is also true with the resolution of the meteorological input data you're using, and the first (parent) domain in your simulation. The difference in resolution between those shouldn't be more than about a 5:1 ratio. This is all explained in the tutorial presentation on Nesting.

2. I want to high-resolution data from 2022-05-05_00:00 at 1.5km to data from 2022-05-05_00:00:00:00 at 500m. (Input data is GFS, FNL etc.)
To be more specific, I would like to get high resolution wind speed data at 00:00:00 on May 05, 2022. At this time, can I put only one input data from 2022-05-05_00:00:00 to extract the output from 2022-05-05_00:00? If you don't recommend this way, can you suggest how many hours of input data to pull out the output as of 2022-05-05_00:00:00?

If you're interested in a specific time, it's best to have a bit of spin-up to make sure the model has time to stabilize - about 12 hours in advance is best.

3. It's a question that follows number 2. The figure below is part of the single-domain WRF calculation process set to time_step=9, and it seems that the calculation is performed every 9 seconds. What calculations are being performed every 9 seconds? I guess the calculation is repeated every 9 seconds, resulting in the accumulated result value in the final wrfout_d01 file. However, as I said in number 2, I want to get the result value at a certain point in time, so isn't it unnecessary to calculate every 9 seconds as below? I'm not sure what exactly the process is for calculations performed every 9 seconds. I think the file size is getting too big as it's calculated every 9 seconds. Is there a way to get rid of this process if it's not necessary?

The time_step is how often the model is integrated forward. This is how the model is run and works, integrating through all calculations each time to update the model state, and therefore is a mandatory component. time_step should be no more than 6xDX (for e.g., if dx=30000, time_step should = 180, or less). The file size should not be getting larger based on the time_step. The file size is based on the size of your domain, how high the resolution is, the physics components used, how often you are asking the model to write-out history (history_interval), and how many history_intervals you are asking for in each history file (frames_per_outfile), in addition to some other components.

3. When setting the time_step within the &domains of the namelist_input, I understand that it should be set to less than 6 times the dx value(km). Is that right?
Can I set the time_step large regardless of dx? For example, would putting a value of time_step=100 or 1000 when dx is 1.5km cause problems with the calculation process?

Yes, this would absolutely cause issues! It would cause the model to become unstable and crash.

4. When I ran a single domain WRF with the attached namelist.wps, namelist.input, the two files were printed out as below.
What does each of these two output files mean? I wonder if it represents the average value for an hour or the value of 05-05_00:00, 05-05_01:00:00, that is, the pinpoint value.

These are the output files for each of those times. The value of each variable is specific to that exact time. It is not an average.