Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Resolution Bounds for First Domain When Downscaling GCM Output

D

dstabile

New member
Hi all,

I have global climate data (100km resolution) that I'd like to dynamically downscale using nested domains. I was told that the first domain does not need to necessarily match the resolution of the global climate data. In order to get sub-kilometer resolution I've been downscaling with the following resolutions for my domains: 75km -> 15 km -> 3 km -> 0.6 km (grid ratio of 5 between each domain).

My question is, what are considered the best-practice bounds for the first domain's resolution? For example, can I stray from the starting 100km resolution by 25% (75km - 125km) as the starting resolution for domain 1? Can I use 125km -> 25km -> 5km -> 1km -> .2km? Or can I go as far down to 62.5km -> 12.5km -> 2.5km -> 0.5km -> 0.1 km?

Thanks for your help.
 
Last edited:
It is recommended that the ratio of resolution of large scale forcing data to the resolution of outermost WRF domain can be around 5-6. In your case, with the 100km resolution forcing data, I would say 15-20km would be fine for WRF outermost domain.

I am concerned of your finest resolution, --- why do you want to go down to resolution of a few hundred meters?
 
Hi @Ming Chen, thanks for your response. I'm glad I asked you - a different WRF expert told me the lowest we could go is 75km resolution.

I'm working on a program where we want to generate very high resolution climate data for the future (we're using CMIP6 data since it goes out to 2100). We're trying to understand how climate change might impact specific sites at the sub-kilometer level. I'd be interested to hear what exactly is your concern - I'm new at this and want to make sure I'm doing everything according to best practice!
 
Last edited:
For long-term climate simulation, personally I think 1-km resolution should be high enough. We have done climate simulations at the resolution of 3-4km, and the results meet our purpose for climate and climate change study. Is there any special reason you want to go to sub-kilometer scale? At sub-kilometer scale, the PBL scheme no longer works but the resolution is still way coarse for LES implementation. Over complex terrain area, numerical stability is also a concern when running WRF at high resolution.
The above is my personal opinions. Please refer to the literature and figure out what is best for your case.
 
Thanks @Ming Chen, your feedback and opinion are very appreciated. We are hoping to collect climate data at very localized sites (essentially specific grid cells in our final domain) which is why we wanted sub-kilometer resolution. I've done some preliminary research into WRF-LES. It seems that we can't use global climate data to drive WRF-LES. Instead, we'd have to downscale our global climate data via a meso-scale model (like vanilla WRF) and then use that output to drive WRF-LES. Is that correct?
 
Last edited:
You can run a real-data case in WRF using an LES grid and LES physics. One aspect to consider is where lateral boundary conditions will come from. Researchers typically have used WRF nesting to increase resolution from dx~several km down to a fine LES mesh of order 10s of meters, and driving the coarse mesh with outside analyses.
 
Thanks again Ming. So it sounds like the following setup could work if our goal was to downscale to 50m resolution:
  1. Interpolate our GCM output (100km) down to 31.25 km to use as the driving data for WRFs first domain
  2. Use the following domain setup:
    • d0 - 31.25 km PBL settings
    • d1 - 6.25 km PBL settings
    • d2 - 1.25 km PBL settings
    • d3 - 250 m LES settings
    • d4 - 50 m LES settings
Please let me know if I misunderstood what you said!

For anyone else who's reading this thread you may also be interested in the following papers:

Simultaneous nested modeling from the synoptic scale to the LES scale for wind energy applications​

Nested Mesoscale Large-Eddy Simulations with WRF: Performance in Real Test Cases​

 
Last edited:
I agree that PBL should be turned on for d0, d1 and d2, and LES can be applied for d4. I am concerned of LES implementation in d3, which has a resolution of 250m that is way too coarse for LES simulation. There is no consensus how to handle PBL process in the grey-zone. I have to say that WRF has limitations in certain applications.
 
Ming Chen said:
I agree that PBL should be turned on for d0, d1 and d2, and LES can be applied for d4. I am concerned of LES implementation in d3, which has a resolution of 250m that is way too coarse for LES simulation. There is no consensus how to handle PBL process in the grey-zone. I have to say that WRF has limitations in certain applications.
Click to expand...
It appears that using different PBL schemes for various domains is not feasible. Could you please confirm if the sample lines below work well for resolutions ranging from 9 km to 333 m?

dx = 9000, 3000, 1000, 333.33
physics_suite = 'CONUS'
bl_pbl_physics = 11, 11, 11, 11,

I believe there is a typo in the wrf_user_guide. In the table for physics_suite=’CONUS’, it is written as follows:
Surface LayerMYJsf_sfclay_physics=2
However, it might actually refer to Eta.
 
Last edited:
You must log in or register to reply here.
Top