Is this model configuration likely to produce reliable results?

[thiagods]

Dear all,

I am a pretty seasoned land modeller but this is my first time handling an atmospheric model. I am preparing a set of WRF v4.1.3 simulations that aim to investigate land/atmosphere interactions by altering the default WRF land use map. I plan to publish the results in a research paper.

My study region is central/northeastern Brazil, with 27/9 Km nested domains and two one-year simulations (one control with default geographic data and one experiment with altered LU map). I will be using CFSv2 for the year of 2013 to drive the WRF runs.

Since I currently have no affiliation with a large institution that provides access to a HPC, I will be using some remaining student credits on AWS to set up a cluster (3 nodes with 16 real cores each) that will run WRF. I can afford to run this cluster 24/7 for around 20 days. With such tight computing resources to run these simulations, I have little margin to make mistakes.

Therefore, I was wondering if more experienced WRF users, perhaps even some staff members, would be willing to take a look at my WPS/WRF namelists to see what potentially could go wrong with my runs and where there is room for improvements. I am particularly unsure as to the which physics parameterizations to use. It doesn't seem to exist any consensus on that. Ideally, I would perform sensitivity tests to determine the best physics combinations for my specific study domain. However, spending computing resources on this is currently out of question as I have to focus on the main runs.

Hopefully there is some "safe" set of physics that produces realistic output for the domain configuration I am using (or at least that will be accepted by the usually finicky journal reviewers). So far, I have successfully used the WPS tools to prepare the input data (file "thiago_met_em_files.tar" placed in Nextcloud) and I am ready to trigger the run when I am confident to.

Hence, any comments or insights would much appreciated!

Cheers,
Thiago.

 

[Ming Chen]

Hi,
I looked at your names files. I have several suggestions:
(1) Your model domain (based on the namelist.wps) is located from around the equator to about 17 degrees south (please see the attached png file). I would suggest that you change the map projection from lambert to mercator. This is because with the mercator projection, the areas and shapes are well preserved between about 15° north and south. However, if for some reasons you want to stay with lambert, that is fine. But you need to pay attention to map factor, which should be around 1. if it is too large, you need to reduce time step accordingly.

(2) In your namelist.input, it seemed that ou try to use the 'tropical' suite. However, you later reset all the physics scheme. Note that and 'TROPICAL' is equivalent to
mp_physics = 6,
cu_physics = 16,
ra_lw_physics = 4,
ra_sw_physics = 4,
bl_pbl_physics = 1,
sf_sfclay_physics = 91,
sf_surface_physics = 2,

If you want to use this suite, then you don't need to reset the above physics options in your namelist.

All others look fine.

 

[thiagods]

Hi,
I looked at your names files. I have several suggestions:
(1) Your model domain (based on the namelist.wps) is located from around the equator to the about 17 degrees south (please see the attached png file). I would suggest that you change the map projection from lambert to mercator. This is because with the mercator projection, the areas and shapes are well preserved between about 15° north and south. However, if for some reasons you was to stay with lambert, that is fine. But you need to pay attention to map factor, which should be around 1. if it is too large, you need to reduce time step accordingly.

(2) In your namelist.input, it seemed that ou try to use the 'tropical' suite. However, you later reset all the physics scheme. Note that and 'TROPICAL' is equivalent to
mp_physics = 6,
cu_physics = 16,
ra_lw_physics = 4,
ra_sw_physics = 4,
bl_pbl_physics = 1,
sf_sfclay_physics = 91,
sf_surface_physics = 2,

If you want to use this suite, then you don't need to reset the above physics options in your namelist.

All others look fine.

Thanks for your feedback Ming. After changing the projection and starting the run, I was getting the infamous CFL errors. I then reduced the time step and the simulation seems to be stable now.

Thanks again for your input.