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

Stochastic perturbation for initial potential temperature

M

MSdueon12

Member
Hello,
I will have an ensemble simulation with the initial temperature randomly perturbed. I followed the User guide to set the stochastic perturbation, please look at the attached namelist I used.

I am not sure if these settings are correct. Also, I checked the wrfint_03 where I thought the initial potential temperature should be varied to the result without the perturbation, however, I am not able to find any difference between the two simulations. May I ask if any setting was wrong or the wrong variable I chose to compare?

I appreciate it if you would like to instruct me on how to get the perturbation done.

Thanks
 

Attachments

  • &time_control.txt
    5.1 KB · Views: 8
Did you see any difference in your wrtfout_d03? I won't expect changes in wrfinput because the perturbation will be imposed to variables read into WRF from the wrfinput file, while the wrfinput itself won't change.
For the options in your &stoch section, I am not 100% sure whether they are appropriate. I will talk to our expert and get back to you if any changes are necessary.
 
Ming Chen said:
Did you see any difference in your wrtfout_d03? I won't expect changes in wrfinput because the perturbation will be imposed to variables read into WRF from the wrfinput file, while the wrfinput itself won't change.
For the options in your &stoch section, I am not 100% sure whether they are appropriate. I will talk to our expert and get back to you if any changes are necessary.
Click to expand...
I checked the wrfout_d03 for the first output. and I set the history interval to 1 min, expecting some changes, but I didn't get variation. I would appreciate it if any further updates here.
 
The following are stochastic options in WRF, and you should choose one from them:
rand_perturb=1
sppt=1
skebs=1
spp=1
It is not recommended to turn on multiple options simultaneously.
 
Ming Chen said:
The following are stochastic options in WRF, and you should choose one from them:
rand_perturb=1
sppt=1
skebs=1
spp=1
It is not recommended to turn on multiple options simultaneously.
Click to expand...
Thank you so much for your update.
I have tested only turn on rand_perturb = 0,0,1 but I can't find any difference between on and off. Should I check the result of T in wrfoutd03 of the first output?
 
Also, starting from a single domain run to check the impact of rand_perturb could be more efficient.
 
Ming Chen said:
Also, starting from a single domain run to check the impact of rand_perturb could be more efficient.
Click to expand...
Hello,
yes, I set rand_perturb = 1,1,1 in my namelist, but I can't find any perturbation of T in wrfout files on the first output time. So I wonder if the variable I checked was incorrect or the setting should be adjusted. I appreciate it if you would like to help to figure this issue out.

Another question is what is the T_INIT in wrfinout file? From the definition it is the initial potential temperature, but from the value, it is not equal to T+300, thank you very much if you would like throw the answer to me as well.
 
wrfout at the initial time saves information from wrfinut. if you compare wrfout files with and without perturbation at later times of integration, I suppose you should find differences.

T_INIT is initial potential temperature, unit is "K".
 
Ming Chen said:
wrfout at the initial time saves information from wrfinut. if you compare wrfout files with and without perturbation at later times of integration, I suppose you should find differences.

T_INIT is initial potential temperature, unit is "K".
Click to expand...
Thank you for your update, following your information, the variable T in the later times seems unperturbed when comparing with and without the &stoch.
rand_perturb = 1,1,1
 
Please set the following options in &stoch

rand_perturb (max_dom) = 1, 1, 1,
gridpt_stddev_rand_pert (max_dom) = 0.03, 0.03, 0.03
lengthscale_rand_pert (max_dom) = 500000.0, 500000.0, 500000.0
timescale_rand_pert (max_dom) = 21600.0, 21600.0, 21600.0
stddev_cutoff_rand_pert (max_dom) = 3.0, 3.0, 3.0
rand_pert_vertstruc = 0

Then try again. Note that I specify default values to the above options, but these values should be adjusted based on your case. Please refer to the website here for more detailed information about the stochastic approach.
 
hi, i have a question. if i want to increase the number of ensemble members to 100 or even larger, I must change the "nens" in namelist.input to 2,3,4,...,100 each time? or could some automatic methods be use to realise 100 ensembles to simulate just at one time?
 
Please see my answer to this question in another form, --- we would appreciate that you don't post the same question in different forums.
 
You must log in or register to reply here.
Top