How does pertlim "work"?

[cnrwil004@myuct_ac_za]

Hi all

We are wanting to use the pertlim CAM(4) namelist variable in a CESM1.2.0 B compset run, in order to produce an initial condition ensemble. I apologise if this is documented elsewhere, but I have unfortunately not managed to find any clear description of the nature of the temperature perturbation that is applied when the pertlim namelist variable is set to a non-zero value. Could someone please point me to such a description?

Thank you in advance
-Stefaan-

 

[hannay]

pertlim is a perturbation you apply to the initial temperature field.Usually, you add a perturbation of the order of 1e-14. It is add to the initial temperature field using a random_number generator based on global column index                 call random_seed(put=rndm_seed)                do k = 1, km                   call random_number(pertval)                   pertval = 2.0 *pertlim*(0.5 - pertval)                   t3xy(i,j,k) = t3xy(i,j,k)* (1.0 + pertval)                end do 

 

[cnrwil004@myuct_ac_za]

Thank you very much for posting the do loop, that should fully resolves our question. In essence then, if I understand correctly, the temperature field at each grid point across all levels is perturbed by a proportion p of its original value, where the random number p is uniformly distributed in the range (-pertlim, pertlim]?

 

[edyer@atmosp_physics_utoronto_ca]

Hi, So if one wanted to do an initial condition ensemble could you just add the line pertlim=1.e-14 to all user_nl_cam_000x in cesm1.2? My interpretation of the loop is that this will lead to a different perturbation being applied to each ensemble member when using instances. Thanks, Ellen

 

[eaton]

You need to add a different value of pertlim for each ensemble member.  Two runs with the same value of pertlim will be identical. 

 

[hannay]

Also, pertlim only works for initial simulations. It doesn't work for an hybrid run. So, check which type of run you are doing.

Errata on Feb 25: pertlim only works for initial simulations. An hybrid run basically starts as an initial simulation. So pertlim should work for hybrid. 

 

[edyer@atmosp_physics_utoronto_ca]

I am doing a startup run - so hopefully this is the right approach? Thanks for your responses, Ellen

 

[cnrwil004@myuct_ac_za]

We have performed numerous experiments by applying different values of pertlim with hybrid runs and output variables from these simulations appear to diverge as expected for micro-scale IC perturbation. This did not occur when equivalent hybrid restarts were setup with pertlim =0; and neither did such divergence occur when the perturbations were applied for branch restarts. What could be responsible for such diveregence if pertlim is not applied for hybrid cases?ThanksStefaan

 

[Yuan Wang]

Exactly the same experience here. I guess pertlim does work for hybrid runs, since they need to read IC files.

Look forward to the answer from Hannay.

 

[hannay]

You are correct. pertlim will work for an hybrid run. I misinterpreted the documentation.  
pertlim only works for initial simulations. An hybrid run basically starts as an initial simulation. So pertlim should work for hyrbid. 

I am putting an errata on the comment above.