WRF LES - ERROR: Nest start locations do not match: namelist.input vs gridded input file

[kermarrec]

Dear forum member,
I am trying to start a WRF LES case and get the following error:
Input data is acceptable to use: wrfndi_d02
i_parent_start from namelist.input file = 36
i_parent_start from gridded input file = 1
j_parent_start from namelist.input file = 47
j_parent_start from gridded input file = 1
---- ERROR: Nest start locations do not match: namelist.input vs gridded input file
NOTE: 1 namelist vs input data inconsistencies found.

I have no idea how to fix it. I attach the input files for both LES and meso domains. Thanks for any help.
best

 

[kwerner]

Hi,
Can you let me know why you want to use ndown? Typically you would only need to use ndown if your domain sizes (e_we and e_sn) are much different between each domain. Your domains are all very small (actually, domains 2 and 3 are probably too small - they should never be smaller than 100x100 grid spaces), so ndown shouldn't be needed. If you still want to run ndown, you can only run it for 2 domains at a time, so you would need to run for domains 1 and 2, and then again for domains 2 and 3 (see the instructions here).

 

[kermarrec]

Many thanks for your kind answer. Actually, I would have liked to follow the methodology for LES as proposed in Introduction to Running the WRF in LES Mode Jeff Massey April 1, 2014 University of Utah WRF Users Group. - ppt download by Jeff Massey to use a mesoscale forcing. That is the reason why I started from a simulation with 3 nestings to have a wrfouput and a second one with smaller domains. Do I understand from your post that I was wrong in setting the LES simulation? I guess that I should definitively increase the size of the domains from the second simulation, but also the first simulation? How many domains would you suggest me to use and of which size? Is the proposal of https://rmets.onlinelibrary.wiley.com/doi/epdf/10.1002/qj.4263 Table 1 adequate? I was wondering about the need to use such large domains.

 

[kwerner]

It's fine to run an LES simulation, and you don't have to have an extremely large domain to do so. It's just that if you do have very large domains for, say, domain 3 and maybe domain 2, while the size of domain 1 is much smaller, then you would possibly need to run with ndown to ensure you can use the right number of processors for each domain. If your domains do not differ that much in size from one another, it's not necessary to use ndown. You can still run in LES mode without running ndown - those options are not exclusively related. But all of your domains should be at least 100x100 in size, regardless of what other options you choose.

In case you're interested, you can find some additional information on LES case set-ups in the PBL section of the WRF Users' Guide, and in this physics presentation (or see the .pdf version of the presentation).

 

[kermarrec]

I am grateful for your answer as I have nobody who has experience in WRF-LES (we use the PALM software at our university). Thanks to your answers to all my silly questions, I am starting to understand much better the philosophy behind all those variables. I have the feelings that sometime what is found in some articles is incorrect or would not have been mandatory.
What do you mean with "If your domains do not differ that much in size from one another," is 2* times much already? ndown should be run between the rans and the les domain always?

 

[kwerner]

Take a look at the link I sent that discusses choosing the right number of processors. It should explain what a problematic difference would be.