Domain sizes and more fine-tuning for .25 GFS

All issues/questions about EMS v3.4 package, please ask here.
theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 6:35 am

Hi All,

So I am running an operational nmm version for the office, it is currently still on GFS .5 degree initialization and ptile inputs. My project for this week is to start using the .25 degree GFS data for both startup and boundary files.

First question is with GFS being .25 it equates roughly to a 25km grid here at my location. NMM steps down by one thirds. Do I have to precisely make my MOAD 8.3333 grid, with the nest then at 2.7777? What is best practice?

Best regards,
Theo

meteoadriatic
Posts: 1587
Joined: Wed Aug 19, 2009 10:05 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by meteoadriatic » Mon May 25, 2015 7:03 am

Resolution ratio is only mandatory within synchronous nesting and in NMM has to equal 3. However in asynchronous nesting can be anything. The same is true for your ratio between mother domain and initial/boundary conditions. So basically your mother domain can have any resolution you want, but your nest, if you use synchronous nesting must have 3x smaller resolution.

Now this is from will work / will not work point of view. From quality point of view, you should not exceed ~5 times scaling down at mother domain from your boundary conditions resolution. This means, for 0.25° GFS, it is not advisable to run mother domain with less than 0.05° resolution, or wierd things can start to happen near domain borders, and propagate inside domain with time.

Hope this helps,
Ivan

theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 9:00 am

Hi Ivan,
Thanks for the reply! :-) That makes great sense, I wanted to make sure I was on the right track. I have created an nmm MOAD of 9km and that is then 1/3rd nested down to 3km.

Do forgive me, I will have a handful of questions as I go along. For interest's sake, I have followed an idea you had brought up in a post somewhere this morning - which is to start my mesoscale run off the T03 GFS instead of the T00 one, that makes a lot of sense.

My next steps will be to follow your instructions in the "Using GFS 0.25 deg grid dataset" thread - are those instructions for using GFSPARA still the right way to go? (gfs-para_gribinfo.conf and Vtable.GFSPARA and then "gfs-para for DSETS)

I will be using the 230 MB .25deg gfs file for initialization, and then gfsptiles for boundaries. From what I can see, still only the 0.5 ptiles are available, no .25 yet?

Best regards,
Theo

meteoadriatic
Posts: 1587
Joined: Wed Aug 19, 2009 10:05 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by meteoadriatic » Mon May 25, 2015 9:08 am

No, parallel GFS run is not available anymore, so those instructions are obsolete and will not work.

I recommend that you upgrade your EMS installation to the latest, and then use gfsp25ptiles for your 0.25° input.

theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 9:42 am

Hee hee... good thing I asked! I was worried about updating as there is a GrADS update that may kill some of my backup graphical output, but will have to bite that bullet as I go along, the .25 gfs is just too important now.

Then another question about DSET.. I currently have DSETS = gfs%gfsptile,gfs which I took on advice on this forum, but cannot remember why the '%' is there?

What would my DSETS argument be to preferably initialize with .25 gfs, if that is not available then .5 (or even full gfs) for initialization, and then boundary files of .25 ptiles, with a backup of full gfs if the ptiles are not available. I make an assumption here that if the .25 ptiles are not available, then it is likely that the .5 ptiles will also not be - so I have to use full gfs for boundary.

Best regards,
Theo

theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 11:21 am

In order to not appear lazy, is it true to say that DSETS = gfs%gfsptile,gfs translates to use full gfs as initialization with gfsptile as boundary files, or if not available use full gfs for everything?

In which case my new DSETS = (name_for_gfs25)%gfsp25ptiles,(name_ for_gfs.5) would be almost right, except how to add in the full gfs as last resort backup if the .5 deg gfs is also not available?

Would that just be DSETS = (name_for_gfs25)%gfsp25ptiles,(name_ for_gfs.5),gfs

Regards,
Theo

meteoadriatic
Posts: 1587
Joined: Wed Aug 19, 2009 10:05 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by meteoadriatic » Mon May 25, 2015 12:07 pm

It looks that you answered all questions yourself :)
theocarter2911 wrote:In order to not appear lazy, is it true to say that DSETS = gfs%gfsptile,gfs translates to use full gfs as initialization with gfsptile as boundary files, or if not available use full gfs for everything?
Yes! However I don't see any point of setting that example. gfs and gfsptiles are same data, so equal would be DSETS=gfsptiles,gfs
theocarter2911 wrote:In which case my new DSETS = (name_for_gfs25)%gfsp25ptiles,(name_ for_gfs.5) would be almost right, except how to add in the full gfs as last resort backup if the .5 deg gfs is also not available?

Would that just be DSETS = (name_for_gfs25)%gfsp25ptiles,(name_ for_gfs.5),gfs

Regards,
Theo
I believe that when you upgrade, you would like to set:
DSETS=gfsp25ptiles,gfsptiles,gfs

Which will try 0.25° ptiles first, then 0.5°, and lastly full gfs if both resolution ptiles aren't ready.

About grads, yes, new EMS has new grads and there are differences. Just backup old one ($EMS/util/grads) and replace that directory after you upgrade.

theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 12:52 pm

Hi Ivan,

Okay, am a little bit lost now. Are the ptile data not just the thin strip of "edge" data on the boundary of one's domain which is used as boundary conditions through the forecast run period? I thought you would have to pull one full set of gfs (whether 1.0, 0,5 or 0.25 deg) to use as analysis/initialization for the mesoscale run, and then the ptiles are only for the 3 hourly boundary conditions input?

I like your workaround for grads, I should have thought of that! :-) Only, I had set my server to update this afternoon after I already left work, so it is done by now! Aaaaaarghhhhh

Just teamviewed in, the update did not work, could not find my current install version number, so I was able to save grads. It is suggesting a fresh install, which I will have to look at tomorrow.

meteoadriatic
Posts: 1587
Joined: Wed Aug 19, 2009 10:05 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by meteoadriatic » Mon May 25, 2015 1:29 pm

Forget update, do fresh install always, that's the best option. Then transfer run directory and other things you might need from old. Just don't delete it too soon :) Best way, rename wrfems directory to wrfems_old, then do fresh install into wrfems.

gfsptiles are your mother domain cuts from full GFS, having only variables that WRF will use. This is why they are so small. They do cover whole mother domain, not just borders. Can be used equally for initialization and for boundary conditions.

theocarter2911
Posts: 70
Joined: Fri Mar 08, 2013 5:22 am

Re: Domain sizes and more fine-tuning for .25 GFS

Post by theocarter2911 » Mon May 25, 2015 1:59 pm

I really like the idea of renaming the old one and keeping it, followed by fresh install, will certainly take that advice.

And as for the ptiles, I have been under the wrong impression all these years! You just schooled me big time! :-)

Thank you kindly for all the advice and help,
Theo

Post Reply