GFS next major version - 4D-VAR and hourly output to 120h

All other topics about postprocessing model data (GrADS and other software), about other numerical weather prediction software (including WRF-NMM and WRF-ARW discussion unrelated to UEMS/WRF EMS), and general meteorology talk go in this forum.
windyweek
Posts: 28
Joined: Thu Aug 21, 2014 8:46 am

GFS next major version - 4D-VAR and hourly output to 120h

Post by windyweek » Sat Mar 12, 2016 2:52 pm

It looks like the next major version of the GFS model is coming soon!

Major improvements include:
- 4D-VAR data assimilation (or more precisely 4DHybrid En-VAR). This is seen by many as one of the weaknesses of the GFS model (currently using 3D-VAR). The other major global models - CMC (the Canadian), ECMWF and UKMO are all running 4D-VAR data assimilation for long enough.
- Hourly output to the 120th hour.
- Improved use of satellite data.
- Land use improvements (unclear to me however if this is only for US regions or globally).

Horizontal and vertical resolutions remain the same.
Preliminary date for the upgrade becoming operational is May, 17th 2016.

I'm particularly very excited to see how the improved data assimilation will impact GFS quality and skill.
The hourly output to the 120th hour should also add a good value to the LBC when we all run our regional models.

Below is also a nice presentation which reveals more details about the upcoming upgrade:
http://www.emc.ncep.noaa.gov/impdoc/GFS ... al2016.pdf

Cheers,
Ivan

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by meteoadriatic » Sat Mar 12, 2016 4:31 pm

Could say finally about new assimilation, but at the same time ECMWF upgraded their systems so they run now at just 9km native.... GFS is again way behind even after it gets new planned stuff :/

windyweek
Posts: 28
Joined: Thu Aug 21, 2014 8:46 am

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by windyweek » Mon Mar 14, 2016 3:38 pm

I don't particularly think that the native model grid resolution of the GFS model is really that bad at the moment. For a global model, 13km is a good resolution. Improved physics and better data assimilation techniques are much more needed now. And hopefully we'll see one of these issues addressed soon.

One thing is for sure - we are living in exciting times from numerical weather prediction perspective. The competition b/w the global models (ECMWF vs. GFS vs. UKMO vs. CMC) is good for everybody, striving for continuous improvements just all the time.

Although ECMWF has the lead now (and for many years), NCEP/NWS may implement completely new and fresh global model - MPAS in the upcoming years. GFS is aging already and NWS are well aware of that. Choosing a new global model for the next several decades might be a game changer for the entire numerical weather prediction industry.

surgeon
Posts: 44
Joined: Wed Dec 08, 2010 1:34 pm
Location: Poland
Contact:

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by surgeon » Thu Apr 14, 2016 8:08 pm

Any new Vtable for new GFS? Anywhere? ;)
Any changes to APCPsfc? (in old GFS: grib006 has APCPsfc t, grib009 has APCPsfc t +APCPsfc t-1, grib012 has APCPsfc t, grib015 has APCPsfc t + APCPsfc t-1 etc).

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by meteoadriatic » Fri Apr 15, 2016 4:47 am

My NMMB works with old Vtable, I don't think anything is changed in regard to that. Of course, precipitation fields are not used in Vtable.

Yes those are mess... I have no idea how to script that in GrADS to display correctly in hourly output :evil:

surgeon
Posts: 44
Joined: Wed Dec 08, 2010 1:34 pm
Location: Poland
Contact:

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by surgeon » Fri Apr 15, 2016 2:29 pm

meteoadriatic wrote:My NMMB works with old Vtable, I don't think anything is changed in regard to that.
I thought that levels in new GFS are changed. :oops:

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by meteoadriatic » Fri Apr 15, 2016 4:28 pm

No idea, levels look OK to me, and output from model that runs on it, is pretty much good:
http://gamma.meteoadriatic.net/nmmb/maps/

Do you know something particular about levels that was changed?

surgeon
Posts: 44
Joined: Wed Dec 08, 2010 1:34 pm
Location: Poland
Contact:

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by surgeon » Fri Apr 15, 2016 10:12 pm

Seems like few more levels only on the top:
New: zdef 31 levels 100000 97500 95000 92500 90000 85000 80000 75000 70000 65000 60000 55000 50000 45000 40000 35000 30000 25000 20000 15000 10000 7000 5000 3000 2000 1000 700 500 300 200 100
Old: zdef 26 levels 100000 97500 95000 92500 90000 85000 80000 75000 70000 65000 60000 55000 50000 45000 40000 35000 30000 25000 20000 15000 10000 7000 5000 3000 2000 1000

Real.exe wants num_metgrid_levels 32 instead of 27.
At least I run succesfully WRF 3.8 on new GFS. 8-)

And I still have problem to draw correct APCPsfc for pure GFS. :x

meteoadriatic
Posts: 1510
Joined: Wed Aug 19, 2009 10:05 am

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by meteoadriatic » Sat Apr 16, 2016 6:28 am

surgeon wrote:Seems like few more levels only on the top:
New: zdef 31 levels 100000 97500 95000 92500 90000 85000 80000 75000 70000 65000 60000 55000 50000 45000 40000 35000 30000 25000 20000 15000 10000 7000 5000 3000 2000 1000 700 500 300 200 100
Old: zdef 26 levels 100000 97500 95000 92500 90000 85000 80000 75000 70000 65000 60000 55000 50000 45000 40000 35000 30000 25000 20000 15000 10000 7000 5000 3000 2000 1000

Real.exe wants num_metgrid_levels 32 instead of 27.
At least I run succesfully WRF 3.8 on new GFS. 8-)
I see. NMMB's nemsinterp.exe (does the same thing as real.exe for WRF) works fine with this and doesn't require changes. I didn't tried run WRF with new GFS yet.

surgeon wrote:And I still have problem to draw correct APCPsfc for pure GFS. :x
Me too, looks horrible. Also, cloud cover looks strange.. like it is interpolated between 6h points, just like precipitation... I have no idea:
http://gamma.meteoadriatic.net/images/gfs/para/

oib
Posts: 117
Joined: Mon Jan 31, 2011 3:29 pm

Re: GFS next major version - 4D-VAR and hourly output to 120h

Post by oib » Sat Apr 16, 2016 1:57 pm

surgeon wrote:And I still have problem to draw correct APCPsfc for pure GFS. :x
meteoadriatic wrote:Me too, looks horrible.
Accumulation period in current GFS official is:

Code: Select all

ForecastHour    AccPeriod (hour)
03              3
06              6
09              3
12              6
15              3
...             ...
Are you saying that accumulation period is changed in current GFS para?

Post Reply

Who is online

Users browsing this forum: Baidu [Spider] and 1 guest