T1534 GFS .. GFS at 13KM ( along with other upgrades ) can be found HERE

[DTWXRISK]

  YES  the improved  GFS T1534  is   available...   wx bell has it

 

 

  and so does   TROPICALTIDBITS  web site

 

 

 

  if anyone    asks    what is the URL for the  tropical  tidbits  web site ... I am   sorry  BUT I am going to have to  come over to their   house and  beat the sh!t out of them 

[RaleighWx]

I have it on our site too http://models.americanwx.com under the normal GFS tab the 1degree version with some graphics. Also working to add some 13km graphics as well

[das]

  YES  the improved  GFS T1534  is   available...   wx bell has it and so does   TROPICALTIDBITS  web site

 

Never heard of Tropical Tidbits.  What's the URL?

 

[Minnesota Meso]

Never heard of Tropical Tidbits.  What's the URL?

 

C&H.png

http://www.tropicaltidbits.com/analysis/models/

[Absolute Humidity]

way cool

[Crocodile23]

Speaking about GFS, how one can take data for one specific area for a specific parameter(e.g 500 hPa heights)?

The data can be taken from here http://www.nco.ncep.noaa.gov/pmb/products/gfs/

From any of the 2 ftp sites they give(e.g ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/)

 

But there, how i can take for example the 500 hPa heights or surface pressure data for Europe or for Norway?

Any ideas in what folder and what procedure exactly i have to use?

[Absolute Humidity]

Speaking about GFS, how one can take data for one specific area for a specific parameter(e.g 500 hPa heights)?

The data can be taken from here http://www.nco.ncep.noaa.gov/pmb/products/gfs/

From any of the 2 ftp sites they give(e.g ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/)

 

But there, how i can take for example the 500 hPa heights or surface pressure data for Europe or for Norway?

Any ideas in what folder and what procedure exactly i have to use?

Don't know if this helps

http://mag.ncep.noaa.gov/model-guidance-model-parameter.php?group=Model%20Guidance&model=GFS&area=EUROPE&ps=area#

[DTWXRISK]

 

 

Never heard of Tropical Tidbits.  What's the URL?

 

C&H.png

 

 

[bdgwx]

This may only be tangentially related (or not related at all), but I didn't really want to create a separate thread either. But, what is "PRX" that has just now started showing up on the verification metrics?

 

[dtk]

This may only be tangentially related (or not related at all), but I didn't really want to create a separate thread either. But, what is "PRX" that has just now started showing up on the verification metrics?

 

 

Most likely it is the NCO-run version of the 13km Semi-Lagrangian GFS package.  This is the version that is part of the official "30 day evaluation".  In theory, it is essentially the same as HW14 which is run by scientists at EMC (though as you can see, even the same system can give slightly different results once it starts cycling).

[bdgwx]

Great explanation dtk. At first my attention was drawn to the higher score, but then I quickly realized there were fewer samples and the ones that are there are pretty close to PRHW14. Thanks!

[ohleary]

Speaking about GFS, how one can take data for one specific area for a specific parameter(e.g 500 hPa heights)?

The data can be taken from here http://www.nco.ncep.noaa.gov/pmb/products/gfs/

From any of the 2 ftp sites they give(e.g ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/)

 

But there, how i can take for example the 500 hPa heights or surface pressure data for Europe or for Norway?

Any ideas in what folder and what procedure exactly i have to use?

 

Use NOMADS. 

 

http://nomads.ncep.noaa.gov/

 

For example, using the 0.5 degree GFS, click "grib filter" then the date/cycle, then "master".  On this page, you can create a grib file with any parameters/levels you want and specify a subregion by lat/lon. 

 

As a side note, you can click the box at the bottom "Show the URL only for web programming" and it'll show you a URL that you can use in a script to grab that exact data set.

 

Pretty nifty.

[InstantWeatherMaps]

Use NOMADS. 

 

http://nomads.ncep.noaa.gov/

 

For example, using the 0.5 degree GFS, click "grib filter" then the date/cycle, then "master".  On this page, you can create a grib file with any parameters/levels you want and specify a subregion by lat/lon. 

 

As a side note, you can click the box at the bottom "Show the URL only for web programming" and it'll show you a URL that you can use in a script to grab that exact data set.

 

Pretty nifty.

Just a fair warning, this method is quite slow (a few megabits per second) due to the post-processing involved, so if you have the bandwidth and are using a large enough part of the file (and especially if it's part of an automated script), you'll almost always want to download all of it (or just individual variables) from the FTP site without subsetting by region.

[Chinook]

With the GFS changing in December, will this help the initialization of the NAM?

[jgf]

Who is running the 13km GFS?

 

 

edit - ok i see that the GFS T1534 is part of a planned upgrade by NCEP later this year.

 

does that mean that the model output used for the graphics posted in this thread came from NCEP?

 

if so, where can I find the output?

 

Is it available in GRIB format?

 

 

edit again...

 

so i found GFS 0.25 degree at the NOMADS site in the Parallels section

 

but 0.35 degree isn't 13km.., so i'm assuming that's not the  GFS T1534 - is that correct?

[dtk]

Who is running the 13km GFS?

 

 

edit - ok i see that the GFS T1534 is part of a planned upgrade by NCEP later this year.

 

does that mean that the model output used for the graphics posted in this thread came from NCEP?

 

if so, where can I find the output?

 

Is it available in GRIB format?

 

 

edit again...

 

so i found GFS 0.25 degree at the NOMADS site in the Parallels section

 

but 0.35 degree isn't 13km.., so i'm assuming that's not the  GFS T1534 - is that correct?

Actually, there are two separate groups now running versions of the 13 km GFS in real time.  The developers at EMC have been running it for some time.  NCO has recently started running the official 30 day evaluation parallel.  I'm assuming the graphics are being generated by the NCO "official parallel" since the data delivery (including time) should be much more reliable than that being done by developers.

 

Your last point is likely incorrect.  The model doesn't run on a regular grid nor does it run on pressure levels.  All of the pressure-grib data that is available is post-processed.  The 0.25 degree files are interpolated from the model native grid (and the post processing includes the generation of many, many derived fields).

[dtk]

With the GFS changing in December, will this help the initialization of the NAM?

Yes, in two ways (one more direct than the other).

 

For one, the NAM uses partial cycling whereby it is effectively restarted from the GFS state twice per day.  The way it actually works is that twice per day, the NAM goes "backwards" in time, starts from the GFS, and then performs catch up data assimilation cycles to the initialization time.  The RAP works similarly, btw.

 

Secondly, the NAM utilizes a hybrid-variational solver for its data assimilation, where the ensemble part is taken from a GFS-based EnKF. Since the 13km GFS upgrade also includes assimilation changes including an increase in the ensemble resolution to T574 (from T254) for the EnKF part (not to be confused with the GEFS), the NAM and RAP will benefit.

[bdgwx]

Is 4DVAR/EnKF still in the works for the GFS? 

 

Also, what's the latest on the FIM? Is it ever going to go into operational use?

[jgf]

will the old 0.5 degree GRIB output continue to be available?

[ohleary]

Just a fair warning, this method is quite slow (a few megabits per second) due to the post-processing involved, so if you have the bandwidth and are using a large enough part of the file (and especially if it's part of an automated script), you'll almost always want to download all of it (or just individual variables) from the FTP site without subsetting by region.

 

He/she is looking for one variable for a pretty small region.  Should be pretty instantaneous, I just did it and it gave me the file right away.  One parm over Europe is a 17KB file. 

[Crocodile23]

Use NOMADS. 

 

http://nomads.ncep.noaa.gov/

 

For example, using the 0.5 degree GFS, click "grib filter" then the date/cycle, then "master".  On this page, you can create a grib file with any parameters/levels you want and specify a subregion by lat/lon. 

 

As a side note, you can click the box at the bottom "Show the URL only for web programming" and it'll show you a URL that you can use in a script to grab that exact data set.

 

Pretty nifty.

 

Thank you. Indeed pretty straightforward.

 

 

Just a fair warning, this method is quite slow (a few megabits per second) due to the post-processing involved, so if you have the bandwidth and are using a large enough part of the file (and especially if it's part of an automated script), you'll almost always want to download all of it (or just individual variables) from the FTP site without subsetting by region.

 

Thank you too. No harm for any alternatives. :-)

 

 

[dtk]

Is 4DVAR/EnKF still in the works for the GFS? 

 

Also, what's the latest on the FIM? Is it ever going to go into operational use?

In terms of your first question, yes, but with one clarification.  It is not actually 4DVAR (or hybrid variants thereof), but a different ensemble-based algorithm called 4DEnVar.  I am happy to pass along details if interested.  The main difference is that 4DVAR based algorithms actually use a linearized model (and adjoint) to propagate back and forth within a window for each iteration of the minimization process in the analysis procedure.  In 4DEnVar, the "4d-ness" is instead prescribed by using linear combinations of 4D Ensemble perturbations instead of explicitly using the (linear) model itself in the minimization.  Or, another way to say it is that 4DVAR creates a 4D analysis that is actually a trajectory of a model, whereas 4D EnVar creates a (time-discretized) 4D analysis that is a linear combination of ensemble states valid at each of the specified times.  There is a huge computational advantage in doing things this way. Also, the full nonlinear model (with all the physics) is used instead of a linearized version, which may have some theoretical advantages.

 

For the FIM, I remain skeptical.  There is a project within NWS to explore the "next generation global prediction system (NGGPS)", which has several candidate models, all of which are non-hydrostatic.  One of these candidates is the nonhydrostatic version of the FIM (called the NIM).  My hunch is that if the FIM/NIM is ever to see the light of day in terms of NWS operations, it will only be in the form of a multi-model ensemble.  I remain skeptical that even this will happen, however.

[Harry]

In terms of your first question, yes, but with one clarification.  It is not actually 4DVAR (or hybrid variants thereof), but a different ensemble-based algorithm called 4DEnVar.  I am happy to pass along details if interested.  The main difference is that 4DVAR based algorithms actually use a linearized model (and adjoint) to propagate back and forth within a window for each iteration of the minimization process in the analysis procedure.  In 4DEnVar, the "4d-ness" is instead prescribed by using linear combinations of 4D Ensemble perturbations instead of explicitly using the (linear) model itself in the minimization.  Or, another way to say it is that 4DVAR creates a 4D analysis that is actually a trajectory of a model, whereas 4D EnVar creates a (time-discretized) 4D analysis that is a linear combination of ensemble states valid at each of the specified times.  There is a huge computational advantage in doing things this way. Also, the full nonlinear model (with all the physics) is used instead of a linearized version, which may have some theoretical advantages.

 

For the FIM, I remain skeptical.  There is a project within NWS to explore the "next generation global prediction system (NGGPS)", which has several candidate models, all of which are non-hydrostatic.  One of these candidates is the nonhydrostatic version of the FIM (called the NIM).  My hunch is that if the FIM/NIM is ever to see the light of day in terms of NWS operations, it will only be in the form of a multi-model ensemble.  I remain skeptical that even this will happen, however.

 

 

Interesting.. That FIM is often cited in AFD's at this nws office here ( GRR ) and has been for some time.

[thunderbolt]

http://www.nws.noaa.gov/os/notification/tin14-46gfs.htm

[bluewave]

http://www.nws.noaa.gov/os/notification/tin14-46gfs.htm

 

MOS refresh at the same time.

 

http://www.nws.noaa.gov/mdl/synop/2014refresh/gfsmos2014.php

[bdgwx]

In terms of your first question, yes, but with one clarification.  It is not actually 4DVAR (or hybrid variants thereof), but a different ensemble-based algorithm called 4DEnVar.  I am happy to pass along details if interested.  The main difference is that 4DVAR based algorithms actually use a linearized model (and adjoint) to propagate back and forth within a window for each iteration of the minimization process in the analysis procedure.  In 4DEnVar, the "4d-ness" is instead prescribed by using linear combinations of 4D Ensemble perturbations instead of explicitly using the (linear) model itself in the minimization.  Or, another way to say it is that 4DVAR creates a 4D analysis that is actually a trajectory of a model, whereas 4D EnVar creates a (time-discretized) 4D analysis that is a linear combination of ensemble states valid at each of the specified times.  There is a huge computational advantage in doing things this way. Also, the full nonlinear model (with all the physics) is used instead of a linearized version, which may have some theoretical advantages.

 

For the FIM, I remain skeptical.  There is a project within NWS to explore the "next generation global prediction system (NGGPS)", which has several candidate models, all of which are non-hydrostatic.  One of these candidates is the nonhydrostatic version of the FIM (called the NIM).  My hunch is that if the FIM/NIM is ever to see the light of day in terms of NWS operations, it will only be in the form of a multi-model ensemble.  I remain skeptical that even this will happen, however.

 

Great information. It looks like there is a lot information available on the net already if you google for "4DEnVar" including some journal articles on the AMS site...too bad I'll have to wait a couple years until they're free to access. Thanks!

[Crocodile23]

The new GFS(PRHW14 in the diagram) doesn't impress me. And quite obviously why. Because the old one(GFS) is quite on par with the new one. :(

(PRX which might be the new one also with some changes does seemingly quite well but it's only for the last 18 days where as we see the prediction ability is higher in all models.)

 

Any comments on this and any comments from the NCEP/NOAA about these not so good news?

 

[Ellinwood]

The new GFS(PRHW14 in the diagram) doesn't impress me. And quite obviously why. Because the old one(GFS) is quite on par with the new one. :(

(PRX which might be the new one also with some changes does seemingly quite well but it's only for the last 18 days where as we see the prediction ability is higher in all models.)

 

Any comments on this and any comments from the NCEP/NOAA about these not so good news?

 

Granted, that's just one parameter to base the skill off of. I would be interested to see the gain and setback analysis of all of the pertinent variables (especially things like QPF and 2mt) before praising/condemning the new version.

[olafminesaw]

The tropical tidbits site seems to be down, anywhere I can find it, on a free site?

[bdgwx]

The new GFS(PRHW14 in the diagram) doesn't impress me. And quite obviously why. Because the old one(GFS) is quite on par with the new one. :(

(PRX which might be the new one also with some changes does seemingly quite well but it's only for the last 18 days where as we see the prediction ability is higher in all models.)

 

Any comments on this and any comments from the NCEP/NOAA about these not so good news?

It looks like PRX wasn't being evaluated during the period of time where there was a big drop off in data being assimilated so that (along with the fewer samples) probably explains why it is higher. During the warm season PRHW14 was consistently scoring higher than the operational version; not a huge jump, but improvement nonetheless.