[THREDDS #GHM-262217]: Archived GFS forecasts

Greetings,

Sorry to hear you're having some trouble with the GFS data, but I think I can 
shed some light on this for you.


> This url does work for me for the archived data using the THREDDS server:
> https://www.ncei.noaa.gov/thredds/catalog/model-gfs-004-files/catalog.html

This url works for me as well, but I notice the files listed there don't look 
like gfs-004 files...


> I am new to the usage of GFS dataset and I would be very thankful to you if
> you could please kindly help clear my confusion with respect to the
> following points:

I'll do my best  :)


> (1) The url with model-gfs-004 files contains three-hourly forecasts for
> one day generated at 00, 06,12 and 18 hours
> The url with model-gfs-003 files contains three-hourly forecasts from
> the chosen archived date to the next 16 days generated at those same hours.

The 00, 06, 12 and 18 times you see are the initialization times for the model, 
effectively what time that day the run began.  This is true for all GFS output, 
so all the times should line up between the two sets.  That much does make 
sense.


> For the purpose of getting the weather forecast time series
> observations say from day 1 to day 3 should I individually download from
> the url with model-gfs-004 files for the three separate dates or should I
> get the time series from the url with model-gfs-003 files for the
> corresponding hours? I mean would they amount to the same thing?

My expectation is each set would contain the same data variables.  The only 
difference between the 003 grid and the 004 grid should be the spatial 
resolution.


> (2) What is the spatial resolution of the dataset in this archived url?

The gfs-003 grid is 1 degree spacing while the gfs-004 is 0.5 degree spacing.  


> From the attributes I could find that it uses the GFS T1534 model which has
> the spatial resolution of 12.5 km at the equator. On the other hand when
> using the best dataset from the Best GFS data from UNIDATA the best
> resolution we could get was 0.25 degree, which would be around 27 km at the
> equator. Does grid and pixel resolution mean the same here? 

I'm not sure it does.  I'm not quite as familiar with the T1534 output, but I 
quickly get the impression that data has other differences, such as the 
variables it contains.

The current version of the GFS is initially calculated on a ~13km grid, but in 
a somewhat irregular grid pattern (see finite volume cubed or FV3 for more 
info).  After that it is post-processed to produce output in several other 
forms... gfs-003, gfs-004 and others are the results of this post-processing.  
Since the post-processed data is typically what gets used we're a little more 
used to hearing those 0.25, 0.5, 1 degree resolutions; I hope this made sense.  
NCEI probably doesn't archive the 0.25 degree data due to the amount of space 
that data consumes, it would add up in a hurry.


> I am terribly confused.

After looking at that NCEI url plus some of the others on that site, I think 
there's a problem with that gfs-004 catalog where it's showing you the wrong 
files!  No wonder you're getting confused, I would be too!

It seems to me what that gfs-004 THREDDS catalog is showing is the analysis 
hours of the gfs-003 data.  Analysis hours are simply F000, F003 and F006 for 
each of the four runs, and that's what I see in those listings.  Those files 
also say Grid 3 instead of Grid 4 (Grid 3 and 003 would mean the same thing 
here, as does Grid 4 and 004).  When I download those files, they are indeed 1 
degree data and not 0.5 degree.  And when I compare this to a gfs-003 file, 
they are identical.  This does not appear to be working as intended!

 
> It would be much appreciated if you could please help me clear these doubts
> if you happen to get some time to spare.
> 
> Thanks in advance in anticipation of your kind response and support again.

At this point I would recommend two things...

(1)  I would encourage you to write NCEI at address@hidden and let them know 
their gfs-004 THREDDS catalog may have a problem.  If I'm right about the above 
they will definitely want to know about it.

(2)  Apart from the issues with the NCEI THREDDS links, it's up to you if you 
should use gfs-003 or gfs-004 data.  If you want the increased resolution then 
I'd go with gfs-004, but if you have concerns over bandwidth or data volume you 
may prefer the gfs-003 grid.

Now the good news here is even though the NCEI THREDDS catalog may be broken, 
you can still get gfs-004 data from them using the HTML option.  If you go here 
(https://www.ncei.noaa.gov/products/weather-climate-models/global-forecast) and 
expand Data on NCEI Servers, click on the HTTPS Data Access Links and you can 
drill down from there.  OR, if you prefer or need to use THREDDS, the gfs-003 
set appears to be hosted correctly: 
https://www.ncei.noaa.gov/thredds/catalog/model-gfs-003-files/catalog.html

Sorry for confusion, hopefully NCEI will correct that THREDDS catalog soon.  If 
you have any other questions about that data, writing NCEI at address@hidden 
should get you more answers regarding their data and services.

Hope this helps!

Best,
-Mike

Ticket Details
===================
Ticket ID: GHM-262217
Department: Support THREDDS
Priority: Normal
Status: Closed
===================
NOTE: All email exchanges with Unidata User Support are recorded in the Unidata 
inquiry tracking system and then made publicly available through the web.  If 
you do not want to have your interactions made available in this way, you must 
let us know in each email you send to us.