[ncl-talk] Confusion on GFS files in GRIB2

[dr.hieronymus at posteo.net]

Hello,

I am trying to dig through GFS files in GRIB2 format on NCEP's FTP 
server (https://www.nco.ncep.noaa.gov/pmb/products/gfs/).

Unfortunately, it is not very easy to get a proper overview. Therefore, 
my first question is whether there is documentation somewhere that 
describes the different formats/variants in detail.

What one quickly notice and easy understand is that the GFS files are 
available in different resolutions (0.25°, 0.50°, 1.00°) and each GFS 
file is described by an index file. The GFS files with 0.25° resolution 
have an hourly granularity and the GFS files with 0.50° and 1.00° 
resolution have a three-hourly granularity.

Then, there are the .anl and .f000 GFS files. I don't understand the 
difference between .anl and .f000. Is .f000 not a forecast? Furthermore, 
I noticed that the index files are different regarding .f000 and 
.f001-384 (0.25°) and .f000 and .f003-384 (0.50°, 1.00°).

What also confuses me are the different GFS files, namely pgrb2, pgrb2b, 
pgrb2full and sfluxgrbfFFF. What are the differences exactly? According 
to the corresponding index files, the same physical variables are 
available in all different GFS files. Do these variables always have the 
same values ​​in all GFS files? Doesn't it matter which file I use when 
I have decided on a resolution?

I am actually only interested in the temperature 2 meters above the 
ground. Which file should I simply use?

In addition, I also wanted to know if the initial values for each model 
run (00, 06, 12, 18) are published somewhere.

Thank you in advance for your help.