Question: I use -small_grib to make a smaller domain.  Why is the "smaller"
file almost 3x larger?

Answer:

-small_grib creates a new grib message.  The default compression is "off".
You can use jpeg2000 compression by "-set_grib_type jpeg".

Question: What is g2clib and should I prefer the internal or g2clib routines for decoding?

Answer: g2clib is a library for encoding/decoding grib2. Some fields have to be unpacked by g2clib and other have to be unpacked using wgrib2 routines. Some files can handled by either set of codes. The wgrib2 codes are slightly faster. The g2clib codes do not like new templates.

In the fall of 2008, a problem was discovered with the NCEP-NCO encode/decode routines when processing a constant field with non-zero decimal scaling. The problem remained hidden because if you used the NCEP-NCO encode and decode routines, you got the correct result. Fortunately constant fields are rare and people rarely look at them. However, you may prefer the g2clib routine if using files created with the NCEP-NCO routines.

Fall 2009: The current wgrib2 writes constant fields that WMO standard and can be read correctly by the NCEP-NCO routines. For most common packing (jpeg, png, complex, simple), you can decode message using the NCEP-NCO routines (constant field problem) or with the wgrib2 internal routines which follow the WMO standard. the WMO standard.

Question: Why does wgrib2 create smaller complex-packed files than cnvgrib?

Answer: Complex packing can be done different ways. wgrib2 uses a better algorithm than cnvgrib.