Where to get GRIB2 files for testing...
1. According to this post, there seems to be some trouble with GRIB2 files made by conversion from other sources. (e.g. ibi.nc to GRIB2 using CDO).
What is needed here is to discover whether the conversion process is in error, or the OCPN GRIB2 decoding is faulty, or both. The way to determine that is by comparing OCPN display against another well known GRIB2 viewer, and looking for differences. Frankly, there are very few simple GRIB2 viewers available online. The best seem to be on Android...
You may also find the wgrib2 utility to be useful.
Climate Prediction Center - wgrib2: grib2 utility
We should focus on determining if the OCPN GRIB2 decode/display is accurate and reliable for a properly formed GRIB2 file. Lets leave the "how do I convert/manipulate GRIB2 files?" workflow questions for another thread.
2. did-g has shown us a way to download filtered GRIB2 files from the French Arome model. As has been noted, the full model available by online download is huge, and so not of much practical use "in-the-raw" for actual navigation
. For practical testing, take a look at this post, and give it a try.
As an example, the following line:
curl -o today.grb --compressed "http://22.214.171.124/grib/arome?x=-2X=2&y=47&Y=48&r=w"
(36 hour wind forecast
by hour in a quite small area) gave a file size of 1.2 Mb.
I find that the current OCPN GRIB2 decodes and displays correctly all of the parameters available by this server.
e.g. current,wind 10m,gust,temp 2m,pressure,rain,cloud cover
This suggests that at least the minimum standard GRIB2 variables available from the AROME model are compatible with OCPN GRIB2.
3. For NOAA GFS model forecasts, I find it best to use the grib_filter interface provided here:
NOMADS-NOAA Operational Model Archive and Distribution System
Choose the model you want, select the "grib_filter" link, and then choose the variables, timescale, and geographical area you want downloaded. This interface produces very small file download sizes, so is quite useful for testing various options of GRIB2.