ICON-D2 for more precise 48h forecasts

[an0nfunc]

I was wondering if it would be possible to add support for ICON-D2.

It seems to provide significantly more accurate data (thanks to higher resolutions it seems) than the "normal" ICON model.

Since it's in grib2 format, it would probably require a different method of parsing.

DWD Documentation of ICON-D2 (Section 6.3. Near surface products) DWD OpenData Folder for ICON-D2 (the subfolders are different generation times if I understand correctly)

Also this data would need to be merged with the normal ICON (or ICON-EU, see below) model for >45h forecast.

I would be happy to help, but I thought getting an overview first (and your opinion on this) would probably help.

If you want to see it in action, windy has this model included.

EDIT: After some further reading, the normal station data seems to depend on the ICON 13km resolution. Seeing this, it would probably also help to switch to ICON-EU for the 45h+ forecast.

Resolutions ICON: 13km ICON-EU: 6.5km ICON-D2: 2.2km

Motivation: ICON data differs significantly where I live (in the 5° range temperature wise, precipitation is even worse), so getting more accurate forecast (into hass) is always helpful.