Gerrit Holl

For the file I just described, we have `ls["3a"].attrs["start_time"] = datetime.datetime(1950, 1, 1, 0, 0)`. This leads to pyninjotiff trying to set a negative value to a tag which has...

It _is_ in the data... what do we do here then? ``` ipdb> self._data["scnlinyr"][4017] 1950 ipdb> self._data["scnlindy"][4017] 1 ipdb> self._data["scnlintime"][4017] 0 ```

Should this PR handle this and put something sensible in the `start_time` and `end_time` attributes?

The header is normal: ``` In : print([(x, self._header[x]) for x in self._header.dtype.names if x.startswith("start") or x.startswith("end")]) [('startdatajd', memmap([26060], dtype=int32)), ('startdatayr', memmap([2021], dtype=int16)), ('startdatady', memmap([128], dtype=int16)), ('startdatatime', memmap([52532877], dtype=int32)), ('enddatajd',...

I could add a check and reject any start/end date that isn't strictly between the header start/end date.

Yes, the `standard_name` has changed. I think the most correct would be to remove the `units` attribute altogether or set it to `None`, if it cannot be meaningfully defined. The...

When I produce RGBs for NinJo, I'm setting units attribute to `"N/A"`. In the Satpy ninjotiff writer, there is a guard in which `convert_units` is only called when the `units`...

Yes, but is `hncc_dnb` a quantity? Or `histogram_dnb`? It doesn't represent a physical value anymore, does it?

From [the HNCC paper](https://www.tandfonline.com/doi/full/10.1080/01431161.2017.1338838), although the gain factor is a complicated calculation, it is itself unitless, so does not change the units of the radiance. So I guess that it's...

The unit "1" is correct for a unitless quantity, but `adaptive_dnb` and `histogram_dnb` are not quantities. The CF conventions do not define what a quantity is, but ISO 80000-1 does:...