Make the CRS / Scale / Geometry defaults more user-friendly.

[alxmrs]

There are some differences between how Xee opens data and "what users expect" (thanks for your feedback, @shoyer).

[tylere]

This issue doesn't have enough information to be actionable. @alxmrs could you provide more details on "what users expect" for the defaults?

[alxmrs]

Alas, I should have written my conversation with Stephan down. I can still recall what it was about, however. When you open a dataset with Xee, today, does the Xarray dataset resemble the underlying data as if it were opened from a well-organized file, like a NetCDF or Zarr (or COG)? That’s the crux of the issue, albeit subjective.

What is expected anyway? I’m reminded a lot of discussions I’ve had with @briannapagan on the GeoZarr spec. “Natural” will feel differently for users of NetCDF (e.g. atmosphere, oceans, and land geosciences folks) vs users of COGs (e.g. EO scientists, satellite industry folks). In addition, EE varies in the types of datasets presented as rasters. Each dataset will make more sense in one of the two models just mentioned.

While EE is a commercial SaaS, I suspect this interface is a useful prototype of what GeoZarr could end up looking like. Not that we have to boil the ocean right now…

I offer this background to provide context to specific concerns. I’ve imagined that all datasets opened with Xee should look like opening ERA5-Land by default. Is that reasonable? What data or users would those assumptions not be suitable?