Upfront: I have very little experience dealing with geo-data properly. But here a few thoughts.
Types of geo-data I'd be interested in:
- raster (1d, 2d and 3d)
- station data (time series at 1 location)
- trajectories (time series at a moving point)
- vector data, like polygons and such (this is probably hard)
The closest I've ever come to deal with these things "properly" is when doing NetCDF data. The conventions they use there might be of some use/guidance:
We could/should build on DataArrays of the DataFrames package as that has support for missing data, and on Calender.jl for time/date.
I couldn't lead such an effort, but I would help a little and would certainly be interested in following the development.
Lastly, better keep it simple to not get overextended like they did with the DataFrame package: