zonal statistics, region masks and xarray GroupBy

[emiliom]

Describe the strengths and weaknesses of each of these overlapping packages and functionality. Design examples. For these examples, initially use the CSO snowmodel output and DEM geotiff binning.

• xarray GroupBy
• rioxarray zonal statistics and masking
• regionmask; can be used to create masks of the regions for arbitrary longitude and latitude grids with numpy and xarray; arbitrary regions can be defined easily
• rasterstats. See my brief example from the GeoHackweek vector tutorial. Unlike rioxarray and regionmask, it includes no explicit use of or interaction with xarray.
• salem: use geometries or shapefiles to subset your data. salem is receiving little new development, and should probably be ignored.

[emiliom]

xarray-spatial New package that seems to implement the "raster algebra" functionality on top of xarray. It seems to have broad-based contribution and support.