What makes zarr unique is the chunk size. Instead of loading everything from cloud storage, you can just stream the chunks.
- Multispectral data where we can chunk along a cross section of lat long
- Weather data where we can chunk on each variable, e.g. temp, pressure
We use xarray to read zarr files.
Microsoft Planetary computer is equivalent to Google Earth data STAC is spatio temporal asset catalogue
Downscaling means increasing the resolution; from coarse global grid hundreds of kms to regional level. It does this by ScaDo - statistical downscaling vs Dynamic downscaling; think of it as interpolation We can downscale spatialy as well as temporally, from days to hours
Latlong needs a CRS. A geographic crs the lat long lives on a globe, while a projected crs the lat long lives on a flat map Geographic CRS - WGS 84 Projected CRS - mercator, UTM, Lambert, Albers projection; for mapping - introduces distortion
Thus with Geographic you cannot directly measure distances, need cosine distance