Zarr is a community project that develops specifications and software for storing large N-dimensional typed arrays. It’s efficient for storing data across multiple files with a single index entrypoint of metadata that also enables loading actual data only when you need it.
GeoZarr enhances the specification for geospatial data by standardizing projections, pyramids, and geographic transformations. Virtual Zarr then lets you build unified, performant access to multiple files from many archival formats, so the data are usable as Zarr without duplicating the original data in NetCDF4, HDF5, GRIB, GeoTIFF.

Why you should care about this:
- Many large data distributors are providing data in Zarr format including; NASA, ESA, and NOAA
- Zarr enables cloud native access for much easier sub-sampling, parallelization, and distributed computing
- Virtual Zarr simplifies access by abstracting away how data are split across files

Come to this talk:
- To understand what mulit-dimensional (N-dimensional) data cubes are
- See some examples of GeoZarr and Virtual Zarr in the wild
- What the latest advances in the GeoZarr specification means for you, and what’s next
- To learn about open source libraries for working with Zarr, including computation and browser based visualization.