Technical Features

Additional information on technical features and their maturity levels

Mature Support

Current mature implementations share the following characteristics:

• Virtual stores are used to provide alternate views (defined below) into the data. These views can range from an entire collection, to subsets of a collection, to a subset of a granule.
• Consistent spatial and temporal resolution across each variable (i.e. array) within a virtual store
• Uniform chunk grid and compression scheme across each variable within a virtual store

Note

We define the term “view” in a virtual store context to mean the different ways the same dataset can be represented. A view of the same dataset could be representing the entire dataset as a cube or aggregating some a subset of the dataset. Different views will make sense for any particular dataset given its characteristics and the intended use cases; see Virtual Stores at NASA for some real-world examples.

Supported data formats

Data format support is limited to formats with VirtualiZarr parser implementations. Parsers maintained by VirtualiZarr include:

• HDFParser
• Kerchunk (Parquet or JSON)
• DMRPPParser
• FITSParser
• NetCDF3Parser
• ZarrParser

Parsers maintained outside of VirtualiZarr include:

• virtual-tiff
• hrrr-parser

Custom parsers for additional formats can be implemented using the VirtualiZarr CustomParsers guide.

Developing Support

Non-uniform grids

Some NASA collections present additional complexity due to having varying compression schemes and chunk grids across a logical array (usually across granules). Currently, Zarr assumes a consistent chunk shape and compression across the entire array.

However, support for both varying chunk shapes and compression schemes is in active development and should be available by summer 2026.

Non-cubable data

COMING SOON

No plans for support

• There are no plans to support formats outside of the “Supported data formats” list above at this time.