geozarr generated by V1 converter (convert_s2_optimized) doesn't meet all cf compliance requirements from geozarr minispec

[lhoupert]

When exploring the v0 and v1 geozarr data I realized a few attributes for CF compliance requirements (described in geozarr-minispec ) were missing from the v1 geozarr (but present in the v0). For example:

• Standard Names: Data variables MUST include CF-compliant standard_name attributes. The implementation validates these against the official CF standard names table
• Array Dimensions: All arrays MUST include _ARRAY_DIMENSIONS attributes for Zarr V2 compatibility

The geozarr_examples have these variables:

data-model/tests/test_data_api/geozarr_examples/S2B_MSIL1C_20250113T103309_N0511_R108_T32TLQ_20250113T122458.json

Lines 150 to 154 in 06df224

	"standard_name": "toa_bidirectional_reflectance",
	"_ARRAY_DIMENSIONS": [
	"y",
	"x"
	],

while the optimized_geozarr_examples don't.

Should we create additional tests that validate the CF-compliance of our format or we have all we need already?

[emmanuelmathot]

This is indeed part of the geozarr spec 0.4 but we do not use it. Maybe it is worth adding it to be compliant

[lhoupert]

In the doc https://eopf-explorer.github.io/data-model/#standards-compliance, we are saying :

📋 Standards Compliance[¶](https://eopf-explorer.github.io/data-model/#standards-compliance)

    - GeoZarr 0.4 specification compliance
    - CF conventions for scientific metadata
    - _ARRAY_DIMENSIONS attributes on all arrays
    - Grid mapping variables with proper CRS information
    - Multiscales metadata structure

[emmanuelmathot]

As we use only Zarr V3, let's discard _ARRAY_DIMENSIONS

[emmanuelmathot]

pending https://github.com/EOPF-Explorer/coordination/issues/199

[lhoupert]

relates to #17

[emmanuelmathot]

insteresting tooling to use: https://github.com/zarr-developers/geozarr-toolkit