A wiki to put your tips, links, references

Below a short description of STAC (Spatio-Temporal Asset Catalog).

What is STAC ?

A STAC collection or catalog is a nested dict (json) listing the metadata of a number of items (e.g. a remote sensing acquisition) and their assets (e.g. raster bands, quicklook, thumbnails, xml files, ...).

In order to limit size and allow fast queries, the nested dict is split into a tree of dict linking to each other (parent and child), from top level to bottom:

catalog > collections > sub-collections > ... > items > assets.

An "item collection" is usually what you will be using at the end, that is a list of items that was subset from a catalog or collection.

Items

The required fields for an item (without type and stac_version that are automatically filled by pystac.Item):

• id: name under which the item will be registered
• geometry: a geojson geometry in dict, see shapely.to_geojson followed by json.loads
• bbox (List[float]): the bounding box of the item
• properties (dict): with at least field "datetime" for the date of acquisition.
• links List[dict]: the links to the parent collections. Thus it is optional for ItemCollection. Optional:
• stac_extensions (List[str]): list of the schema URIs used for validation, e.g. pystac.extensions.eo.SCHEMA_URI

Assets

The only required field for a stac asset is the href (i.e. URI to the file, e.g. a raster file) and the key (usually the band name) under which it will be registered in the Item.

Optional but recommended:

• title (str): usually the band name
• type (str): see list(pystac.MediaType) or https://github.com/radiantearth/stac-spec/blob/master/best-practices.md#working-with-media-types
• roles (List[str]): thumbnail, overview, data, metadata
• extra_fields (dict): all the rest of the metadata, in particular exetnsion fields

Other fields are also recommended, especially for satellite data: https://github.com/radiantearth/stac-spec/blob/master/item-spec/common-metadata.md and https://github.com/radiantearth/stac-spec/blob/master/item-spec/common-metadata.md#instrument

Extensions

Several extensions define the commonly used properties when searching (e.g. the band name/wavelength/bandwidth, the coordinate system, the MGRS code, ...). These standards are referenced in https://stac-extensions.github.io/#grouped-by-maturity.

With the Electro-Optical extension the additionnal fields (extra_fields) are

• eo:bands (List[dict]): each item of the list corresponds to a band of the band file see https://github.com/stac-extensions/eo/#band-object for specs (names and types)
  • name (str): band name
  • common_name (str): the color name for the reflectance bands
  • description (str)
  • center_wavelength (number)
  • full_width_half_max (number)
  • solar_illumination (number)

With the Projection extension the additionnal fields are: See https://github.com/stac-extensions/projection/#projection-extension-specification and https://github.com/stac-extensions/projection/#best-practices

• proj:epsg
• proj:wkt2
• proj:projjson
• proj:geometry
• proj:bbox
• proj:centroid (dict(lat, lon))
• proj:shape (List[height, width]): https://github.com/stac-extensions/projection/#projshape
• proj:transform The projection information can be set at the item level if it is the same for all bands. Recommendations are to set it at the asset level if it differs from

Although the extension fields are all "optional", if used it must be in the good format, the item validation fails otherwise.

Ressources

A good place to get started is: https://stacspec.org/en/tutorials/

In order to build your own STAC with your local data: https://forgemia.inra.fr/umr-tetis/stac/simplestac

Other tutorials:

• on Landsat, but convertible to Sentinel is available here: https://pystac.readthedocs.io/en/stable/tutorials/creating-a-landsat-stac.html

Catalogs

• Copernicus: on the roadmap historical Sentinel assets should be available from STAC API in May 2024