Catalogue integrity report

Figures are recomputed each night from the public catalogue API. Documents are validated against the published schemas, and declared assets are matched against the attached files and their storage replica states. Items whose declared assets or stored files changed, and newly published items, wait in a queue until a byte check covers them. Edits elsewhere in a document keep the earlier check. Each night the oldest waiting items and a random sample are downloaded and verified against their recorded checksums, within a budget per collection of which the sample keeps up to half, and failed checks are retried. A rotating seventh of the catalogue is inventoried again every night.

Counts include results carried over from earlier runs for items whose document has not changed since. Changed counts the items whose document differs from the previous run. The version is the processing version recorded in the document, or the identifier suffix for items without one. Assets stored outside the catalogue, such as the Copernicus product links on Sentinel-2 items, are counted as external and are not verified. Surplus files are attached to an item but referenced by none of its assets. They do not make an item incomplete, yet they occupy storage.

Sentinel-2 source images with at least one product, per use case and version. Reasons for a missing product are outside the scope of this report and are tracked by the processing platform.