AIP catalogues: two discovery models, shared programme delivery

An institutional catalogue helps people understand what a dataset describes, where it came from and how to access it. Earth-observation discovery adds another question: which asset covers a place and time and can be used by a downstream application?

FAO’s catalogue and agro-informatics services support a broad range of projects. My work connects that institutional need with two distinct discovery models: CKAN and ISO metadata for descriptive records, and STAC for spatial and temporal asset discovery.

The CKAN lane builds on structured metadata, schema-based editing and validation, and documented Solr projections in ckanext-jsonschema. It is the metadata-catalogue part of the work, associated with FAO’s public data catalogue and agro-informatics portal.

The STAC lane uses SFEOS, the STAC FastAPI implementation with Elasticsearch and OpenSearch backend support. I introduced it at FAO for the Agro-informatics Platform’s STAC catalogue. Its public documentation describes item and collection search across spatial, temporal and property criteria.

The diagram shows those discovery responsibilities side by side. It does not assert automatic synchronization between the catalogues, a shared database or CKAN authority over every STAC record. Each lane needs its own publishing, identity and access rules.

I contributed SFEOS patches and suggestions while adopting it for the FAO catalogue. Public pull requests include an Elasticsearch Serverless working draft and a collection-search draft in 2024; those proposals were closed without merging. A team contribution for custom mappings and sortable fields, submitted through my account, was merged in April 2026.

My public issue reports address multi-catalog organization, collection bounding-box mappings, index refresh visibility and pagination with sorting. These are concrete integration concerns: they connect data semantics with the search behavior a user sees. They do not establish that I authored every later upstream capability.

I also participated in standardization work on ISO 19115-4. The official ISO entry identifies ISO/TR 19115-4 as the Technical Report on JSON Schema implementation of metadata fundamentals, under publication when inspected in October 2026. My participation is recorded in my own account; I do not claim sole authorship or certification.

A retrospective alternative is to force descriptive metadata and asset discovery into one undifferentiated catalogue model. A single interface can be convenient, but it still needs to preserve provenance, spatial/temporal semantics and the ownership of each record type. The two-lane view makes those responsibilities easier to discuss before proposing any integration.

The programme work includes the FAO ISO metadata catalogue, SFEOS adoption for AIP STAC discovery and support for FAO projects through the public data and agro-informatics services. That institutional scope reflects my direct contribution; the platform is work delivered with colleagues.

Public upstream evidence includes the merged team patch for configurable mappings and sortable fields and earlier design discussions. This distinguishes delivered upstream work from proposals that informed discussion but were not merged.

My AIP STAC work also supported the institutional work presented in a 2025 UN geospatial presentation. The linked presentation is programme context; I do not claim authorship of its slides.

A metadata schema is not enough by itself. A collection bounding box must be represented in a way the search backend can query; a pagination token depends on ordering; a newly indexed record may not be immediately visible. These issues turn standards language into implementation work.

For a further review I would trace publishing, updates, deletions and visibility rules separately in both catalogues, then document any intended links between them. A shared programme does not by itself establish a shared consistency or authorization contract.