If you have lost time searching for data collections across catalogs or if you maintain a STAC catalog, this talk will give you concrete tools to improve discoverability for your or your users.

There is one question that everyone must answer when starting a project: which dataset should I use? This is a critically important question and the right answer requires considering many factors, but before you can begin to answer that question you need to know where you should even look for data.

At Development Seed we believe there is a lot of room for improvement in the realm of data discoverability. There are two separate phases of improvements that the community can make today: greater adoption of collection-search in STAC and federated search interface(s) that allow users to search for collections across many catalogs simultaneously.

To address this challenge, Development Seed has been working in partnership with scientists from the NASA/ESA Multi-Mission Algorithm and Analysis Platform on the problem of data discoverability by filling in the gaps in STAC API tooling and building a few new applications that make it possible to discover data collections from a variety of STAC API sources.

stac-fastapi-collection-discovery: a collection-search-only STAC API implementation that performs simultaneous collection search operations on multiple upstream STAC APIs
- api: https://discover-api.maap-project.org
- repo: https://github.com/developmentseed/stac-fastapi-collection-discovery

STAC Collection Discovery web app: a visual interface for searching for data collections across many STAC APIs (via the collection-search-only STAC API)
- app: https://discover.maap-project.org
- repo: https://github.com/developmentseed/stac-collection-discovery

Beyond keyword search, we are also actively evaluating AI-assisted enhancements to collection discovery including semantic similarity search and relevance ranking across heterogeneous catalog metadata. These capabilities are particularly valuable in a federated context where collection metadata quality and schema consistency vary. We will share early findings from this scoping work.