The data-family matrix
Rows are facilities; columns are data families, grouped by the six stages above. A filled cell means the facility’s capture-to-science chain contains that family—not that every product is downloadable today or was recorded directly. Profiles distinguish current public holdings, official-pipeline derivatives, and scheduled products.
Select any row to open that facility’s profile · rows are keyboard-navigable with ↑ ↓ and Enter
Detector-to-discovery profiles
Astronomical field dictionary
These are the recurring coordinates, measurements, uncertainties, masks, calibrations, and inferred quantities found across archives. Names vary by mission; meanings often rhyme.
How the data lands on disk
The same scientific quantity can arrive as a multidimensional image extension, a binary table row, a database column, or a streaming packet. Format determines how you query and validate it.
What “all data” means here
Scope
This atlas inventories science-relevant data classes, not every telemetry mnemonic or every column in every release. The unit is a meaningful measurement or product family: detector ramps, visibilities, event lists, WCS, extracted spectra, covariance, classifications, and the like. Field names shown in facility profiles are representative and intentionally preserve mission vocabulary. The matrix describes each facility’s end-to-end data chain; availability is release-dependent, and scheduled products are labeled in their profiles.
- Captured means recorded by detector or observatory telemetry.
- Measured means calibrated or fitted from captured samples.
- Inferred means model-dependent physical interpretation.
- Delivered means exposed in a public archive, release, or science-ready package.
Do not erase provenance
A catalog value without its uncertainty, flag, calibration context, epoch, pipeline version, and selection function is an incomplete observation. Null is not zero. A non-detection is not a flux of zero. A coadd is not an epoch. A model parameter is not a detector count. Covariance can dominate a derived result even when every individual error bar looks small.
Official data models & handbooks
Product names and schema examples are grounded in facility documentation. Archive schemas evolve; follow the linked release documentation before production analysis.