ATMODAT

ATMODAT is a recommendations framework for making atmospheric research data more FAIR. It focuses on the metadata, identifiers, file formats, and documentation needed so that observational and modeling datasets remain interpretable beyond the original project that produced them.

The gap it addresses

Atmospheric datasets are often technically readable but still hard to reuse because key context is missing: sensor setup, calibration history, variable definitions, coordinate systems, processing steps, or license information. ATMODAT addresses that gap by defining what should be documented alongside the data itself.

Core expectations

1. Persistent identifiers: People, institutions, datasets, and referenced resources should use stable identifiers such as ORCID, ROR, or DOI where possible.
2. Structured metadata: Temporal coverage, spatial reference, variables, units, and methodological context should be described in a machine-readable way.
3. Controlled vocabularies: Keywords and variable descriptions should use shared vocabularies where possible to improve interoperability.
4. Clear licensing: Reuse conditions should be stated explicitly and ideally machine readable.

File-format guidance

ATMODAT strongly aligns with self-describing scientific data formats such as netCDF, including clear coordinate definitions, rich variable attributes, and standard conventions such as CF conventions. In practice, that means datasets should expose time, space, and variable metadata directly in the file rather than leaving interpretation to a PDF or README alone.

Relation to FAIR and provenance

ATMODAT is best understood as an implementation layer for FAIR atmospheric data. It turns general principles into concrete documentation requirements and encourages stronger provenance for derived products and workflow outputs.

What makes it usable rather than aspirational

ATMODAT ships a checker. That is the difference between it and a set of recommendations, and it is the whole reason it works.

“Document your data well” is advice nobody follows consistently, because the producer knows everything and finds most of it too obvious to record, while the consumer reads it years later knowing nothing. A validator that passes or fails removes the judgment call. This is the same mechanism behind the UC2 standard, arrived at independently, and the pattern is worth generalising: FAIR states principles, and something has to turn them into a check before anything changes.

Beneath it sit metadata as the descriptive layer, provenance as the processing record, and netCDF with CF conventions as the file-level substrate ATMODAT profiles.

See also: FAIR, Metadata, Provenance, UC2 data standard, netCDF, CF conventions, MOC Open Science Data and Knowledge Graphs