Written and maintained by CASRAI Editorial Board
Last updated
The Climate and Forecast (CF) Metadata Conventions are the standard community rules for describing Earth and environmental science datasets stored within NetCDF (Network Common Data Form) containers. Designed to make scientific binary files fully self-describing, the CF conventions define precise attributes and coordinate variables that document what physical quantity a variable represents, its exact spatial-temporal coordinates, its measurement units, and its statistical processing history. CF compliance is mandatory across major international scientific efforts, including the Coupled Model Intercomparison Project (CMIP), the Copernicus Climate Data Store, and ECMWF reanalysis archives.
By enforcing standardized metadata syntax and controlled vocabularies, CF conventions allow automated data processing libraries (such as Python’s xarray) and visual tools to parse multidimensional grids without manual intervention. Understanding how to structure datasets to comply with CF conventions—currently at Version 1.13—is essential for environmental researchers seeking to adhere to FAIR data principles and deposit reproducible data into scientific repositories like the PANGAEA Data Publisher. To learn how to open and inspect these files directly, consult our practical tutorial on how to open NetCDF files.
What It Is
The CF conventions define a standardized schema of global attributes, variable attributes, and coordinate variables designed for implementation with the NetCDF API. Initially developed for atmospheric, oceanic, and climate numerical simulation models, CF has expanded across geosciences, hydrology, and remote sensing observation systems. Rather than altering the underlying binary storage layout of NetCDF-3 or NetCDF-4/HDF5, CF provides semantic metadata layer rules.
Under the CF conventions, every compliant dataset includes a global Conventions attribute (for example, :Conventions = "CF-1.13") that asserts adherence to a specific version of the standard. At the variable level, CF eliminates ambiguity by requiring variables to pair standardized names with valid physical units. For instance, an atmospheric temperature field must declare its units using Unidata’s UDUNITS library (such as units = "K" or units = "degC") and bind to a recognized entry from the controlled CF Standard Name Table (such as standard_name = "air_temperature").
Who Runs It and What It Covers
The CF metadata conventions are maintained as an open, community-governed standard overseen by three formal leadership bodies established in the CF Constitution:
- CF Governance Panel: Chaired by Ethan Davis, this body oversees high-level strategy, constitutional updates, and relationships with partner standards organizations such as the Open Geospatial Consortium (OGC) and the World Meteorological Organization (WMO).
- CF Conventions Committee: Chaired by Jonathan Gregory with David Hassell serving as secretary, this technical committee evaluates, reviews, and votes on proposed structural modifications and extensions to the core conventions document.
- CF Standard Names Committee: With Alison Pamment serving as secretary, this committee reviews and moderates proposals for adding new physical quantities to the CF Standard Name Table and standardized region vocabularies.
The standard covers a comprehensive scope of multidimensional geoscientific structures:
- Spatial and Temporal Coordinate Systems: Explicit attribute standards for latitude, longitude, vertical coordinates (pressure, height, depth, dimensionless hybrid sigma-pressure levels), and calendar time.
- Horizontal Grid Mappings: Parameter definitions for projected coordinate reference systems, including Transverse Mercator, Lambert Conformal, Polar Stereographic, and Rotated Pole grids.
- Statistical Cell Methods: Syntax (
cell_methods) that specifies whether values represent instantaneous samplings, spatial averages, maximums, or climatological temporal means. - Discrete Sampling Geometries (DSG): Standardized structures for in-situ observations that are not regular grids, including time-series stations, vertical soundings/profiles, trajectory paths (such as ocean gliders or research aircraft), and spatial point collections.
How to Search or Use It
Researchers creating, cataloging, or publishing Earth observation datasets implement the CF conventions through a structured four-stage workflow:
- Consult the Active Specification: Review the official CF Conventions specification document on
cfconventions.org(version 1.13) to identify the required data structures for your array dimensions (e.g., gridded model output vs. discrete trajectory features). - Search the Controlled Standard Name Table: Access the online CF Standard Name Browser to identify the exact string corresponding to your physical variables. Standard names follow strict grammatical rules (e.g.,
surface_downwelling_shortwave_flux_in_air) and must not be modified or truncated. - Define Valid UDUNITS Dimensions: Assign physical units that conform to Unidata UDUNITS-2 syntax. Units must be dimensionally equivalent to the canonical units listed in the standard name table (for example, velocity must use units equivalent to
m s-1). - Establish Coordinate Bounds: If your variables represent grid-cell averages or time-period accumulations rather than point measurements, declare boundary variables (via the
boundsattribute, such astime:bounds = "time_bnds") to record the precise start and end coordinates of each cell. - Add Required Global Attributes: Set the mandatory global attribute
:Conventions = "CF-1.13"alongside discovery metadata (title,institution,source, andhistory). - Validate Using Automated CF Linters: Verify your generated NetCDF file using the official command-line CF-Checker (developed by the National Centre for Atmospheric Science / CEDA) or web validation tools to catch syntax errors, missing coordinate bounds, or invalid unit strings prior to repository submission.
- Propose New Vocabulary Terms via GitHub: If your research introduces a physical parameter not currently in the Standard Name Table, open a structured proposal issue in the
cf-convention/vocabulariesGitHub repository for community review and committee moderation.
Strengths and Limits
Strengths
- Universal Cross-Tool Interoperability: CF-compliant NetCDF files open automatically across analysis environments (Python xarray, R, MATLAB, Panoply, Ncview, QGIS, and CDO) with spatial coordinates and maps aligned without manual calibration.
- Machine-Actionable Physical Semantics: By distinguishing between variable names (which can be arbitrary) and controlled
standard_nameattributes, CF enables automated data aggregation across heterogeneous climate models. - Transparent Community Governance: Proposed modifications, standard name additions, and committee deliberations occur openly in public GitHub repositories and documented committee minutes.
- Extensible Discrete Observation Handling: Robust support for non-gridded platforms like autonomous ocean floats, weather stations, and drone transects.
Limits
- Steep Learning Curve for Complex Geometries: Correctly implementing coordinate bounds, rotated poles, and discrete sampling geometries requires substantial metadata authoring expertise.
- Vocabulary Moderation Latency: Proposing and formally approving new terms in the Standard Name Table requires a public comment period, which can introduce delays for emerging sensor types.
- Partial Redundancy with General Discovery Metadata: CF focuses primarily on internal data structure and physical meaning; comprehensive repository discovery often requires pairing CF with complementary schemas like the Attribute Convention for Data Discovery (ACDD).
Alternatives
Depending on data storage formats and research domains, several alternative and complementary metadata standards exist alongside CF:
| Metadata Standard / Schema | Primary Scope | Primary Storage Container | Key Distinguishing Feature |
|---|---|---|---|
| CF Conventions (Version 1.13) | Earth science multidimensional gridded and in-situ data | NetCDF-3, NetCDF-4, HDF5, Zarr | Physical standard names, spatial-temporal coordinate definitions, and cell methods |
| ACDD (Attribute Convention for Data Discovery) | High-level dataset discovery and catalog indexing | NetCDF, HDF | Defines bibliographic metadata (title, summary, keywords, creator_email, license) intended to complement CF |
| ISO 19115 / ISO 19139 | Geospatial catalog records and spatial data infrastructures | XML / JSON catalog records | Broad geographic catalogue standard used by national libraries and spatial portals, rather than an array-level specification |
| Darwin Core (DwC) | Biodiversity, taxon occurrences, and biological sampling | CSV, Darwin Core Archive (DwC-A) | Focused specifically on biological specimen observations, taxa classifications, and sampling events |
| OGC WaterML / GeoTIFF | Point hydrology time series and georeferenced raster imagery | XML / TIFF | Specialized format-specific encodings developed by the Open Geospatial Consortium for web services |
FAQ
What does CF stand for in CF conventions?
CF stands for “Climate and Forecast”. Although originally established for climate simulation models and atmospheric weather forecasts, the conventions now cover physical observations and models across oceanography, geology, hydrology, and remote sensing.
What is the latest version of the CF conventions?
As documented on the official conventions portal (cfconventions.org), the active production release is CF Conventions Version 1.13. Compliant datasets should declare this version in their global attributes (:Conventions = "CF-1.13").
What is the difference between CF and NetCDF?
NetCDF is the underlying binary data format and software API (developed by Unidata) for storing multidimensional arrays. CF is the metadata convention layer built on top of NetCDF that defines what the arrays mean, how coordinates map to the Earth, and which units are valid.
What is a CF Standard Name?
A CF standard name is a unique, controlled string that identifies a specific physical quantity (for example, sea_surface_temperature or precipitation_flux). Standard names eliminate ambiguity caused by idiosyncratic local variable abbreviations.
Sources
- CF Metadata Conventions: Official Specification (Version 1.13). Published by CF Conventions Committee. URL: https://cfconventions.org/Data/cf-conventions/cf-conventions-1.13/cf-conventions.html (Read: 2026-11-08).
- CF Conventions Governance: Constitution, Governance Panel, and Committee Membership. URL: https://cfconventions.org/governance.html (Read: 2026-11-08).
- CF Conventions: Rules for Changes to CF Standard Names, Area Types and Standardized Region List Vocabularies. URL: https://cfconventions.org/standard_name_rules.html (Read: 2026-11-08).
