Skip to content

CMIP7 Global Attributes

Version: 1.0
Published: 6 October 2025
DOI: 10.5281/zenodo.17250297
Authors: Karl E. Taylor, Laurent Troussellier, Sasha Ames, David Hassell, Maria Molina, Zebedee Nicholls, Martin Schupfner, James Anstey, Daniel Ellis, Elisabeth Dingley, Paul J. Durack, Guillaume Levavasseur, Matthew Mizielinski, and Marie-Pierre Moine

1. Executive Summary

This document summarizes the definitions of the global attributes, Data Reference Syntax (DRS), file naming conventions, and directory structure for CMIP7 (Coupled Model Intercomparison Project Phase 7).

It is intended as a convenient reference and overview.
The authoritative and complete specification is available on Zenodo:
➡️ https://doi.org/10.5281/zenodo.17250297

Each CMIP7 model output file includes standardized metadata (global attributes) describing:

  • The source model and experiment conditions
  • Data content and licensing
  • Controlled vocabulary (CV) values ensuring consistency across models

These attributes ensure interoperability, facilitate search and retrieval, and provide traceable metadata for analysis.
The DRS system connects file naming, directory organization, and dataset metadata.

2. Introduction

CMIP7 continues the CMIP framework with updated and expanded metadata definitions.
Each file records a well-defined set of global attributes — some mandatory, some optional — required to:

  • Identify the model and experiment
  • Ensure reproducibility and data discovery
  • Enable automated validation and search

Values for many attributes are governed by CMIP7 Controlled Vocabularies (CVs) available at:
➡️ https://github.com/WCRP-CMIP/CMIP7_CVs/
https://github.com/WCRP-CMIP/WCRP-Universe/
Guidance documents: https://wcrp-cmip.github.io/cmip7-guidance/

3. Data Reference Syntax (DRS) Components

The DRS defines how datasets are uniquely identified and organized. It uses a subset of global attributes to construct:

  • Filenames
  • Directory paths
  • URLs and search facets

Key Global Attributes

Category Attribute Example Description
Identification activity_id CMIP, ScenarioMIP Activity responsible for experiment
experiment_id historical, piControl Short label for experiment
mip_era CMIP7 CMIP phase
source_id CanESM6-MR Model identifier
Versioning data_specs_version MIP-DS7.1.0.0 Version of data specifications
drs_specs MIP-DRS7 Label for DRS version
Simulation Variants realization_index r1 Ensemble members differing only by initial conditions
initialization_index i1 Initialization method/date
physics_index p1 Physics configuration
forcing_index f1 Forcing variant
Structure & Sampling temporal_label tavg, tpt Time sampling type
vertical_label h2m, 200hPa Vertical level sampling
horizontal_label hxy, hs Horizontal sampling
area_label u, lnd, sea Area mask
Other Required Fields frequency, grid_label, region, realm, variable_id, tracking_id Define data frequency, grid, domain, and unique file IDs

Conditionally Required Attributes:
Used mainly when datasets branch from parent simulations (e.g., branch_time_in_parent, parent_activity_id).

Optional Attributes:
Provide additional context (e.g., experiment, institution, license, source, title, variant_info).

4. File Name Template

Each CMIP7 file follows a standardized pattern derived from global attributes: 

<variable_id>_<branding_suffix>_<frequency>_<region>_<grid_label>_<source_id>_<experiment_id>_<variant_label>[_<timeRangeDD>].nc

Example:

tas_tavg-h2m-hxy-u_mon_glb_g13s_CanESM6-MR_historical_r2i1p1f1_190001-190912.nc

Notes:

  • timeRangeDD is omitted for fixed (time-independent) variables.
  • Filenames directly encode key experiment and model metadata.

5. Directory Structure Template

The CMIP7 archive organizes files hierarchically using the following DRS pattern:

<drs_specs>/<mip_era>/<activity_id>/<institution_id>/<source_id>/<experiment_id>/<variant_label>/<region>/<frequency>/variable_d>/<branding_suffix>/<grid_label>/<directoryDateDD>

Example:

MIP-DRS7/CMIP7/CMIP/CCCma/CanESM6-MR/historical/r2i1p1f1/glb/mon/tas/tavg-h2m-hxy-u/g13s/v20250622

datasetEditionDD (e.g., v20250622) marks the dataset creation or revision date.

6. Appendix 1: Global Attributes for Labeling Experiments

  • Branching Attributes:
    Required when a simulation derives from a parent run:
  • branch_time_in_child
  • branch_time_in_parent

  • parent_time_units

  • Parent Identifiers:
    parent_activity_id, parent_experiment_id, parent_source_id, parent_variant_label, etc.

These ensure reproducibility and traceability between related simulations.

7. Appendix 2: Algorithms for Defining the “nominal_resolution” Attribute

The nominal_resolution describes approximate horizontal grid spacing (e.g., "1 km", "250 km").
It is computed using a standard algorithm (updated from CMIP6 Appendix 2) that ensures consistent comparison of model resolutions across CMIP7 datasets.