Skip to content

Citing this work

If you use gbfs-toolkit in a publication, please cite both the software and the accompanying dataset paper. Citing software explicitly supports reproducibility and gives credit to the research-software effort behind the toolkit.

The machine-readable metadata is maintained in CITATION.cff, which GitHub renders as a "Cite this repository" panel and which reference managers can import directly.

Software

@software{fosse_gbfs_toolkit_2026,
  author    = {Fossé, Rohan and Pallares, Gaël},
  title     = {{gbfs-toolkit: research-grade ingestion and semantic
               quality audit for GBFS feeds}},
  year      = {2026},
  version   = {1.4.0},
  doi       = {10.5281/zenodo.20992153},
  license   = {MIT},
  url       = {https://github.com/cycling-data-lab/gbfs-toolkit},
  note      = {ORCID: 0009-0002-2195-0198, 0009-0002-8680-604X}
}

Cite the exact version you used. The canonical schema and public API are frozen under semantic versioning, so reporting the version is sufficient to identify the behaviour your study relied on.

Dataset paper

The A1–A7 semantic taxonomy implemented here comes from the gbfs-audit-catalogue dataset paper.

@article{fosse_gbfs_catalogue_2026,
  author  = {Fossé, Rohan and Pallares, Gaël},
  title   = {{A certified, anomaly-flagged reference catalogue for
             GBFS bike-sharing feeds}},
  year    = {2026},
  note    = {gbfs-audit-catalogue dataset paper, in preparation}
}

The validation data behind the taxonomy live in the companion repository gbfs-audit-catalogue, so the audit rules rest on a verifiable corpus rather than on assertion.

Authors

Author ORCID Affiliation
Rohan Fossé 0009-0002-2195-0198 CESI Engineering School, Montpellier, France
Gaël Pallares 0009-0002-8680-604X CESI LINEACT (EA 7527), Montpellier, France

Archival DOI

Each tagged GitHub release is archived on Zenodo with a citable DOI, minted automatically from the .zenodo.json metadata (authorship, ORCID identifiers, licence).

Cite the concept DOI for the software in general, or the version DOI when exact reproducibility requires the precise archived release.

Provenance for derived datasets

When a study deposits a collected dataset, record its provenance alongside the deposit so reviewers can verify that the data were frozen as described:

import gbfs_toolkit as gb

gb.generate_manifest("lake/")          # SHA-256 per Parquet partition plus a system/date summary
gb.coverage_report(panel, expected_freq="5min")   # per-station uptime and longest gap, no imputation

The toolkit never imputes missing data. Gaps stay NaN and are quantified rather than smoothed, so the reported coverage reflects what was actually observed.