rohanfosse/gbfs-audit-catalogue

GBFS Audit Catalogue

A reproducible audit of 1,509 GBFS bike-sharing feeds worldwide, with the resulting 46-column reference dataset for 46,307 certified stations across 123 French operators.

TL;DR

from datasets import load_dataset
gs = load_dataset("rohanfosse/gbfs-audit-catalogue", split="train").to_pandas()

# High-confidence dock-based stations
clean = gs[(gs.station_type == "docked_bike") & (gs.audit_confidence == "high")]
print(len(clean))  # 5,402

# Per-operator anomaly rates
gs.groupby("operator_name").agg(
    n=("uid", "size"),
    A3_rate=("flag_A3", "mean"),
    A7_rate=("flag_A7", "mean"),
).sort_values("n", ascending=False).head(10)

What is this?

The General Bikeshare Feed Specification (GBFS) is the open standard that French bike-sharing operators must publish on transport.data.gouv.fr under the 2019 Mobility Orientation Law (LOM). The standard guarantees syntactic interoperability but not semantic consistency: identical fields carry mutually incompatible meanings across operators.

This dataset is the output of a systematic audit of the 1,509 GBFS systems catalogued by MobilityData worldwide. Seven recurring anomaly classes (A1–A7) are detected at the row level; 30.9% of the raw French stations are reclassified, removed, or relabelled. The remaining 46,307 stations are released here with per-row anomaly flags and contextual enrichment so that researchers can reuse the audited data without rerunning the pipeline.

Schema (46 columns)

Group	Columns	Source
Identifiers	uid, station_id, system_id, system_name, source	GBFS
Spatial	lat, lon, city, commune_name, code_commune, region_id	GBFS + INSEE COG
Station description	station_name, address, capacity, n_stations_system	GBFS
Audit pipeline (11)	station_type, capacity_raw, capacity_audited, flag_A1–flag_A7, operator_name, audit_confidence, fetched_at	This work
Network geometry (5)	dist_to_nearest_station_m, n_stations_within_500m, n_stations_within_1km, nearest_system_dist_m, catchment_density_per_km2	KNN on this work
Topography	elevation_m, topography_roughness_index	IGN BD ALTI
Cycling infrastructure	infra_cyclable_km, infra_cyclable_pct	BD TOPO 300 m buffer
Safety	baac_accidents_cyclistes	ONISR BAAC 500 m, 5 yr
Multimodal access	gtfs_heavy_stops_300m, gtfs_stops_within_300m_pct	National GTFS aggregation
Socio-economy	revenu_median_uc, gini_revenu, revenu_d1, ecart_interquar, part_menages_voit0	INSEE Filosofi
Modal share	part_velo_travail	INSEE Recensement

The seven anomaly classes

Class	Name	Signature	FR systems	Global systems
A1	Out-of-domain inclusion	car-sharing advertised as BSS	14	46
A2	Placeholder capacity	constant non-zero c across stations	3	48
A3	Structural over-capacity	conditional averaging on free-floating	8	33
A4	Geospatial error	transposed coords or >3σ outliers	3.8% stations	81
A5	Out-of-perimeter	system area >50,000 km² or overseas	5	17
A6	Zero-capacity dock	≥1% stations with c=0	0	14
A7	Null capacity field	≥50% stations with c=NaN	19 (FF)	215

Provenance

• Raw source: 142 candidate GBFS feeds from transport.data.gouv.fr (FR national portal) + the MobilityData canonical catalogue (1,509 systems globally).
• Audit code: Python, MIT licence, github.com/rohanfosse/bikeshare-data-explorer.
• Reference Zenodo deposit: 10.5281/zenodo.20125460 (concept DOI, current release 1.0.0).

Citation

If you use the catalogue in your research, please cite both the paper and the Zenodo deposit:

@article{Fosse2026gbfs,
  author  = {Foss\'e, Rohan and Pallares, Ga\"el},
  title   = {Auditing {GBFS} bike-sharing feeds at country and global scale:
             A reproducible anomaly taxonomy for open mobility data},
  journal = {Computer Standards \& Interfaces},
  year    = {2026},
  note    = {Under review}
}

@dataset{Fosse2026gbfsdata,
  author    = {Foss\'e, Rohan and Pallares, Ga\"el},
  title     = {{GBFS Audit Catalogue} v1.0},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.20125460}
}

Licence

• Data: Open Data Commons Open Database License (ODbL) v1.0 — share-alike, attribution required.
• Code: MIT.

Issues, contributions, contact

GitHub issues for bugs, schema requests or new audit classes: https://github.com/rohanfosse/bikeshare-data-explorer/issues

Lead contact: Rohan Fossé (rfosse@cesi.fr), CESI LINEACT, Montpellier, France.