US National Transit Database safety

Use scripts/cli.py for two United States Federal Transit Administration resources on data.transportation.gov. These are not Auckland Transport feeds. The catalogue scope tag that says Auckland does not describe this publisher.

Read references/source-notes.md for licences, withheld fields, units and lag.

python3 skills/us-ntd-safety/scripts/cli.py events --limit 20 --json
python3 skills/us-ntd-safety/scripts/cli.py events --limit 20 --year 2024 --mode MB --ntd-id 00000 --json
python3 skills/us-ntd-safety/scripts/cli.py modes --limit 20 --year 2024 --month January --json

Commands

• events [--limit N] [--year YYYY] [--mode CODE] [--ntd-id DDDDD] reads SODA resource 9ivb-8ae9 (Major Safety and Security Events). Output is coded fields and counts only.
• modes [--limit N] [--year YYYY] [--month Month] [--mode CODE] [--ntd-id DDDDD] reads SODA resource 5ti2-5uiv (Monthly Modal Time Series). Output is monthly agency-mode aggregates.

--limit is an integer from 1 to 100. The default is 20. --year is 1900–2099. --mode is a 1–4 letter NTD code such as MB or HR. --month is a full English month name such as January and applies only to modes. Bad arguments fail before any network call.

Each command checks that resource's view metadata before reading rows. events continues only when licenseId is USGOV_WORKS. modes continues only when licenseId is PUBLIC_DOMAIN. A missing or different id is refused. The other resource's id does not satisfy the check.

Output

--json returns meta and results. meta includes source_url, publisher, retrieved_at, the verified licence, geography (United States), stated units, and latest_data when the view has rowsUpdatedAt. That timestamp is the Socrata update instant in UTC. It is not the newest event date and not an Auckland date.

incident_date, incident_time and month_year are copied as source strings. They have no timezone. Do not relabel them as UTC or Auckland time.

Absent measures are omitted, except an explicit JSON null, which stays null. Do not treat either as zero. A non-numeric token in a measure stays a string and is listed in unparsed_measures.

The row set never includes incident narratives, person lists, approximate addresses, coordinates, free-text comments, or other-description fields. Use --json when you need the full coded set. Text output is a short summary.