Skip to content

NZTA (Waka Kotahi) data via eolas

Waka Kotahi NZ Transport Agency (NZTA) is the agency responsible for the state highway network, vehicle and driver licensing, road safety, and the EV charging rollout. eolas serves 5 datasets from NZTA — small in count but heavy in research value: crash records since 2000, the entire licence-holder population, and traffic counts at every state highway telemetry site.

If you're doing road-safety research, transport-policy analysis, EV-infrastructure planning, or modelling traffic flows — NZTA is the source.

What's in the catalogue

Dataset Description
nzta_cas_crashes All recorded NZ road crashes since 2000 (~900k records). Includes severity (fatal / serious / minor / non-injury), location, road conditions, vehicle types, intoxication flags. The canonical road-safety dataset.
nzta_driver_licences Count of NZ driver licence holders by region, class (car, motorcycle, heavy combination, etc.), and stage (learner / restricted / full).
nzta_ev_charging All public EV charging stations in NZ — operator, address, connector types, plug count.
nzta_tms_daily_traffic Daily vehicle counts at state highway Telemetry Monitoring Sites since 2018 (~7.3M records), broken down by vehicle weight class. The granular traffic-flow source.
nzta_traffic_monitoring_sites Annual Average Daily Traffic (AADT) at all state highway monitoring sites, with five-year trend snapshots. The summary version of TMS daily traffic.

Refresh schedule

Weekly, Wednesday morning NZ time.

  • CAS crashes update as new investigations close — typically a 1-3 month lag between an incident and its appearance in the published data.
  • Driver licences and EV charging refresh from the NZTA Open Data Portal.
  • TMS daily traffic publishes T+1 (next day) at the source — our weekly refresh catches up to ~5-7 days of fresh data each Wednesday.
meta = client.info("nzta_cas_crashes")
meta["last_refreshed_at"]
meta["source_last_modified_at"]

License

All NZTA data is published under CC-BY 4.0 via the NZTA Open Data Portal. Commercial use is fine; attribution required.

Recommended attribution: "Source: Waka Kotahi NZ Transport Agency, served via eolas (eolas.fyi). CC-BY 4.0."

A privacy note for nzta_cas_crashes: NZTA has already aggregated/redacted personal information (no driver names, no plate numbers). What you get is incident-level data suitable for analysis.

Common patterns

Fatal-crash trend by region

import matplotlib.pyplot as plt

crashes = client.nzta("nzta_cas_crashes", start="2015-01-01")
fatal = crashes[crashes["crash_severity"] == "Fatal Crash"]
by_year = fatal.groupby(["crash_year", "region"])["objectid"].count().unstack()
by_year.plot(figsize=(10, 6), title="Fatal crashes by region")
plt.show()

library(dplyr)
library(ggplot2)

crashes <- eolas_get_nzta("nzta_cas_crashes", start = "2015-01-01")
fatal <- crashes |>
  filter(crash_severity == "Fatal Crash") |>
  count(crash_year, region)

ggplot(fatal, aes(crash_year, n, colour = region)) +
  geom_line() +
  labs(title = "Fatal crashes by region", y = "Count")

State highway traffic over time

tms = client.nzta("nzta_tms_daily_traffic", start="2020-01-01")
# tms columns: site_ref, date, light_volume, heavy_volume, total_volume

# State Highway 1 around Wellington — one site code
sh1_well = tms[tms["site_ref"] == "01N00251"]
sh1_well.plot(x="date", y="total_volume", title="SH1 Ngauranga daily traffic")

EV charging coverage by territorial authority

import geopandas as gpd

ev = client.nzta("nzta_ev_charging", as_sf=True)
# Join to TLA boundaries (LINZ) for coverage analysis
ta = client.linz("nz_territorial_authority_2023", as_sf=True)
ev_by_ta = gpd.sjoin(ev, ta, how="inner", predicate="within")
coverage = ev_by_ta.groupby("ta_name").size().sort_values(ascending=False).head(15)
print(coverage)

Active monitoring sites with rising AADT

mon = client.nzta("nzta_traffic_monitoring_sites")
# Five-year growth >= 20%
growing = mon[mon["aadt_5yr_pct_change"] >= 20].sort_values("aadt_5yr_pct_change", ascending=False)
print(growing.head(20))

Source-specific notes

  • CAS reporting lag: a fatal crash that happens today won't appear in the CAS feed for ~30-90 days while the police investigation closes. For up-to-the-minute road-toll stats, use NZTA's Road Toll dashboard. For longitudinal research, CAS is authoritative.
  • TMS site coverage: ~600 telemetry sites covering the State Highway network only. Council-managed roads aren't in TMS — for those you'd need to ask each council (most don't publish equivalent open data).
  • Light vs heavy split: TMS volumes split at ~3.5t GVM. "Heavy" here means trucks + buses; "light" means cars + light commercials.
  • EV charging is operator-published: NZTA aggregates from operator submissions. Newly-opened stations may take 1-2 weeks to appear. For the absolute latest, PlugShare and operator apps are faster — but NZTA is canonical.
  • Driver licences are stock not flow: this is a snapshot of current licence holders, not the count of licences issued. For new-licence flow you'd need the (smaller) test-pass data, which isn't currently in eolas.
  • Geometry: nzta_cas_crashes, nzta_ev_charging, nzta_tms_daily_traffic, and nzta_traffic_monitoring_sites all have a geometry_wkt column. Use as_sf=True to get a GeoDataFrame in Python.

Where to find more

  • LINZ source guide — for road centrelines + addresses that geocode against TMS sites
  • Examples — boundary mapping recipes