Skip to content

Manaaki Whenua / LRIS data via eolas

Manaaki Whenua – Landcare Research is New Zealand's primary CRI for terrestrial ecosystem science. It runs the Land Resource Information Systems (LRIS) portal, which hosts the canonical national datasets on land cover, land use, soils, and protected areas. eolas serves 20 datasets from LRIS — the foundational layers behind every "what's on the ground" analysis in NZ.

If you're doing environmental research, agricultural / forestry / horticultural analytics, climate-policy modelling, or ecological-impact studies — LRIS is the source.

What's in the catalogue

Land Cover Database (LCDB)

The LCDB is the canonical national land-cover classification — what type of vegetation, surface, or use covers each ~10m² of NZ. It's been re-derived from satellite imagery at successive vintages, so you can compare class transitions over time. Six versions are in eolas:

Dataset Description
lcdb_v6_mainland LCDB v6 polygon land-cover for mainland NZ (latest).
lcdb_v6_chatham LCDB v6 for the Chatham Islands.
lcdb_v6_version_trace_mainland Full change history of every polygon across all LCDB versions (mainland).
lcdb_v6_version_trace_chatham Same trace table for Chatham.
lcdb_v5_mainland, lcdb_v5_chatham, lcdb_v5_version_trace_* LCDB v5.
lcdb_v41_mainland, lcdb_v41_chatham, lcdb_v41_version_trace_mainland LCDB v4.1.
lcdb_v40_mainland, lcdb_v40_change LCDB v4.0 + change table.
lcdb_v33_mainland, lcdb_v33_change LCDB v3.3.
lcdb_v30_mainland, lcdb_v30_change LCDB v3.0.

For a single point-in-time analysis, use the latest (lcdb_v6_mainland). For change-over-time analysis, the version-trace tables give you a polygon-by-polygon transition record without having to join multiple versions manually.

Land use + protected areas

Dataset Description
nzlum_v03 NZ Land Use Map v0.3 — polygon land-use classification (~170k features). Distinct from LCDB: classifies use (dairy, exotic forestry, urban) rather than cover (grassland, native bush, built).
pan_nz_2025_draft Protected Areas Network of NZ 2025 (~60k features). All formally-protected land — DOC + QEII covenants + iwi-protected + ngā whenua rāhui. The comprehensive protection layer.
pbc_h3_r9 Pastoral beef + sheep farming intensity modelled on the H3 geospatial indexing system at resolution 9.

Refresh schedule

Weekly, Wednesday morning NZ time.

LRIS datasets are research outputs — they refresh on the LRIS portal's publication cadence (sometimes annually, sometimes longer between major versions). Our weekly refresh catches any changes promptly.

meta = client.info("lcdb_v6_mainland")
meta["last_refreshed_at"]
meta["source_last_modified_at"]

License

LRIS data is published under CC-BY 4.0 in most cases — but always check the per-dataset metadata. A few research outputs have stricter or attribution-only licences.

Recommended attribution: "Source: Manaaki Whenua – Landcare Research / LRIS Portal, served via eolas (eolas.fyi). CC-BY 4.0."

If you cite LCDB in academic work, Manaaki Whenua publishes a recommended citation per version on the LRIS dataset page.

Common patterns

Land-cover composition of a region

import geopandas as gpd

lcdb = client.lris("lcdb_v6_mainland", as_sf=True)
# Filter to one region (clip against LINZ regional council boundaries)
rc = client.linz("nz_regional_council_2023", as_sf=True)
canterbury = rc[rc["regc2023_v1_00_name"] == "Canterbury"]
cant_cover = gpd.clip(lcdb, canterbury)

# Area by cover class
cant_cover["area_ha"] = cant_cover.to_crs("EPSG:2193").area / 10_000
composition = cant_cover.groupby("class_name")["area_ha"].sum().sort_values(ascending=False)
print(composition.head(15))

library(sf)
library(dplyr)

lcdb <- eolas_get_lris("lcdb_v6_mainland", as_sf = TRUE)
rc   <- eolas_get_linz("nz_regional_council_2023", as_sf = TRUE)
cant <- rc[rc$regc2023_v1_00_name == "Canterbury", ]
cant_cover <- st_intersection(lcdb, cant)

cant_cover$area_ha <- as.numeric(st_area(st_transform(cant_cover, 2193))) / 10000
composition <- cant_cover |>
  st_drop_geometry() |>
  group_by(class_name) |>
  summarise(area_ha = sum(area_ha)) |>
  arrange(desc(area_ha))
print(head(composition, 15))

Land-cover change 2008 vs 2018 (v3.3 → v5)

# Use the version-trace table — already has class_2008 + class_2018 on each polygon
trace = client.lris("lcdb_v5_version_trace_mainland")
# Polygons that flipped from "Indigenous Forest" to "Exotic Grassland" (i.e. cleared)
cleared = trace[(trace["class_2008"] == "Indigenous Forest") & (trace["class_2018"] == "Exotic Grassland")]
print(f"Indigenous forest cleared to grassland 2008-2018: {cleared['area_ha'].sum():.0f} ha")

What's protected near me?

import geopandas as gpd
from shapely.geometry import Point

pan = client.lris("pan_nz_2025_draft", as_sf=True)
# 50km buffer around a point (in NZTM)
pt = gpd.GeoSeries([Point(174.7762, -41.2865)], crs="EPSG:4326").to_crs("EPSG:2193")
buffer = pt.buffer(50_000).iloc[0]
nearby = pan[pan.to_crs("EPSG:2193").intersects(buffer)]
print(f"Protected areas within 50km of Wellington: {len(nearby)} polygons, {nearby['area_ha'].sum():.0f} ha")

Pastoral intensity heatmap

import matplotlib.pyplot as plt

pbc = client.lris("pbc_h3_r9", as_sf=True)
# pbc has H3 hex polygons + a modelled intensity score
pbc.plot(column="intensity", cmap="RdYlGn_r", figsize=(10, 12), legend=True)
plt.title("Pastoral beef + sheep intensity (H3 R9)")

Source-specific notes

  • LCDB versions are vintaged, not "updated": each LCDB version was independently re-derived from satellite imagery at a specific epoch (v3.0 ≈ 2001/02, v3.3 ≈ 2001/02 reprocessed, v4.0 ≈ 2008, v4.1 ≈ 2008 reprocessed, v5 ≈ 2012/13, v6 ≈ 2018). Use the version-trace tables for proper change analysis — don't just join two raw versions on geometry, the polygon boundaries themselves shift.
  • LCDB classes (33 total): see the LCDB documentation — top-level categories are Artificial, Bare/Lightly-vegetated, Water, Cropland, Grassland, Scrub, Forest. Each splits into specific classes (e.g. "Indigenous Forest", "Exotic Forest", "Manuka and/or Kanuka").
  • Cover vs use: LCDB tells you what's growing/present; NZLUM tells you what it's used for. Forestry land in LCDB might be "Exotic Forest" cover; in NZLUM it'd be "Forestry — exotic — production". Use the right one for the question you're answering.
  • PAN includes private protection: the Protected Areas Network is broader than just DOC's doc_public_conservation_land. It includes QEII covenants on private land, ngā whenua rāhui on Māori land, and other formal protections. For "what's protected somewhere" use PAN; for "what does DOC manage" use the DOC layer.
  • Geometry: all LRIS datasets expose geometry_wkt. The polygon datasets (LCDB, NZLUM, PAN) can be large — lcdb_v6_mainland is ~430k polygons. Free tier 50,000-row cap will truncate; either upgrade or filter with a region clip first.
  • CRS: LRIS publishes in NZTM2000 (EPSG:2193). The eolas pipeline reprojects to WGS84 (EPSG:4326) for the wire format. For area / distance calculations, reproject back to NZTM (.to_crs("EPSG:2193")).

Where to find more