National Wilding Conifer Control Programme

AI Drone Surveys for Wilding Pine Detection & Compliance Reporting

Rapid aerial surveys using computer vision to map wilding conifer spread — with generated compliance reports.

Request a Survey Quote See the Report

Wilding Conifers Are Spreading

Self-seeding pines, larches, and firs are spreading across New Zealand's high country, threatening biodiversity, water catchments, and pastoral land. Early detection is the most cost-effective form of control.

1.8M+

hectares affected by wilding conifers across New Zealand

$110M

committed by the National Programme over 10 years

5×

cheaper to control a tree at the scattered stage vs. dense infestation

Hours

to survey and map a property — not weeks of manual ground search

How It Works

Three steps from flight to council submission.

1

Autonomous Survey Flight

Our drone flies a pre-programmed grid at 100 m AGL with the gimbal pointed nadir (straight down). GStreamer feeds live 1080p footage to the onboard AI at 5 frames per second.

2

AI Detection & Deduplication

A fine-tuned AI model identifies wilding conifers by species in every frame. A 3D ray-cast engine projects each pixel detection to ground GPS. Our Detection Register then deduplicates across frames by ground proximity, producing a clean count of distinct trees.

3

Compliance Report Generation

After landing, a post-flight batch step reads the accumulated detections and generates a council-facing PDF and GeoJSON export. Density classes, control method recommendations, species breakdown, and cost estimates are pre-populated — aligned to WCIS recording conventions so the output maps directly onto programme data fields.

What You Receive

Every survey produces a structured dataset and a ready-to-submit PDF report. No manual data entry. No transcription errors.

📄

Council Compliance PDF

  • Site metadata and survey conditions
  • Total distinct trees with GPS coordinates
  • Species breakdown and confidence scores
  • Density class per 25 m grid cell
  • Recommended control method per density class
  • Indicative cost weighting for prioritisation
  • Pre-control and post-control report types
🗺️

GeoJSON Detection Map

  • Point feature per distinct tree
  • Latitude / longitude centroid (running-mean across all observations)
  • Crown diameter estimate in metres
  • n_observations — frames the tree appeared in
  • Density class and first / last seen timestamps
  • Importable directly into QGIS, ArcGIS, or Google Earth
📊

Summary Statistics

  • Total distinct trees and total raw observations
  • Breakdown by density class (scattered → very dense)
  • Breakdown by species
  • Density choropleth map embedded in PDF
  • Raw detection CSV for your own analysis
🖼️

Thumbnail Archive

  • Cropped frame image saved for each new detection
  • Labelled with entry ID for cross-reference
  • Supports dispute resolution and contractor briefing
  • Delivered as a compressed archive alongside the report

Sample Report Extract

Real output from a survey flight over a 64 ha Mackenzie Basin block. 120 distinct wilding conifers identified from 1,668 raw frame detections.

Dominion Aero — Compliance Report
Wilding Conifer Survey — Site Assessment
Pre-control infestation survey
SCATTERED — 120 trees

1. Operational Details

Site / location
Example Station — North Block
Management unit
Mackenzie Basin MU-12
Region
Canterbury
Aircraft / sensor
Quad VTOL + RGB gimbal  ·  RGB 20MP, nadir
Flight altitude
100 m AGL
Area surveyed
64 ha

2. Infestation Summary

120
distinct trees
1,668
raw frame detections
64 ha
area surveyed
Recommended control: Ground / manual — hand-pull or cut-and-stump; spot follow-up

3. Species Composition

Pinus contorta
50  42%
Larix decidua
41  34%
Pseudotsuga menziesii
29  24%

Species attribution from aerial imagery is indicative; low-confidence detections should be ground-truthed. Pinus contorta is the highest spread-risk species and should be prioritised where present.

4. Detection Register (extract — first 10 of 120)

ID Lat Lon Species Conf Crown m
1-44.1904170.1405Larix decidua0.953.3
2-44.1934170.1367Pseudotsuga menziesii0.903.4
3-44.1907170.1423Pinus contorta0.953.4
4-44.1915170.1423Pinus contorta0.943.5
5-44.1869170.1361Pinus contorta0.943.5
6-44.1866170.1382Pinus contorta0.893.0
7-44.1938170.1421Pseudotsuga menziesii0.912.8
8-44.1910170.1368Larix decidua0.923.4
9-44.1904170.1376Pinus contorta0.922.9
10-44.1918170.1420Pseudotsuga menziesii0.933.0

Showing 10 of 120 records. Full register delivered as detection_register.csv and detections.geojson.

5. Methodology & Data Quality

  • ›Each tree is geolocated by projecting its image position through gimbal and aircraft attitude to a ground coordinate, then de-duplicated across frames by proximity.
  • ›Altitude is measured relative to the takeoff point. Over sloping terrain this differs from true height-above-ground; figures should be treated as survey-grade indicative, not cadastral-survey accurate.
  • ›Stem counts reflect canopy visible from above; seedlings beneath larger crowns or below the detection size threshold may be undercounted.

Want to see a full report for your site? Request a survey quote ›

Density Classification

Each 25 m grid cell is classified by stems per hectare. Control method recommendations are pre-populated in the report.

Class Stems / ha Recommended Control
SCATTERED 0 – 50 Ground / manual — hand-pull or cut-and-stump; spot follow-up
SPARSE 50 – 200 Ground / manual — cut-and-stump or basal-bark herbicide
MODERATE 200 – 1,000 Ground — manual + mechanical mix; drill-and-fill on larger stems
DENSE 1,000 – 4,000 Mechanical clearance; basal-bark / drill-and-fill herbicide
VERY DENSE > 4,000 Mechanical or aerial control where ground access is limited

Density cutoffs are calibrated to NWCCP guidance and adjusted per species and terrain on request.

Why Dominion Aero

Built in New Zealand for New Zealand's land management challenges.

🇳🇿

New Zealand Owned

Based in Wellington. We know the terrain, the programme, and the councils involved.

🤖

Purpose-Built AI

Species-trained computer vision. Not a generic YOLO model — tuned for wilding conifer detection from nadir aerial imagery.

📋

WCIS-Aligned Output

Report fields map directly onto Wilding Conifer Information System conventions. Less reformatting for your team.

⚡

Survey in Hours

What takes a ground crew days to walk can be surveyed, detected, and reported in a single flight day.

🔁

Pre & Post-Control

Fly before control to scope the work. Fly again after to document results and meet programme reporting requirements.

📍

GPS-Accurate Mapping

3D ray-cast geolocation projects detections to ground coordinates. Every tree in the report has a verifiable GPS pin.

Request a Survey

Tell us about your site and we'll come back with a quote and timeline.

Who We Work With

  • Regional and district councils
  • Department of Conservation
  • High country pastoral stations
  • Private landowners with notification obligations
  • Contractors managing control programmes
  • Environmental consultants preparing management plans

Based in: Wellington, New Zealand

Operating area: New Zealand nationwide

Email: operations@dominionaero.co.nz

Phone: +64 21 428 799