# Eco-index.

URL: https://levystreet.com/portfolio/eco-index/
Updated: 2026-10-19

Eco-index was a New Zealand programme using AI remote detection and open geospatial data to guide native ecosystem reconstruction. Its tools are now open access.

## Short answer

Eco-index was a New Zealand biodiversity programme that built open data products to guide where and how to reconstruct native ecosystems, including two AI-trained models that detect native vegetation from satellite and Google Earth imagery, mapping tools and reconstruction costings. It began under the Biological Heritage National Science Challenge, closed in March 2026, and released all of its GIS layers and code as open access.

- Observer detected mature native forest from Planet satellite imagery.
- A second model detected young regenerating vegetation from Google Earth imagery.
- Navigator and Navigator X located the best places to reconstruct native ecosystems.
- Everything is now open access under CC BY 4.0.

## What Eco-index was

Eco-index set out to answer a practical question for Aotearoa: where should native ecosystems be rebuilt, what would it cost, and what would it be worth? It started as a publicly funded research programme within the Biological Heritage National Science Challenge, co-directed from the University of Waikato's Environmental Research Institute, became a purpose-driven company in 2024 and a charitable company in 2025.

## The tools


- **Observer**: Two AI-trained remote detection models. One finds mature native forest using Planet multispectral satellite imagery with radar and elevation data; a minimum viable model finds young regenerating vegetation on Google Earth imagery.
- **Navigator and Navigator X**: GIS and browser-based mapping tools that locate the best places to reconstruct native ecosystems from scratch.
- **Ecosystem Reconstruction Map**: A national map shading catchments by reconstruction priority.
- **Valuer and costings**: NZ dollar valuations for ecosystem services and reconstruction and maintenance cost layers.


*[Image: The Eco-index Ecosystem Reconstruction Map of New Zealand]*

## Why it is a useful AI example

Eco-index shows computer vision doing real public-good work: turning imagery that no one has time to inspect by hand into maps people can plan with. It also shows the limits honestly. The regeneration model was released as a minimum viable product, with Eco-index noting "further R&D required", and every detection still needed human interpretation before decisions were made.

Those are the same lessons we apply in vision work for clients, from [choosing between OCR and vision models](/engineering/ocr-versus-vision-models/) to [preparing training data](/engineering/ai-data-preparation/): be clear about what the model was trained on, measure where it fails, and keep people in charge of decisions.

*[Image: Navigator X for Waitaha Canterbury]*

## Status

Levy Street worked with the Eco-index team. The programme closed in March 2026 with all layers and code open access under CC BY 4.0. Its website remains live until March 2027 and its code at [github.com/eco-index](https://github.com/eco-index). For AI in farming and land management today, see our guide to [AI for New Zealand agriculture](/industries/agriculture/).

## Questions people ask

### What was Eco-index?

A New Zealand programme, later a charitable company, that produced open data, mapping tools and AI remote detection models to guide biodiversity investment and native ecosystem reconstruction. It began under the Biological Heritage National Science Challenge and was co-directed from the University of Waikato.

### Is Eco-index still running?

No. Eco-index closed in March 2026. Its GIS layers and code were released as open access under CC BY 4.0, its website stays live until March 2027, and its GitHub account remains available.

### How did Eco-index use AI?

It trained two remote detection models: one detecting mature native forest from Planet multispectral satellite imagery, radar and elevation data, and a minimum viable model detecting young regenerating vegetation from Google Earth imagery.


Published by Levy Street (https://levystreet.com), Wellington, New Zealand.