# Land structure analysis

Last updated: 2026-10-18

> Break a property into six kinds of land in 2020, from primary forest to other land, and outline the other land worth a field visit for reforestation. Uses Nature Trace Forest Typology at 10 m.

What kind of tree cover is on this land: primary forest, regrowth, plantations, tree crops, or none? The answer shapes what a reforestation project can do there. The analysis splits the property into six classes for 2020 and outlines the other land the model is most confident about, as candidates for a field visit.

## Who it is for

- **Carbon project developers.** Separate open land from forest, plantations and tree crops early, so ARR planning starts from land that may qualify.
- **Project and land teams.** Get a first map of where planting could go before a site visit, with every area traced to its dataset.
- **Restoration organizations and funders.** Describe a site’s starting point by kind of tree cover, not only as trees or no trees.

## What goes in

A property boundary as a polygon, and these settings:

| Setting | Default | Range |
| --- | --- | --- |
| Minimum uncalibrated Other Land model probability for candidate screening. | 0.6 | 0.5–1 |

## Data sources

- [Nature Trace Forest Typology (ForTy) 2020 v1.0](https://developers.google.com/earth-engine/datasets/catalog/projects_nature-trace_assets_forest_typology_forest_typology_2020_v1_0_collection), by Google DeepMind, Google Research. Spatial resolution: 10 m. Temporal resolution: One snapshot, 2020. Coverage: Land from 65°S to 84°N. License: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). Model probabilities for five kinds of tree cover; other land is what remains. Cite: [Neumann et al. 2026, preprint](https://doi.org/10.31223/X58R27).

## What you get

- Hectares and share of the property’s observed area in each class: primary forest, naturally regenerating forest, planted forest, plantation forest, tree crops and agroforestry, and other land.
- Other land where the model’s score passes your threshold (0.6 by default), in hectares and as outlines for field review.
- How much of the property had valid data.

### Map layers

- **Forest typology 2020.** Nature Trace forest types in 2020: primary, naturally regenerating, planted, plantation, tree crops and other land. Land may have changed since.
- **Reforestation candidate screening.** Other land pixels where the model is sure enough to list them as reforestation candidates.
- **Candidate Other Land areas.** Candidate other land outlined and cut to the property boundary, for field review.

## Limits

- The data shows 2020. Land cleared, planted or burned since then will not show; check recent imagery before deciding.
- The scores are uncalibrated model outputs: 0.6 is a screening cut-off, not a 60% chance.
- Other land is not the same as plantable land. Wetlands, crops, settlements and protected areas still need ruling out.
- It points to places worth a field visit. It does not decide eligibility.

## Where it fits

Run it at pre-feasibility, beside forest cover history. It narrows a property to the open land worth walking; the field visit, recent imagery and your standard’s eligibility rules make the call.

## Related data layers

- Nature Trace Forest Typology 2020: Global 10 m 2020 forest typology shown as the most likely class: primary, naturally regenerating, planted, plantation, tree crops and agroforestry, or other land.
- ESA WorldCover: ESA’s 2021 global 10 m land-cover map, classifying tree cover, shrubland, grassland, cropland, built-up areas, wetlands, water, and other surfaces.
- WDPA Protected Areas: Protected Planet’s global inventory of terrestrial and marine protected-area boundaries reported by governments and other authoritative data providers.

## Tools and guides

- [Carbon standards for nature projects compared](https://restorationscope.com/blog/carbon-standards-for-nature-projects)
- [Software for ARR carbon projects](https://restorationscope.com/blog/arr-carbon-monitoring-software)
- [Geolocation file checker for Verra projects](https://restorationscope.com/tools/verra-geolocation-file-checker)

## Questions teams ask

### How is this different from forest cover analysis?

Forest cover analysis tells you how much tree cover there was in 2000 and where it was lost through 2025. Land structure tells you what kind of tree cover was there in 2020, so natural forest, plantations and tree crops are told apart. Carbon teams usually run both.

### Can I use the outlined other land as my project area?

Not as it is. The outlines are places to check. Confirm them with recent imagery and a field visit, rule out land your standard excludes, and draw the project area from what holds up.

### Why does the resolution matter?

The dataset’s pixels are 10 m across, so field edges, strips along rivers and small woodlots show up that 30 m data would blur.

### What do I need to run it?

A property boundary as a polygon, drawn on the map or imported from a KML, shapefile or GeoJSON file. The result is saved on the property, beside its other records and maps.

More: [All analyses](https://restorationscope.com/analyses). Book a call: https://cal.com/bernat/restoration-scope-initial-call