Land Eligibility Methods

Overview

Land eligibility assessment determines which areas within a project boundary qualify for ARR carbon crediting under Verra VM0047. An area is considered eligible if it has been in a degraded or non-forest state for a minimum of ten years prior to the project start date to prevent perverse incentives, is not located on organic soils or wetlands, and is suitable for native ecosystem restoration.

LandOS provides three distinct land eligibility computations, each suited to different data environments and project stages, as well as an automated selector that routes to the appropriate method based on geographic coverage. Access to different land eligibility approaches varies by subscription tier.

---

Land Eligibility (Terra Vista)

Computation type: land_eligibility_tv | Tier: Free, Pro, Enterprise

Data Sources

Land Eligibility (Terra Vista) applies two approaches depending on geographic coverage:

Approach 1 — MapBiomas Collection (where available)

Where MapBiomas data is available, the computation uses annual land-use/land-cover (LULC) maps from the MapBiomas Collection (1985–present). Two LULC maps are selected: one approximately ten years before the project start date and one in the most recent available year. MapBiomas classification accuracy ranges from 81.8% to 96.6% at Level 1 classification across biomes.

Original MapBiomas classes are reclassified into four eligibility categories:

Approach 2 — Sentinel-2 Supervised Classification (outside MapBiomas coverage)

For regions outside MapBiomas coverage, two supervised classifications are performed using Copernicus Sentinel-2 MSI Level-2A imagery and SRTM v4 digital elevation model data. Both classifications span the same ten-year window as Approach 1.

Technical specifications:

• Surface-reflectance composite at 10-meter resolution
• Random Forest classifier with 100 trees
• Input bands: B2, B3, B4, B5, B6, B7, B8A, B11, B12
• Cloud masking via Sentinel-2 QA60 and SCL bands
• Spectral indices: NDVI, NDII, NDWI, EVI
• Ancillary inputs: canopy height models, global LULC datasets, ground truth sample points

Eligibility Assessment

In both approaches, eligible areas are defined as pixels classified as non-forest converted from forest into productive use (cropland, pastureland) or degraded forest in the historical map, and remaining in an eligible class in the current map. Areas with greater than 50% wetland coverage are assigned zero eligible area.

All calculations are performed in Google Earth Engine.

Geographic Coverage

Accuracy boundaries are defined in two tiers. High-accuracy coverage corresponds to regions with dense MapBiomas data and validated training datasets. Low-accuracy coverage extends to additional regions where Sentinel-2 classification is applied. Polygons outside both boundaries are rejected.

---

Land Eligibility (Forest Loop)

Computation type: land_eligibility_fl | Tier: Enterprise

Methodology

Land Eligibility (Forest Loop) performs two independent Forest Loop land cover classifications internally — one for a selected reference year and one for ten years prior — and derives eligible area from the temporal comparison. This approach does not rely on external LULC datasets; all classification is driven by the same Random Forest model used in the Land Cover (Single Year) computation.

The end year is specified by the user; the start year is calculated as end_year − 9. Both classifications are run using the same sample points retrieved from the platform database and the same satellite composite (Landsat 8 by default).

Classification and Comparison

Each classification produces a land cover map with the following classes: Forest, Water, Built, Degraded, Savannah, Clouds, Degraded Forest, Bare, Grassland, Shrubland.

Each pixel is then remapped to an eligibility state and classified according to the following rules:

Areas classified as forest in the start year but not in the end year are flagged as deforested and visualized separately.

Outputs

The computation returns:

• Land Eligibility Executive Summary: structured JSON with eligible area (ha), per-class statistics for both years, and analysis metadata
• Land Cover Distribution: pie chart of LULC classes for the end year
• Land Cover Change: line chart of all LULC classes across the ten-year window
• ARR Eligible Land Map: Earth Engine tile layer showing eligible pixels
• Deforestation Map: Earth Engine tile layer showing pixels that transitioned from forest to non-forest
• Land Cover Maps: tile layers for the start and end year classifications

---

Land Eligibility (2 Year)

Computation type: land_eligibility_2y | Tier: Internal

Methodology

Land Eligibility (2 Year) accepts two previously completed Land Cover (Single Year) computation IDs with one representing the historical reference period and one representing the project start period. These images are compared. This computation does not re-run any classification; it loads exported raster assets and performs the comparison directly.

The two input computations must have sequential start dates (start date of the first must precede start date of the second). Both classifications should be reviewed for accuracy before running this computation; the quality of the eligibility output depends entirely on the quality of the input Forest Loop maps.

Comparison Logic

The pixel-level comparison follows the same rules as Land Eligibility (Forest Loop). Each pixel in both input images is remapped from the ten-class land cover scheme to an eligibility state, and the two states are combined:

Users may supply a custom class_mapping to override the default forest/eligible/non-eligible assignments for any land cover class.

Outputs

The computation returns:

• Land Eligibility Executive Summary — structured JSON with eligible area, per-class statistics, satellite metadata for both input computations, and the applied class mapping
• ARR Eligible Land Map — Earth Engine tile layer showing eligible pixels
• Deforestation Map — Earth Engine tile layer showing forest-to-non-forest transitions between the two input years

---

Land Eligibility (Automatic Selector)

Computation type: land_eligibility | Tier: Free

Selection Logic

The Land Eligibility computation automatically selects the most appropriate method based on the spatial coverage of the project polygon. The selection proceeds as follows:

1. The project polygon is evaluated against the Terra Vista high-accuracy boundary.
2. If the polygon falls within the high-accuracy boundary, Land Eligibility (Terra Vista) is executed.
3. If the polygon falls outside the high-accuracy boundary, Land Eligibility (Forest Loop) is executed using default parameters.

This computation is used as the Land Eligibility step in the ARR Project Pipeline, ensuring that users receive the most appropriate assessment without manual method selection.

---

Limitations

All land eligibility computations produce outputs intended for prefeasibility and feasibility-stage analysis. Results should not be treated as final eligible area for carbon crediting without additional field verification and, where applicable, compliance with Verra VM0047 eligibility documentation requirements.