Insights · Land eligibility
ARR land eligibility: what 17 desktop screenings show
Most people planning ARR carbon projects assume the hard part is growing trees. In our screenings, the first and biggest filter was simpler: how much of the land was allowed to count at all. Across 17 real sites in Africa, Latin America and Asia, the eligible share ran from 99.4% to 0.1%. One layer explained most of that spread. Here is the data, anonymised, and what it means for choosing land.
The 17 sites
Between July and November 2026 we screened 17 distinct sites for tree-planting projects: afforestation, reforestation and revegetation (ARR), including agroforestry, following Verra's VM0047 methodology. Together they cover about 1.3 million hectares, from a 130-hectare montane plot to an 845,000-hectare savanna landscape. Where a site was run more than once, for example after the boundary was redrawn, we use the latest run, so each site counts once. The settings are generalised to a region and the areas rounded, so no project or client can be identified. Agroforestry sites are marked as such; the rest are other ARR tree-planting designs.
Each screening applies the same tests to every 30 m pixel inside the boundary: was it forest in 2000, was forest lost in the last ten years, is it built-up, water or wetland, is the slope over 30°, and was native vegetation recently converted. What survives is the eligible area.
| Setting (generalised) | Type | Site (ha) | Eligible | Forest in 2000 |
|---|---|---|---|---|
| West Africa, savanna landscape | ARR | 845,000 | 99.4% | 0% |
| Kenya, dryland highlands | ARR | 1,700 | 94.7% | 0% |
| Kenya, montane agroforestry | Agroforestry | 420 | 92.7% | 7% |
| Colombia, Llanos savanna pasture | ARR | 1,190 | 90.7% | 9% |
| South Africa, Albany thicket | Agroforestry | 180 | 70.3% | 30% |
| Kenya, coffee and tea landscape | Agroforestry | 2,700 | 61.4% | 39% |
| DR Congo, forest-savanna mosaic | ARR | 690 | 58.1% | 38% |
| Malawi, miombo and farmland | ARR | 258,700 | 40.7% | 53% |
| Southern Africa, miombo | ARR | 15,400 | 37.6% | 62% |
| Uganda, Albertine Rift montane | Agroforestry | 4,070 | 16.3% | 82% |
| Tanzania, Eastern Arc montane | ARR | 130 | 10.4% | 89% |
| Côte d’Ivoire, cocoa and forest landscape | Agroforestry | 3,635 | 7.6% | 91% |
| Cameroon, forest-savanna | ARR | 5,570 | 2.0% | 98% |
| Cameroon, lowland rainforest | ARR | 5,040 | 1.7% | 98% |
| Indonesia, Borneo lowland rainforest | ARR | 41,300 | 0.4% | 99.5% |
| Cameroon, coastal mangroves | ARR | 450 | 0.1% | 66%* |
The patterns in ARR carbon projects
- Median eligible share40.7%
- Sites under 11% eligible6 of 17
- Sites with ≤9% forest in 2000All ≥90% eligible
There is no typical site. The median eligible share was 40.7%, but few sites sat near it. Five sites were above 90% and six were under 11%. A planning assumption like “about half the land will qualify” would have been wrong for most of them.
Forest in 2000 decided the result. All six sites under 11% eligible had at least 66% of their area classed as forest in 2000. Every site with 9% forest or less was at least 90% eligible. In between, the eligible share tracked the non-forest share closely: the Kenyan coffee and tea landscape was 39% forest and 61.4% eligible; the Albany thicket site was 30% forest and 70.3% eligible. The other layers usually took only a few points more.
Project type did not rescue forested land. Agroforestry sites followed the same rule. The Côte d'Ivoire cocoa landscape, at 91% forest, ended at 7.6% eligible. Cocoa grown under shade trees often reads as forest from space, which is a real limit of desktop data and one reason that site needs field checks.
Percentages hide hectares. A low share on a big site can still be a project, and a high share on a small one may not be. The Malawi site was only 40.7% eligible but that is about 105,000 hectares. The Tanzanian montane plot was 10.4% eligible, which leaves about 14 hectares. The Uganda site, at 16.3%, kept 664 hectares; Côte d'Ivoire, at 7.6%, kept about 276. Whether those areas are worth developing depends on growth rates, costs and what the landowners want, not on the percentage. Always read the eligible share next to the eligible area.
Some of the low results are good news. The Cameroon rainforest sites and the Borneo site were 98–99.5% forest. Those landscapes need protection or restoration of standing forest, which is a different project type, not ARR. The screening did its job by saying so early.
Which exclusion layers bit where
After existing forest, the other layers mattered in specific settings:
- Built-up land was the biggest non-forest exclusion on large farmed landscapes. On the Malawi site it removed about 9,500 hectares; on the Kenyan dryland site about 90 hectares of homesteads and roads.
- Recent forest loss under the ten-year look-back took about 3,100 hectares in Malawi, and smaller areas in DR Congo (about 25 ha) and Côte d'Ivoire (about 35 ha). These are the pixels that stop a project being rewarded for clearing.
- Slopes over 30° mattered only in hilly country: about 2,300 hectares in Malawi, none at all on the flat Kenyan dryland site.
- Recent conversion of native vegetation removed about 1,800 hectares in Malawi and none on the Kenyan dryland site.
- Water and wetland was decisive on the Cameroon mangrove site, together with forest cover, leaving 0.1% eligible. Elsewhere it was small: about 36 hectares in Uganda, none on the Kenyan dryland site.
Malawi is the one site where every layer mattered at once: existing miombo woodland, villages, recent clearing, steep hills and converted vegetation. It still kept about 105,000 eligible hectares, because the landscape is so large. The detail is on the Malawi country guide, and the other countries are under country guides.
What it means for choosing sites
Three practical lessons came out of these screenings.
- Look at forest history before anything else. If more than about two-thirds of your boundary carried forest in 2000, expect very little eligible land. You can check this yourself on the Global Forest Change map before paying anyone. The rules behind it are in the forest definition and the 10-year rule.
- Draw tight boundaries. A boundary that includes villages, river corridors and forest patches drags the eligible share down and makes the project harder to explain. Map planting zones on the open land you intend to plant.
- High eligibility is not a carbon case. The open savanna and dryland sites passed eligibility easily, but they sit in the slowest IPCC growth zones. The Kenyan dryland site was 94.7% eligible with a Tier 1 rate of 0.775 tCO2e per hectare per year for young stands. Eligibility is the first gate, not the last. See carbon sequestration per hectare for the growth side.
Open savanna and grassland raise their own question: whether trees belong there at all. A treeless hectare is not automatically degraded land. That is an ecological judgement, covered in afforestation vs reforestation. The eligibility screen is the first step of a full carbon feasibility study, and it is sold on its own as the Eligibility Check.
Method notes and limits
Forest in 2000 and forest loss come from Hansen et al. Global Forest Change, which records tree canopy cover in 2000 for vegetation taller than 5 m, and the year of stand-replacing loss, at about 30 m per pixel. Built-up land comes from GHSL, water and wetland from JRC Global Surface Water, slopes from NASADEM. The “forest in 2000” column is the share of pixels that met the forest test. On forested sites it is higher than the average canopy cover: a site can have 46% mean canopy cover and still have most of its pixels above the forest threshold. On open sites it can be lower, because scattered trees raise the average without any pixel counting as forest.
The 17 runs were made over July to November 2026 on different engine versions. One site (DR Congo) ran on v1.5, before the gross-to-net screen existed; the latest runs used v1.18.0. Results from different versions may differ in detail, so read the table as a pattern, not a benchmark.
Frequently asked questions
Across 17 real desktop screenings, the eligible share ranged from 99.4% to 0.1%, with a median of 40.7%. There is no typical figure. The share depends mostly on how much of the site was already forest, which is why we recommend an eligibility check before any carbon modelling.
ARR means establishing trees on land without forest. Land that already meets the forest definition is not eligible, and land where forest was lost in the last ten years is excluded so nobody is rewarded for clearing to replant. In our data, every site under 11% eligible had at least 66% of its area classed as forest in 2000.
On large farmed landscapes it was built-up land: villages, homesteads and roads. On our Malawi site that alone removed about 9,500 hectares. Water and wetland, recent forest loss, slopes over 30 degrees and recent conversion of native vegetation each took smaller areas, though they can decide the result on small sites.
No. It is a screening on global datasets with 30 m pixels. A validation body will check the national forest definition, land history and field conditions. Desktop results are good at showing which sites are clearly in or clearly out; mixed landscapes such as cocoa under shade trees need field checks.
Sources
- Hansen et al. (2013), Global Forest Change: dataset description and layer definitions (tree cover 2000, loss year, 30 m pixels)
- Verra: VM0047 Afforestation, Reforestation and Revegetation, v1.1
- FAO Global Forest Resources Assessment: terms and definitions
- IPCC (2019) 2019 Refinement, Volume 4, Chapter 4 Forest Land (forest definition used in Table 4.9: at least 10% canopy cover)