Insights · Gross to net
Leakage in carbon credits: what displaced farming costs a tree-planting project
Leakage in carbon credits is the emissions your project causes somewhere else. For tree planting it usually means the cattle, crops or fuelwood collection that used the land before. In the screenings we publish it took between 56% and 79% of gross removals on farmed land, and close to nothing where farming carried on under the trees. Here is how VMD0054 estimates it, and what you can do about it.
What leakage in carbon credits means for tree planting
A tree-planting project takes land out of its old use. If that land fed cattle or grew maize, the demand for beef and maize does not disappear. Some of it is met by farmers elsewhere working harder on land they already have. Some of it is met by clearing new land. That clearing releases carbon, and it would not have happened without your project. That is leakage.
Verra's leakage module for tree planting, VMD0054, separates two routes:
- Activity shifting. The people who used the land before, the “baseline agent” in the module's words, move their herd or their fields to another place.
- Market leakage. Other producers, anywhere in the country, grow more because supply has dropped. Some of that growth comes from newly cleared land.
The module covers displaced agriculture, including grazing, and displaced fuelwood collection. It counts emissions in the host country only. The point is simple: a buyer of your credits wants to know that the carbon your trees store is not cancelled out by a forest cleared down the road.
- ModuleVMD0054 (v1.1 since 13 Jan 2026)
- Used byVM0047 area-based approach
- In our published farmed-land screenings56–79% of gross
How VMD0054 estimates leakage
VMD0054 does not ask how fast your trees grow. It asks how much production you take away and what it costs in carbon to replace it. The module works in five steps.
Measure what the land produced
Production of each commodity in the project area is averaged over a historical reference period: the three years before the start, or one full crop rotation if that is longer. Grower records come first, then calibrated remote sensing, then census averages. The baseline is grown each year by a yield growth rate; the module offers a default of 2.5%.Subtract what the project still produces
Monitored production in the project area is set against that baseline. The gap is the production subject to leakage. If farming carries on, the gap is small.Take off any mitigation
Optional. Extra production on land outside the project, secured by agreement, reduces the gap. See the mitigation section below.Turn production into hectares
Two defaults convert the gap into new land. For displaced crops or grazing, 75% of the lost production is assumed to be replaced by more supply elsewhere, and 40% of that from newly cleared land. For fuelwood both defaults are 100%. Divided by the local yield per hectare, that gives the area of new land brought into production.Count the carbon on that land
The new land is assumed to be forest, losing all its biomass, plus a soil carbon change. A project may use the most common native ecosystem in its ecoregion instead only if forest covers less than 10% of the country, or if the country's gross deforestation rate has been under 0.1% a year for ten years.
Multiply the two crop defaults and you get the core of the module: 75% × 40% = 30% of displaced crop or livestock production is assumed to come from newly cleared land. Leakage is calculated for at least five years from the project start, and displacement is treated as a one-time event, so it is not charged again at every verification.
The forest assumption is why leakage is so heavy. A hectare of new pasture elsewhere is charged as a hectare of cleared forest. On a dryland site where your own trees add carbon slowly, a few hundred hectares of assumed forest clearing can outweigh most of what the project stores.
Published figures: from 0% to 79% of gross
Our screenings follow VM0047 and apply the VMD0054 defaults. When the intake does not say how the land is used today, the screening has to assume. For block planting it treats the part of each planting zone mapped as cropland or rangeland as displaced; on the Kenyan site that share was 98.2% and on the Colombian site the whole planting area counted. For agroforestry the base case assumes the farming carries on under the trees, so nothing is displaced, and only the low case treats the cropland share as displaced. That is the conservative choice, and it shows how much the answer matters. All figures below are base-case screening estimates over 40 years, in tCO2e.
| Gross removals and leakage | Why | |
|---|---|---|
| Kenya, dryland highlands (ARR) | Gross 69,779; leakage 55,310 (79% of gross) | Land use not given, so the planting area counted as displaced grazing in proportion to the 98.2% of the site mapped as crops and rangeland. Slow tropical shrubland growth rates. |
| Colombia, Llanos pasture (ARR) | Gross 372,500; leakage 209,200 (56%) | Whole planting area treated as displaced cattle pasture. Fast rainforest-zone growth rates. |
| Uganda and Côte d’Ivoire (agroforestry) | Leakage 0 in the base case | Farming continues under the trees. The low case treats the cropland share as displaced. |
Compare Kenya and Colombia. Both were treated as almost entirely displaced grazing land. Leakage in Colombia was nearly four times larger in tonnes, but a smaller share, because the trees there grow much faster. In Kenya the slow growth of dryland planting meant leakage swallowed most of the gross figure. The pattern holds generally under VMD0054: leakage scales with what you displace, gross removals scale with how fast your trees grow.
Here is a full deduction table, from the anonymised sample study on our site. Leakage is the largest line, bigger than uncertainty and the buffer together, and it takes the net figure below zero.
| Line | tCO2e | % of gross | Basis |
|---|---|---|---|
| Gross removals | 69,779 | 100% | IPCC Tier 1 projection, average case |
| Uncertainty deduction | −6,978 | −10% | 10%, the VM0047 minimum for ex-ante estimates |
| Non-permanence buffer | −10,467 | −15% | 15%, registry median of ARR projects |
| Leakage | −55,310 | −79% | VMD0054 defaults; land use not given, so the mapped crop and rangeland share counts as displaced |
| Net screening estimate | −2,976 | −4% | over 40 years, average case, on 1,608 ha of eligible land |
The country pages show the same studies in context: Kenya and Colombia.
Why agroforestry often carries little leakage
VMD0054 charges for production that stops. Agroforestry plants trees into working farmland: coffee under shade trees, maize between rows, cattle under scattered trees. If the farm still produces what it did, the gap between baseline and monitored production is small, and so is leakage. That is why the two agroforestry screenings in the table above, in Uganda and Côte d'Ivoire, carried no leakage in the base case.
Two caveats. First, the baseline grows each year by the yield growth rate, so production that merely holds steady can still show a shortfall over time. Second, the claim has to be proven with monitored production, not assumed. A screening shows both: a base case where farming continues, and a low case where the cropland share is displaced.
There is also a methodology route. VM0047 has a census-based approach for counted trees on land with less than 10% woody cover that is under continuous cropping, settlements or other land. For that approach VM0047 sets leakage to zero. Whether your planting design fits it is a question for the method check. More in agroforestry carbon credits.
How to reduce leakage before you design the project
The cheapest leakage is the leakage you design out. Most of these choices are made long before validation, which is why they belong in a feasibility study.
- Tell the screening the current land use, and where it goes. On farmed land this one answer moves the net figure more than anything else. If the herd moves to land that is already pasture, or the field was fallow, say so and give the evidence you have.
- Draw planting zones on unused land. VMD0054 allows leakage to be treated as de minimis only where the land was not used for farming or fuelwood in the ten years before the start, shown by remote sensing and attestations. Abandoned land qualifies only if it really has been out of use that long.
- Keep production under the trees. Agroforestry, silvopasture and fodder trees reduce the production gap. If you introduce a new crop, version 1.1 lets you count it as cross-commodity production, under stricter default values and conditions.
- Intensify elsewhere. Support better yields on farmland outside the project, for example improved pasture for the displaced herd. VMD0054 credits this only in the same country, outside the project area, under an agreement of at least five years, with extra fertiliser and fuel emissions counted. Livestock intensification must stay within carrying capacity, confirmed by a qualified local expert.
In practice, the intake questions that move leakage most are short. What is the land used for today, and by whom? How many animals graze it, or what is harvested and how much? Where would that use go if the land were planted? Is there land nearby that is already farmed or grazed and could absorb it? Has any part of the area been out of use for ten years or more, and can that be shown on imagery? Even rough answers let a screening replace the blanket assumption that everything mapped as cropland or rangeland is displaced.
Leakage also interacts with additionality and the baseline. A project that keeps farming going may find it easier to show that farmers would not have planted trees anyway. See additionality for tree-planting projects and the non-permanence buffer for the other deductions.
What a feasibility screening can and cannot do
A carbon feasibility study maps the cropland and rangeland inside your planting zones, applies the VMD0054 defaults and shows leakage as its own line, next to uncertainty and the buffer. The hub page on gross vs net carbon credits puts all the deductions side by side. Because the headline stays gross and the net has its own section, you can see exactly what changes if you supply better land-use data.
Frequently asked questions
Leakage is an increase in emissions outside a project that the project causes. In tree planting it happens when the crops, cattle or fuelwood collection that used the land before move elsewhere, and land there is cleared to make room. Under VM0047 the leakage estimate is deducted from the removals the project can claim.
VMD0054. Version 1.0 was the "Module for Estimating Leakage from ARR Activities". Version 1.1, active since 13 January 2026, is called "Estimating Leakage from the Displacement of Agricultural Activities". Projects using the VM0047 area-based approach must apply it. VM0047 sets leakage to zero for the census-based approach.
In the Kenya and Colombia screenings the intake did not say how the land was used, so the screening treated the planting area as displaced grazing: in Kenya in proportion to the 98.2% of the site mapped as crops and rangeland, in Colombia in full. Leakage depends on what is displaced, not on how fast the trees grow, so on a slow-growing dryland site it took 79% of gross removals.
VMD0054 lets leakage be treated as de minimis only if the land was not used for farming or fuelwood in the ten years before the project start, shown with remote sensing and attestations from landowners or managers. Agroforestry that keeps the farming going also displaces little production, which is why the Uganda and Côte d'Ivoire agroforestry base cases carried no leakage.
Plant on land with no current agricultural use, keep production going under the trees, or support intensification outside the project area. VMD0054 credits intensification only with a written agreement of at least five years, in the same country, outside the project area, with any extra fertiliser or fuel emissions counted.
Sources
- Verra: VMD0054 module page (v1.0 and v1.1, status and dates)
- Verra: VMD0054 Estimating Leakage from the Displacement of Agricultural Activities, v1.1 (13 January 2026)
- Verra: VM0047 Afforestation, Reforestation and Revegetation (methodology page)
- Verra: VM0047 Afforestation, Reforestation and Revegetation, v1.1 (14 May 2025)