Insights · Gross to net
The dynamic baseline in VM0047: how control plots decide your credits
Land left alone often grows trees by itself. VM0047 does not credit that growth. Instead of fixing a baseline at the start, the area-based approach measures a dynamic baseline at every verification, by watching matched control plots outside the project from space. It is the one deduction nobody can calculate in advance. Here is how it works, and what you can still check before you commit.
Why VM0047 uses a dynamic baseline
A carbon credit should pay for removals that would not have happened anyway. For tree planting, “anyway” is hard to pin down. Abandoned pasture can turn into scrub and then woodland. A drought decade can slow regrowth across a region. A government programme can start paying farmers to plant trees. A baseline frozen at the start would miss all of that.
VM0047 deals with it in its area-based approach. The methodology describes a “dynamic performance benchmark” that is used “to assess additionality and determine the crediting baseline at each verification”. The benchmark is derived from observations after the fact, comparing the change in a vegetative stocking index between remotely sensed project plots and matched control plots. In short: your trees are compared with similar land nearby that had no project, over the same years.
- MethodologyVM0047 v1.1, area-based
- Measured withRemote sensing, at least annually
- UpdatedEach verification, or every 5 years
The census-based approach works differently. For counted trees on land with less than 10% woody cover, or on continuously cropped land, settlements or other land, VM0047 assumes trees would not appear without the project and sets the baseline to zero. Everything below is about the area-based approach. For the overview of both, see VM0047 explained.
How the performance benchmark is built
Appendix 1 of VM0047 v1.1 sets out the procedure. It is a statistical matching method, the kind used in impact evaluation, applied to satellite pixels.
Sample project plots
The project area is split into equal plots of 0.01 to 10 hectares, each at least 75% inside the boundary. At least 30 are picked at random. Grouped projects build a separate benchmark for each annual cohort.Define a donor pool
Control plots can only come from land in the same ecoregion (biome level) and jurisdiction, with the same tree-planting incentive programmes and the same land tenure classes as the project, within 100 km of each project plot. Registered AFOLU projects may be excluded.Match on history
Each project plot is matched to its k nearest control plots on the stocking index at three or more points in time: between ten and eight years before the start, between eight years and one year before, and at the start. Matching is without replacement. The match is valid only if the standardised difference of means is 0.25 or less for every covariate; otherwise the radius grows in 100 km steps or k is reduced.Monitor both
The stocking index is recorded for project and control plots at least annually. Control plots that come under a tree-planting incentive programme the project does not have are dropped and replaced.Compare the slopes
At each verification, the rate of increase in the index is estimated for project and control plots by weighted regression over the whole time series, and the two are compared with a Z test.
The stocking index is left open on purpose. VM0047 calls it an unspecified remote-sensing metric with demonstrated correlation with aboveground carbon, such as a fraction index from Landsat imagery, canopy height from LiDAR, or canopy cover from aerial imagery. Verra's transition guidance notes that the index used to select control plots may differ from the one used for monitoring, but monitoring must use the same metric on project and control plots. The index only measures relative change between project and control plots. The carbon itself is still measured in field plots.
How the dynamic baseline changes credits each monitoring period
The benchmark is a ratio: the control plots' rate of increase in the stocking index divided by the project plots' rate. The project's measured carbon stock change is multiplied by one minus that ratio, before the uncertainty deduction, project emissions and leakage come off.
| Outcome at verification | Benchmark | Effect on credited removals |
|---|---|---|
| Control slope not significant or negative (project slope significantly higher) | Set to 0 | Full measured stock change goes forward to the other deductions |
| Control plots grow, but clearly slower than the project | Control rate ÷ project rate | That share of the stock change is not credited |
| Project and control not significantly different (|Z| < 1.96) | Set to 1 | Nothing credited, and additionality is not shown for that verification |
A simple illustration of the arithmetic: if the index on the control plots rises at one fifth of the rate on the project plots, the benchmark is 0.2 and 20% of the measured stock change is not credited. If the region starts regrowing on its own and the control rate climbs to half the project rate, the cut is 50% for that period.
Three details matter for planning. First, the benchmark is recalculated at every verification, at least every five years, using the full time series since the start, so it can move in either direction. Second, VM0047 treats the benchmark as having zero uncertainty; it does not add to the uncertainty deduction. Third, the same test decides additionality: at each verification the project has to show that its slope is significantly different from the control slope.
Which projects are most exposed
Because the benchmark is a ratio, two things drive it: how fast the land around you regrows, and how fast your own planting lifts the index. Both matter as much as each other.
- Slow planting next to regrowing land. Where natural regrowth in the region is real and your trees grow slowly, the control rate is a large share of the project rate. Dryland planting in a landscape that is greening by itself is the uncomfortable case.
- The early years. Young trees change a satellite index little. If the project slope is not yet significantly different from the control slope, the benchmark is set to 1 for that verification and nothing is credited. Projects should expect early verifications to be the hardest to pass.
- Poor survival. Gaps in the planting pull the project slope down and push the ratio up. Good establishment and fast canopy closure work in your favour twice: more carbon, and a smaller share cut by the benchmark.
The reverse also holds. Where land around the site is being cleared or degraded, the control slope may be negative or not significant. VM0047 then sets the control rate to zero, and provided the project slope is significantly higher, the benchmark takes nothing.
Why ex-ante forecasts are uncertain
Before validation, a project still has to forecast its removals for the whole crediting period, including a forecast benchmark. VM0047 describes how. The project's index is modelled from a growth and yield model, converted to stocking index values with a regression. The control plots' index may be projected as a straight line from its trend over the ten years before the start.
That is a reasonable starting point and a weak predictor. A straight line from the past decade assumes the next 30 or 40 years look the same. It cannot know about a fire season, a drought, a change in grazing pressure or a new national planting programme. Each of those moves the control plots, and so the benchmark, after the forecast is made. The forecast is a requirement for validation; the credits depend on what is measured.
This is why the benchmark is listed among the things a desktop study cannot settle. It is measured during monitoring, from imagery that does not exist yet.
What a feasibility screening can do now
A carbon feasibility study cannot calculate the benchmark, and ours does not pretend to. In the anonymised sample study, 69,779 tCO2e of gross removals became a net screening estimate of −2,976 tCO2e after uncertainty, the buffer and leakage. There is no baseline line: the report marks it “to be determined with comparison plots”. Any credit forecast built on a net figure like that still has the benchmark to come off. The other deductions are explained on the gross vs net carbon credits hub, with leakage covered in depth in leakage in ARR projects.
What a screening can do is look at the same kind of history the matching step uses. Forest-cover history and land cover around and inside the boundary show whether tree cover in the region is stable, being lost or coming back. A landscape where trees are returning by themselves is a warning: the control plots may grow, and the benchmark with them. A landscape with steady clearing or no regrowth is the opposite. The screening flags where natural regrowth looks likely, so you can weigh that risk before paying for design.
You can do a first check yourself. Open the Global Forest Change map and look at tree cover and loss within a few tens of kilometres of your site, in the same kind of land and under the same kind of tenure. Then ask the questions the donor pool will ask: is there a tree-planting subsidy in the area? Is grazing pressure falling?
Frequently asked questions
In the area-based approach of VM0047, the baseline is not fixed at the start. It is a performance benchmark measured at each verification by comparing the growth of a remote-sensing index on your project plots with matched control plots outside the project. Growth that also happens on the control plots is not credited.
The benchmark is the ratio of the control plots’ rate of increase to the project plots’ rate. The project’s measured carbon stock change is multiplied by one minus that ratio. If the control plots grow at a fifth of the project’s rate, a fifth of the stock change is not credited. If the two are not significantly different, nothing is credited.
No. The benchmark is measured during monitoring, from imagery taken after the project starts. A desktop screening can look at forest-cover history around the site and flag where natural regrowth looks likely, which is a risk to the net figure. Our net screening estimate does not include a baseline line for this reason.
No. Under the census-based approach, for counted trees on land with less than 10% woody cover, or on continuously cropped land, settlements or other land, VM0047 sets the baseline to zero. The dynamic performance benchmark applies to the area-based approach only. A project can combine both approaches on separate, non-overlapping areas.