Verra’s New VM0047 v1.0 ARR Methodology

On September 28, 2023, Verra officially released VM0047 v1.0, a new ARR methodology using a dynamic baseline meant to bolster the integrity of reforestation projects. As many readers of this blog will know, Renoster is a proponent of dynamic baselines so we have taken a particular interest in this new methodology. This new methodology allows projects to choose between an Area-based approach (dynamic baseline) and a Census-based approach (traditional baseline), the latter of which is designed for small project areas. We will focus on the Area-based approach. Below we will summarize the key takeaways from the new method and provide honest feedback on where the method excels and how we think it can be improved.

Additionality

The first big change is how additionality is determined. Previously, ARR additionality was based on land-use history where the project area must have been classified as non-forest for the past 10 years, and a common practice analysis was performed on lands in the surrounding region. This new methodology approaches these questions differently. While it still uses a 10-year historical window, instead of using a binary forest/non-forest criteria it uses matching control plots based on a series of factors to track surrounding land similar to the project area over time (more on this in a moment). So even if the project area had cleared its forest only a few years before the project start, the dynamic baseline would look for areas where the forest was cleared within a similar time frame and track the progress of that land moving forward.

What we like:

• This makes implementing projects much easier because there is no longer a burden of proof for additionality placed on the project proponent other than a regulatory surplus requirement and investment barrier analysis, which are both standard checks through Verra protocols. Instead, the historical and regional land-use analysis is done algorithmically.

What we feel can be improved:

• We would prefer the old 10-year historical check be kept in place. The new method opens the door to a project proponent cutting their own trees then immediately replanting and enrolling in ARR, which is a scenario considered non-additional by most standards, including ours at Renoster. While planting trees after clear cutting is certainly better than cattle ranching or some alternative land use, we do not want to incentivize clear cutting at all prior to ARR, especially when a native forest could be replaced by a monoculture or mixed non-native species plantation. This would also allow an industrial landowner to enroll their plantations immediately after a commercial harvest. Another risk posed by the removal of this check is that commercial timber plantations would be eligible to enroll their land following a normal clear cut. They would in essence be able to receive carbon offsets for trees that were being planted anyway.

• It does not account for landowner intent, which asks whether the landowner was going to plant trees regardless of credit incentives. Admittedly, this latter point is difficult to assess. The most straight forward example is Weyerhauser enrolling their eucalyptus plantations into ARR projects in South America when they were clearly going to plant those trees anyway. We feel that a reasonable change to this protocol would be to exclude industrial timber projects taking place in regions where more than 5% of matched neighbors exhibited similar plantations. Thus ruling out projects that don’t deviate from the common practice.

‍

Carbon Stocks

Carbon stocks within the project area will still be measured by hand using traditional inventory field plots.

What we like:

• While remote sensing technology has dramatically improved, it is still difficult to accurately measure carbon stocks of young plantations when using satellite data. This approach seems reasonable for now.

What we feel can be improved:

• LiDAR is being increasingly used by industry to estimate plantation biomass. Allowing projects to use LiDAR under the right circumstances could help lower the cost of verifications and improve monitoring.

‍

Baseline

First, it’s important to understand how dynamic baselines work. The short explanation is that annual baseline emissions are determined by observing areas around the project that share similar ecological and legal characteristics to the project area. These areas are measured in “real-time” (1-year increments, ex-post) and do not depend on modeled future predictions. For a more detailed overview of dynamic baselines, read Renoster’s previous blog post or Renoster’s white paper.

The project’s dynamic baseline will be determined by measuring the stocking index (SI) in control plots located within the surrounding region that share similar legal and ecological characteristics to the project area. Notably, SI is measured in these control plots, not carbon stocks. Baseline emissions will be calculated by multiplying the annual carbon stock change of the project area to the ratio of SI in the control plots compared to the project area.

The control area plots must be a minimum size of .09 hectares (30m x 30m) and a maximum of 10 hectares. They must match the project area in terms of jurisdictional boundary, ecoregion, policy environment, and land tenure before finally comparing stocking index. They must be outside of any registered AFOLU project area, and they must be within 100km of the project area. A minimum of 30 control plots must be selected to determine the project’s baseline.

Statistically, the methodology uses a k-nearest neighbor optimal matching approach without replacement. The project selects the k control plots most similar to the project area. Each match’s quality is evaluated using the standardized difference of means on each covariate.

What we like:

• Dynamic baselines are generally a more sound method of quantifying additionality than static baselines that are designed to predict the future.

• The statistical methods to select matching control areas appear robust and fair.

• The selection factors used to find matching control areas appear relevant, but we think this can be improved even further (see below).

• Measuring stocking index (SI) is a clever use of remote sensing and circumvents directly measuring biomass, which is difficult with satellite data.

What we feel can be improved:

• This new methodology uses five factors to match control areas plus two other requirements. There is no cap to the number of attributes a project developer may use to match control plots, so it is possible a project developer could use more if they wish, but they are not required to. Renoster’s baseline methodology, for example, uses over 20 bioclimatic variables to match individual pixels. We would like to see the matching control areas selected using more stringent factors such as elevation, slope, distance to transportation corridors, etc.

• The methodology assumes the uncertainty related to the selected control areas is zero. This is unrealistic and is something that should be calculated. Renoster is currently developing a method to estimate uncertainty of our dynamic baseline, which we hope to share soon.

• One critical missing element to Verra’s new methodology is specific remote sensing requirements. Instead, it points to two best practices publications: Global Forest Observations Initiative 2016 and Mitchell et al. 2017. Inevitably, different projects will use different remote sensing data sources and processing methods to generate their baselines. Some will be better than others, and each project will have to be evaluated independently to determine the quality of its credits. While we believe dynamic baselines offer a higher level of ecological integrity compared to fixed predictive baselines, the lack of procedural standardization increases the likelihood of malfeasance. We feel this could be improved by Verra dictating what data and algorithms should be used or having projects use publicly available land cover models such as Global Forest Watch, Dynamic World, or ForObs.

‍

Leakage

The VMD0054 module must be applied with this methodology to measure leakage. We will save our thoughts on leakage for a future blog post.

‍

Native Tree Species

The methodology does not specify what species of trees can be planted inside the project other than the trees cannot affect the local water table. Similarly, the project area cannot be drained for the project’s implementation nor can a project be implemented on tidal wetlands. Any significant amounts of pre-existing dead wood must remain on site.

What we like:

• We generally agree with these requirements and restrictions. Protecting the ecological integrity of local ecosystems is paramount to any carbon project.

What we feel can be improved:

• We would like to see this methodology require a majority component of native species planted in the project. Preferably, a mixture of 4-5+ native species should be planted depending on the local ecosystem. Native forests provide ecological co-benefits for wildlife and humans that cannot often be replicated with non-native plantations. This requirement would also address additionality by discouraging industrial timber companies from enrolling their lands into ARR which they always intended to plant in the first place. Specifically, a mixed species requirement would solve the issue* of industrial Eucalyptus plantations being enrolled in ARR.

• A requirement of some portion of native species or land reserved for pure conservation purposes would go a long way towards addressing the methodology’s additionality issues. In general, native species reforestation and conservation are not common practice.

*Eucalyptus plantations being enrolled in ARR is a complex topic that involves more nuance than we can express in this blog.

‍

Overview

In general, we feel this methodology has some positive elements that improve upon traditional ARR methodologies that rely on static predictive baselines, but there are still many areas that must be improved. Our chief concern is that the approach may still allow for the enrollment of non-additional timber plantations.

While we appreciate the use of algorithmic remote sensing to determine and quantify additionality, we are concerned not enough emphasis is being placed on gauging landowner intent to plant trees. Extra safeguards should be put in place in order to prevent the felling and replanting of native rainforest, the crediting of non-additional timber plantations, and sponsoring common practice timber management. We also wish there were more data transparency and that projects were not generating their own maps. Finally, we wish the methodology enacted an explicit native species requirement for trees planted inside the project. Such requirements would strengthen the methodology’s additionality and safeguards considerably.

The future of the carbon market depends on registries’ willingness to adopt more stringent methods using the best available science. While we appreciate Verra’s willingness to revisit their system, more iterations are needed in order to ensure that carbon offsets truly removing the carbon that they say they are from the atmosphere.

We are glad Verra has started down this path with their new ARR methodology, and we hope they continue to improve with future iterations.

‍

References

Dynamic World, GEE.

ForObs, ForObs Explorer.

Global Forest Observations Initiative (2016). Integration of remote-sensing and ground-based observations for estimation of emissions and removals of greenhouse gases in forests: Methods and guidance from the Global Forest Observations Initiative, edition 2.0. U.N. Food and Agriculture Organization., USDA.

Mitchell, A. L., Rosenqvist, A. & Mora, B. (2017). Current remote sensing approaches to monitoring forest degradation in support of countries measurement, reporting and verification (MRV) systems for REDD+. Carbon Balance and Management, 12, 9., BMC.

Verra’s VM0047 v1.0 ARR Methodology, Verra.