The carbon credit system enables companies to fund initiatives aimed at reducing or eliminating greenhouse gas emissions. However, the credibility of these credits, particularly within voluntary carbon markets, has come under scrutiny with regard to a crucial element: additionality.
A project is deemed ‘additional’ if it plays a role in reducing greenhouse gas (GHG) emissions. Specifically, in the context of Reducing Emissions from Deforestation and Forest Degradation (REDD) projects, this means that the associated forests would have been cleared without the intervention of the project.
Researchers from the Climate Policy Initiative (CPI) and the Pontifical Catholic University of Rio de Janeiro (PUC-RIO) have introduced a new methodology for establishing baselines and assessing deforestation on private Amazonian properties in Brazil. This new approach integrates economic factors into land-use dynamics, moving beyond mere statistical models.
Brazil is poised to emerge as a leader in the generation of carbon credits, especially through nature-based solutions like forest conservation.
Historically, assessments of additionality have relied heavily on statistical and spatial data, often extrapolating historical deforestation trends. However, such data can be manipulated to inflate carbon credit generation, which undermines the ecosystem’s credibility.
The new model requires landowners in forested regions to assess the most economically viable use of their land: preserving the forest, participating in a REDD project, or converting it for agricultural purposes. This decision is influenced by factors such as agricultural and carbon prices, transportation costs, agricultural productivity, and carbon stocks.
The proposed methodology estimates the opportunity cost of potential deforestation and aims to eliminate bias since it does not rely on specific time frames or reference regions. The model operates over a historical span from 2010 to the present and employs a substantial sample of over 13,000 properties throughout the Amazon.
When applied to existing REDD projects in the Brazilian Amazon, the research indicates that 77% of the carbon traded under these projects is indeed ‘additional’, contributing to the prevention of approximately 0.5 gigatons of carbon dioxide (CO2) equivalent from being released. Notably, 23% of carbon stocks would have remained protected even without project incentives such as REDD.
The study highlights that additionality rates differ across regions. Non-forested and recently deforested areas exhibit high additionality rates of 98.5% and 93.8%, respectively, largely due to the substantial pressure from agribusiness to clear remaining forests, supported by lower transportation costs and improved infrastructure. Conversely, forested regions show a lower additionality rate of 79.4%.
An example cited in the study is the municipality of Portel in Pará, where despite significant carbon stocks, properties within REDD project areas have low agricultural productivity and high transportation costs, making agriculture economically unfeasible. The CPI/PUC-RIO model predicted these areas would remain forested regardless of carbon credit revenues, reflecting a low additionality rate of 57% based on their analysis.
Properties that are larger, possess higher agricultural productivity, and have lower transportation costs are deemed most suitable for REDD projects, as they face the highest risk of deforestation.
The new methodology includes defining a baseline for a given area, representing the expected outcomes had the project not been implemented. The fundamental assumption of the model hinges on landowner profit maximization and considers a multitude of economic conditions and property-specific characteristics, such as agricultural prices, carbon prices, transportation costs to the nearest port, distance to highways, agricultural productivity, and carbon stocks.
To estimate additionality, the model first predicts transition probabilities for properties that may convert from forest to agriculture or to REDD activities, using data from sources like MAPBIOMAS and Verra projects. Following this, the model develops equations to assess revenue associated with each land use and establishes a prediction equation correlating the probability ratios to revenue differences between REDD and agriculture.
Research parameters are derived from observed data, allowing the model to theorize various future scenarios based on differing prices, costs, and probabilities of landowners opting for forest, agriculture, or REDD over time.
The study posits a baseline scenario where carbon projects, such as REDD, do not exist. It estimates the probability of deforestation for a specific property without carbon credit incentives. Additionality is calculated as the discrepancy between expected deforestation in this baseline scenario and the amount deforested in a ‘scenario of interest’, which may involve current REDD projects or hypothetical situations with specific carbon pricing.
This research was published on September 15, 2025.
