Sylvera response to The Guardian’s Analysis of Rainforest Offsets

Not all carbon credits are high quality, and it’s both valid and important to highlight concerns with credit quality—that’s why Sylvera exists.

Analysis published in The Guardian yesterday claimed that 94% of REDD+ credits do not represent genuine carbon emission reductions. While we agree that unfortunately, only a minority of credits are legitimate – we have said this publicly since launching in 2020 – the 94% claim is hugely overstated, not in line with our own deep analysis, or representative of some of the cited papers.

Scrutiny is sorely needed; it’s what our team of 150 does on a daily basis. However, misrepresentation of a consensus and repeating a flawed analysis are detrimental to the viability of a vital, yet nascent market.

Forests need to be protected to deliver our climate goals. Of course, we should absolutely call out when this is falling short, but we must not sabotage the financing of projects that are delivering sound climate, social, and biodiversity benefits…even if well-intentioned.

The analysis highlighted in The Guardian hinges on an assessment of the strength of the projects’ baselines – in other words, how accurate a project’s claim is, about how much deforestation would have happened if the project had never taken place.

As a counterfactual, “what-if?” exercise, this is difficult to estimate and critical to get right. We do agree that oftentimes baselines are indeed not estimated correctly, leading to the overissuance of unjustified credits. However, the comprehensive analysis that we published in November, found that in ~30% of cases, we can have sufficient confidence that the baseline estimation is sound. So what explains the gap from 30% to 6%?

Unfortunately, despite claiming to be led by three papers, the “94% bad, 6% good” figure seems to have been calculated solely from an academic paper by West et al (2023). That paper created counterfactuals by selecting areas intended to serve as proxies for the project areas using a technique in development called synthetic baselines. 

While this technique can potentially be a valid approach, it’s only as good as the assumptions used. In this case, control or proxy areas meant to serve as mirrors of what would have happened in the project area had it not been protected, were selected solely based on physical characteristics (i.e. distances from roads, rivers, settlements, slope, etc.). Those physical characteristics, relating to accessibility and proximity to markets are all relevant factors. However, the single and obviously most important factor has been completely omitted—the proximity to the active front of deforestation. 

Case Study: Brazilian Amazon Project, Florestal Sta Maria

This is well illustrated in the image below from the Sylvera platform. It shows Florestal Sta Maria (Verra 875), a project in the Brazilian Amazon.  

The project area, outlined in purple, is boxed in by aggressive deforestation on three sides. Before the project began in 2009, deforestation was prevalent in the south, and since then, it has gone on to envelop the project area. To the north lies a protected area (emerald green). To be clear, the area within the purple bounds was not part of the protected area. It is evident that without the project in place, the project area would have been deforested. This threat is disregarded without explanation or justification in the West et al model used to support the claim that “94% of projects are worthless.”
Furthermore, the project actually projected the evident risk in its local area remarkably well.

The above chart from the Sylvera project assessment captures the project's own projection of the future deforestation rate; the project’s own proxy area is the purple line. As you can see, the developer selected a much lower rate of deforestation than the average in the early 2000s. Despite doing this exercise in 2009, the projection of deforestation running through 2019 was remarkably accurate. This proves baselining can be done well. Yet according to the test applied by West et al, this project has done nothing to prevent deforestation. (It’s worth noting that West et al acknowledge the limitations of their method in their paper, but these limitations have not been included in the Guardian report.)

Sylvera’s Research Shows 30% Are High Quality

In our “State of Carbon Credits 2022” report, published in November last year, we provided an in-depth analysis of >85% of REDD+ credits on the market. It found that 31% of the projects (representing 143M issued credits) we’ve rated are, in fact, high quality (falling in our Tier 1 category of ratings). To be clear: 31% isn’t great, but it’s a lot better than 6% and demonstrates there are good quality projects available to buyers today. Our report also showed that the remaining two-thirds of projects we’ve rated are of mixed quality, with 25% being effectively junk. The Guardian has accurately called out some of these bad projects such as Madre De Dios.

Why we trust our assessment

• We’ve invested millions of dollars and thousands of hours conducting field research and gathering above-ground biomass data; synthesizing optical, radar, and LiDAR data to validate emissions reductions claims; and building and calibrating machine learning models with multiple data sources (see how here) to reconstruct individual project models - including, but not limited to, strength of baseline
• We build meticulous frameworks to test project design, carbon accounting, and impact claims. These frameworks are peer-reviewed by a committee that includes market experts, project developers, and academic researchers
• We use verified, independent data to test the accuracy of credit issuances and claims

Although well-intentioned to call out greenwashing, this article risks damaging one of the best mechanisms we currently have for funding the wide-scale conservation of natural carbon sinks. We welcome the ongoing effort from the research community, and we ourselves are adding to the public scientific knowledge base by publishing our work in collaboration with leading institutions.

We have followed up directly with The Guardian and hope to discuss our analysis with them.