Analyzing Forest Loss Trends

Introduction

Forest carbon offset projects are a key tool in mitigating climate change by reducing greenhouse gas emissions from deforestation. However, they have faced scrutiny for not being as effective in reducing deforestation as initially claimed. This has prompted increased demands for improved monitoring and evaluation to ensure these projects fulfill their commitments.

Avoided Deforestation (AD) projects work by preventing conversion of extant forest land into non-forest land uses, while Improved Forest Management (IFM) projects involve the adoption of forest management practices that enhance carbon stocks within forests and/or reduce greenhouse gas emissions from forestry activities compared to standard forestry practices.

In this article we explore the forest loss trends for AD and IFM projects, both before and after a project started. We aim to answer questions such as did forest loss decrease after a project started? What were the global trends? Which projects effectively reduced forest loss, and which ones did not perform well?

Methods

To calculate the forest loss rates inside project boundaries we intersected the geospatial database of forest carbon offset projects (Karnik et al. 2024) with the Global Forest Watch (GFW): Year of gross forest cover loss event data (v1.11_2023) produced by Hansen et al. (2013). This is a global dataset on forest loss by year between 2001 – 2023.

We calculated the percent forest loss (relative to the project area) for each project for each year between 2001-2023. We excluded the project start year for calculation of both percent forest loss (hereon referred to as forest loss) before and after a project started. Average forest loss rate after a project started (AD_after) was calculated a year after the project started till 2023. Average forest loss rate before a project started (AD_before) was calculated for a period of 10 years before the project started. If projects started in 2010 or before, AD_before was calculated by including years from 2001 to a year before project start (e.g. start year if is 2008, AD_before was calculated for years 2001-2007).

Where, D_i is the area (Km²) deforested in the year i, A is the total area of the project, P_i  is the percent forest loss in year i, and n is the number of years in the period.

If, AD_before > AD_after , the project observed a decrease in forest loss after its start. And vice versa if AD_before < AD_after , the project observed an increase in forest loss rate after its start. In a small number of cases projects had no forest loss both before and after (Figure 1).

Caveats

Our analyses come with a set of caveats we mention here, urging the reader to interpret our findings with caution.

• GFW data.
  1. The assigned year for observed tree cover loss was accurate 75.2 percent of the time and was within one year before or after the actual year 96.7 percent of the time. Applying a 3-year smoothing window would have improved the accuracy of our results.
  2. The GFW data are better suited for a landscape scale analyses, and interpretations at a local scale should be made with caution. Performance varied by ecoregion, with certain forest types such as sparse deciduous tropical forests being less accurate. We refer readers to Hansen et al. 2013 to learn more.
  3. In some locations, cloud cover may have been misclassified as forest cover loss.
• We included only a subset of projects (63.8%) for which project boundaries were available in the geospatial database of carbon offset projects. The remaining projects declined to share their project locations either publicly or with Renoster upon request.

Improved Forest Management

Improved forest management (IFM) projects are designed to take forests that have historically been commercially managed, and manage them more sustainably to increase carbon stock. Typically these projects take place in the United States, Mexico, and Canada. Collectively we analyzed 237 number of IFM projects in total, with 161 in the United Sates, 52 in Mexico, and 3 in Canada. We reviewed projects across three databases - The Climate Action Reserve, Verra, and The American Carbon Registry. We analyzed credits that used both the California Air Resource Board compliance forestry protocol, as well as voluntary protocols developed by registries.

On the whole, we observed that forest loss increased on average from 0.29% in the ten years prior to the project, to 0.34 % after project start. A two-tailed t-test comparing forest loss before and after had a p-value of 0.32, suggesting the difference in forest loss before and after project start is not significant. While this means that the noted increase in forest loss is not statistically significant, it also indicates that there is no significant decrease in forest loss overall.

Of the projects analyzed, 48.9% projects saw an increase in forest loss while 47.7% projects had decreased forest loss after the project was implemented (Figure 1). The maximum forest loss rate across projects was 1.96 (AD_before ) and jumped to 5.33%(AD_after).  Eight (3.4%) projects had no forest loss before or after start of projects.

Project-level insights

The total forest loss (percentage) after project start was highest in the Warm Springs project (ACR260) where 48% of the project area suffered a loss (Figure 3b) and large tracts were burned in in a 2020 wildfire (Figure 3a). Likewise, the next most deforested project (Eddie Ranch, CAR1174) also suffered fire losses in 2018. This corroborates with our findings where we classified them in the high fire risk category. Beyond that, losses were a combination of forest harvesting and natural disturbance.

The third most-disturbed project (McCloud River, CAR429) is being clear cut in patches that appear to us to resemble business-as-usual timber harvest. The project was registered in 2012 and most of the forest loss dates to after that period. Under the California Air Resource Board compliance protocol, projects are allowed to harvest down to a baseline that is defined by regional average, or -20% of starting carbon stock. If the project’s forests were more productive than its region (indeed, this project’s region included arid areas with no trees), aggressive harvesting is allowed to take place. Thus the forest loss in this project and many others on this list speak to a flawed baseline methodology. These findings echo those in Stapp et al. 2023 - Little evidence of management change in California’s forest offset program.

Deforestation Inside Project Area After Project Start

Some of the IFM projects that were most successful in reducing forest loss inside the project after its start were CAR1094 (near Washington’s Mt. Rainier), ACR483 (located in Florida on land historically intensely managed for pine plantations), and VCS2609 in Malaysia (Table 2).

Take-aways

On the whole, improved forest management projects are designed to improve upon existing management by promoting more sustainable harvest activities (such as extended harvest rotations). If the programs were working as intended, we would have expected an average decrease in observable harvests following project enrollment, rather than an increase.

In our experience, many IFM projects fail to truly improve upon a prior state, but instead improve upon a hypothetical baseline that is far more aggressive than project’s past actions. Many projects being enrolled for these programs have a previous history of conservation or light forest management, and thus are not improving anything. This factor, coupled with natural disasters, may be leading to the small increases in losses observed.

Avoided Deforestation

Important Context

It is important to note that one might not expect decreases in deforestation inside of avoided deforestation projects as one would in IFM project types. There are several reasons for this:

• Avoided deforestation projects do not enroll land that had historically been deforested. Instead they tend to only enroll intact forests. Therefore, one might anticipate historic rates of deforestation to be low inside a project prior to its start date, even if rates in the surrounding areas were high.
• By their nature, these projects are intended to take place in some of the most at-risk forests in the world. They would not be impactful projects if there were not severe risk off loss. We cannot expect complete effectiveness in these regions, and indeed, many projects plan for partial effectiveness.
• Avoided deforestation projects work on avoiding losses that are likely to happen in the future. Even if they fail to avoid some loss, they may be partially effective at preventing much greater losses in that area.

Many of our favorite AD projects have increases in deforestation for the reasons stated above. They are protecting an area that has so far been spared by deforestation (and thus has a historic deforestation rate of 0), they’re facing enormous deforestation pressures, and minor amounts of deforestation have occurred inside of them but been reported properly. Simply looking at deforestation rates before and after the project might lead one to conclude that it was no successful. Looking at the project’s accomplishments of keeping the area intact on the whole, leads to the conclusion that the project is successful.

For these reasons, we categorically discourage readers from making broad conclusions about the effectiveness of avoided deforestation programs based on this data alone. We encourage users to examine the spatial dataset that accompanies this to explore which projects have reduced deforestation the most.

‍Observed Trends

For AD projects, forest loss increased in 73.8% of the projects and decreased in 25.4% projects after project started. The average forest loss rate after project started was 0.38% which was higher then the AD_after at 0.22%. There was a slight increase in maximum forest loss rate from 5.00% (AD_after) to 5.45% (AD_after). Most projects aim to reduce forest loss compared to their baseline scenarios. Even though forest loss has increased after the projects started, it is possible that these projects have still reduced forest loss rates compared to what would have occurred without them. However, this analysis did not include a comparison to the baseline scenarios.

The Amazonian countries of Colombia (n=20), Brazil (n=28), and Peru(n=12) had the highest number of projects that saw an increase in forest loss rates after project implementation in AD projects. An article by Infoamazonia, an independent media outlet, points out these countries have also had the highest forest loss between 2001 and 2020 (outside project areas) and is attributed to livestock farming, farming, mining, and road construction.

Project-level insights

At a project-level, we calculated the total forest loss (percentage) after project start by summing the yearly percentage forest loss after project start. It is worth noting that projects are not credited for their entire forest area, often creditable carbon amounts to a fraction of the project’s total carbon stock.

The worst project was Tumring (VCS1689) with 43.6% of the project deforested (Figure 2a). This is a project that Renoster refers to as a ‘zombie’ project, abandoned by the project developer but still listed as active by Verra. The project has not been remeasured in nearly ten years. Following this, Serra do Amolar (VCS 2566) was most deforested, this project is in the Pantanal Wetlands of Brazil and was impacted by fires recently. The third most deforested project was Surui in Brazil (VCS1118), deforested once again by humans and like Tumring, essentially a zombie project not verified for seven years but still listed as active on Verra’s platform. We found no indication that Verra’s buffer pool has been used to make up for credits issued to any of these projects.

The projects with greatest decreases in forest loss were all United States-based avoided conversion projects (we classify these as avoided deforestation). All of these were successful at reducing forest loss from unsustainable levels down to sustainable rates. It makes sense that these would top the list, because each of them had a pre-existing history of forest management to improve upon, although they are still based upon the idea that the land could have been deforested and developed. Thus this finding is an artifact of different methodologies, rather than a statement of project’s effectiveness.

Click on the projects in the interactive map below to view their deforestation trends.

‍

‍

References

1. Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D. Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, and J. R. G. Townshend. 2013. High-Resolution Global Maps of 21st-Century Forest Cover Change. Science 342 (15 November): 850-53. Data available on-line from: https://glad.earthengine.app/view/global-forest-change.
2. Karnik, A., Kilbride, J., Goodbody, T., Rachel, R., & Ayrey, E. 2024. A global database of nature-based carbon offset project boundaries [Data set]. Zenodo. https://doi.org/10.5281/zenodo.11459391.
3. Stapp, J., Nolte, C., Potts, M., Baumann, M., Haya, B. K., & Butsic, V. (2023). Little evidence of management change in California’s forest offset program. Communications Earth & Environment, 4(1), 331.
4. West, T.A., Börner, J., Sills, E.O. and Kontoleon, A. 2020. Overstated carbon emission reductions from voluntary REDD+ projects in the Brazilian Amazon. Proceedings of the National Academy of Sciences, 117(39), pp.24188-24194.
5. Zanon, S. 2023. Deforestation in the Amazon: Past, present and future. InfoAmazonia. https://infoamazonia.org/en/2023/03/21/deforestation-in-the-amazon-past-present-and-future/

‍