Journal list menu

Volume 33, Issue 3 e2817
ARTICLE
Open Access

Improved assessment of baseline and additionality for forest carbon crediting

Nina A. Randazzo

Corresponding Author

Nina A. Randazzo

Environmental Defense Fund, Washington, DC, USA

Correspondence

Nina A. Randazzo

Email: [email protected]

Search for more papers by this author
Doria R. Gordon

Doria R. Gordon

Environmental Defense Fund, Washington, DC, USA

Department of Biology, University of Florida, Gainesville, FL, USA

Search for more papers by this author
Steven P. Hamburg

Steven P. Hamburg

Environmental Defense Fund, Washington, DC, USA

Search for more papers by this author
First published: 08 February 2023
Handling Editor: Mingkai Jiang

Funding information: Bezos Earth Fund

Abstract

In the California compliance cap-and-trade carbon market, improved forest management (IFM) projects generate carbon credits in the initial reporting period if their initial carbon stocks are greater than a baseline. This baseline is informed by a “common practice” stocking value, which represents the average carbon stocks of surveyed privately owned forests that are classified into the same general forest type by the California Air Resources Board. Recent work has called attention to the need for more ecologically informed common practice carbon stocking values for IFM projects, particularly those in areas with sharp ecological gradients. Current methods for estimating common practice produce biases in baseline carbon values that lead to a clustering of IFM projects in geographical areas and ecosystem types that in fact support much greater forest carbon stocks than reflected in the common practice. This phenomenon compromises additionality, or the increases in carbon sequestration or decreases in carbon emissions that would not have occurred in the absence of carbon crediting. This study seeks to expand upon recent work on this topic and establish unbiased common practice estimates along sharp ecological gradients using methods that do not rely upon discrete forest classification. We generated common practice values for credited IFM projects in the Southern Cascades using a principal components analysis on species composition over an extensive forest inventory to determine the ecological similarity between inventoried forests and IFM project sites. Our findings strengthen the results of recent research suggesting common practice bias and adverse selection. At several sites, even after controlling for private ownership, 100% of the initial carbon stocks could be explained by ecological variables. This result means that improved management did not preserve or increase carbon stocks above what was typical, suggesting that no carbon offsets should have been issued for these sites. This result reveals greater bias than that been found at project sites in this region by research that has used discrete forest categorization.

INTRODUCTION

Improved forest management (IFM) has emerged as a potential low-cost strategy to mitigate greenhouse gas emissions by increasing forest carbon sequestration (Kaarakka et al., 2021; Ruseva et al., 2017). Toward that end, a variety of carbon crediting protocols, namely the American Carbon Registry, the Climate Action Reserve, and the Verified Carbon Standard, include the opportunity for IFM projects to receive carbon credits, primarily in the USA (Goldstein & Ruef, 2016; Kaarakka et al., 2021). In particular, in the California compliance offset market, established in 2012, IFM projects have come to dominate the forest offsets market (Anderson et al., 2017; Goldstein & Ruef, 2016; Ruseva et al., 2017). Because of this dominance of IFM projects as forest carbon offsets in a large compliance cap-and-trade program, it is important to understand the efficacy of current IFM projects in generating actual climate benefits.

For IFM projects to receive credits, landowners must demonstrate an increase in carbon sequestration or an avoided loss of carbon sequestration via shifts in management. The requirement for a demonstrated increase in or avoided loss of carbon sequestration resulting from enrollment in a carbon offset program addresses what is known as additionality (e.g., Gren & Aklilu, 2016), a decrease in net emissions compared with the outcome in the absence of the carbon offset program. A lack of demonstrated additionality is a major concern regarding the efficacy of carbon offsets projects in meaningfully reducing greenhouse gas emissions (Gren & Aklilu, 2016; Shrestha et al., 2022). If a carbon project does not decrease emissions or increase sequestration beyond the counterfactual (i.e., if a carbon project is not additional), then credits generated from this project will not actually decrease net emissions in the carbon market.

The additional carbon sequestration claimed by an IFM project in the California compliance market is typically expressed as an increase in forest carbon stocks relative to a generalized baseline stocking level (e.g., Ruseva et al., 2017). These baseline levels can be regarded as the carbon stocks that would be present in the forest in the absence of the altered management incentivized by the program. The use of a generalized baseline carbon stock is considered the most direct method by which to assess additionality (Ruseva et al., 2017). Baseline carbon modeling may be informed by a “common practice” figure, which represents the carbon stocks associated with typical harvesting and management over a forest type (Gren & Aklilu, 2016). The use of a common practice value prevents a project from claiming an erroneously low baseline from an unrealistic counterfactual management plan; this method, therefore, can at least partially alleviate concerns of a lack of additionality because of comparisons with an unrealistic counterfactual (Anderson et al., 2017).

In the California compliance offset market, the baseline is established by modeling carbon stocks under a business-as-usual management scenario (California Environmental Protection Agency Air Resources Board, 2015). The modeled baseline carbon often converges to the minimum baseline value that is allowed to be claimed under this protocol. For projects with higher-than-average initial carbon stocks, this minimum baseline value is set by the common practice stocking value (Badgley et al., 2022; California Environmental Protection Agency Air Resources Board, 2015; Ruseva et al., 2017). The common practice stocking value is estimated for each IFM project by averaging the carbon in standing live aboveground tree biomass on privately owned forests that are classified as the corresponding “assessment area” within the project's “supersection,” using forest survey data from the Forest Service's Forest Inventory and Analysis (FIA) program (California Environmental Protection Agency Air Resources Board, 2015; McRoberts et al., 2005). A “supersection” is a geographical area, such as the Southern Cascades or the Sierra Nevadas, whereas an “assessment area” is a vegetation community, such as a mixed conifer or mixed oak. Credits may be generated for projects' initial carbon stocks above the common practice-informed baseline to incentivize the continuation of management practices that have successfully increased or preserved carbon sequestration. Theoretically, in the absence of these carbon credits, existing economic forces would incentivize a decrease in the carbon stocks from current levels to the common practice-informed baseline, which represents the carbon stocks that would be expected under typical private management. Therefore, the carbon credits for all carbon above this baseline would represent an additional reduction in net carbon loss (Badgley et al., 2022; California Environmental Protection Agency Air Resources Board, 2015). These upfront credits constitute an important source of carbon credit payments (Anderson et al., 2017; Badgley et al., 2022; Kelly & Schmitz, 2016).

Despite the theoretical benefits of a common practice-informed baseline for ensuring additionality, concerns about the actual implementation of this baseline approach on carbon stocks in the California offset market remain. For example, critics of the implementation of this method note that the assessment areas within a supersection are often so broad as to encompass a multitude of forest types with distinct carbon-carrying capacities and rates of carbon sequestration. Oversimplification of the assessment area can lead to credits from the initial crediting period, or “upfront credits,” being generated in forests that are more carbon dense simply due to biophysical conditions, such as species composition and soil type, than the baseline carbon density estimate (Badgley et al., 2022). This greater carbon density may not reflect an active increase in carbon sequestration via atypical management, and credited forests may actually have a comparable level of carbon stocking and sequestration as unenrolled forests with similar characteristics (Badgley et al., 2022; Coffield et al., 2022). The ability to generate carbon credits without altering management may incentivize adverse selection, the process by which a disproportionate number of projects applying to the program are projects for which initial carbon stocks are greater than common practice, due to factors other than an increase in carbon stocks, due to atypical management (Badgley et al., 2022). When upfront credits are issued for existing carbon stocks in forests that probably would have produced comparable levels of carbon stocking under typical management, the additionality requirement is not met. Therefore, when adverse selection occurs, a large proportion of offset credits are nonadditional.

This issue is particularly concerning for projects located where sharp ecological and geographical gradients significantly affect carbon-carrying capacity, such as the Mixed Conifer assessment area in the Southern Cascades supersection, a mountainous region that stretches from near the coast of Northern California and Southern Oregon to more inland areas just north of the Sierra Nevada mountain range (Figure 1; Badgley et al., 2022). This assessment area contains extremely carbon-dense forests on the western-most edge of the Southern Cascades, which experiences a climate associated with proximity to the coast, along with much less carbon-dense forests further inland (Badgley et al., 2022). The vast majority of Mixed Conifer sites that have been issued initial reporting period credits are on the coastal edge of this supersection (13 out of 15 projects) and have initial carbon stocks greater than the common practice levels (Badgley et al., 2022). The evidence for adverse selection in the Southern Cascades is especially troubling because this supersection contains more IFM offset projects than any other supersection (Badgley et al., 2022).

Details are in the caption following the image
The portion of the Southern Cascades supersection that falls within California (supersection outlined in green) extends from near the coast to inland regions.

These patterns raise the need to ensure that baseline carbon estimates need to be done in a way that better incentivizes additionality and minimizes or eliminates adverse selection. Such a method must account for the effects of pre-existing ecological factors on carbon density and carbon growth rates, minimizing the gaming of crediting systems with limited benefits to the climate. One method of addressing ecological factors that influence pre-existing carbon stocks is to compute common practices using smaller and more ecologically specific classes of forests within supersections than are described by current assessment areas, as in Badgley et al. (2022). However, methods that continue to use discrete classifications of forest types for common practice calculation, as in Badgley et al. (2022), may preserve bias, as ecological variability that depends upon environmental conditions may still exist within a classification, even one that is more ecologically informed than current assessment areas. Yet, breaking down a discrete classification into even smaller and more ecologically distinct classifications in order to reduce this bias may result in sample sizes of applicable FIA plots being too small to support statistically rigorous claims (Badgley et al., 2022).

Here, we propose a method that produces a more ecologically informed estimate of baseline carbon that still relies on FIA data, yet avoids the need to reduce sample sizes. Our method uses ecological gradients that correspond to California's geographical features to more accurately estimate carbon stocks for a given site. Rather than constraining our sampling to geographically narrow regions and then breaking down our sampling further into discrete categories based on ecology, our methods select reference sites based on gradients of species composition. This approach allows for the selection of the most ecologically similar FIA reference sites for each project site without restricting the sample of FIA sites based on supersection boundaries. While management can influence species composition, we show that our method identifies species variations controlled primarily by geography, not management, so it avoids biasing common practice values based on land-use history. As a case study, we use this method to calculate new common practice estimates for Southern Cascades Mixed Conifer project sites, compare these new estimates to the currently listed common practice carbon stocks, and assess how geography and ecology may influence bias in common practice estimation and, as a result, initial crediting period offset generation.

While we focus here on a California supersection, these methods are applicable to IFM projects in any region with substantial forest inventory data. Because a majority of IFM projects are located in the USA, where FIA provides extensive inventory data, these methods are highly relevant for the study of IFM additionality and baseline bias. This approach has the potential to strengthen the integrity of the forest offset market as an effective climate mitigation tool.

MATERIALS AND METHODS

As in the California Air Resource Board's (CARB) current methods (California Environmental Protection Agency Air Resources Board, 2015), our methods for the calculation of common practice carbon stocks use the United States Forest Service's FIA data, a systematic and long-term collection of forest survey data across the USA (Burrill et al., 2018). The survey plot datasets, organized by state, include the estimated average stand age, tree-level species composition and aboveground carbon estimates, as well as other relevant information. Coordinates are slightly altered for landowner privacy. We used all data collected within the last 10 available years (2010–2019) for the State of California and excluded sites that were noted to have experienced a major disturbance since the last measurement (in the case of California, measurements are made on average once per decade), were missing stand-level data, or contained zero recorded trees. In total, 2727 sites were included in the calculations. Because CARB's current methods calculate common practice stocks using only aboveground live tree biomass, we focused on that carbon pool. While other carbon pools are considered in the allocation of CARB's offset credits, minimum acceptable baseline values are not calculated using FIA samples, so our methods were not applicable.

For each plot, carbon in aboveground live tree biomass per unit area was calculated by summing the tree-level aboveground woody biomass carbon estimates provided by FIA and multiplying by a provided scaling factor that corrects the surface area of the sampled subplot. These carbon stocks allow a direct comparison with the reported common practice values for the carbon stocks in aboveground live biomass. We have expressed all carbon values as megagrams of CO2 (Mg CO2) per hectare.

We then established ecological gradients of FIA sites according to tree species composition. To this end, we constructed a matrix of relative tree species basal area for all of the FIA plots based on tree-level species codes and trunk diameter measurements to identify the primary axes of variability in forest species composition. A principal component analysis (PCA) was performed on the species frequency matrix using the R package FactoMineR. PCA identifies axes of greatest variability across a dataset and over multiple variables with no prior assumptions about the relative importance of variables and therefore can be used to characterize important ecological differences over a large number of sites; for this reason, this method is one of the most commonly used methods in ecology to characterize site differences and has proven useful in characterizing forest sites based on tree species (Syms, 2008; Verburg & van Eijk-Bos, 2003). Because matrix sparsity may lead to excessive sensitivity in the PCA algorithm (Syms, 2008), species constituting the least abundant species at the 30th percentile over all sites were excluded. Due to a highly skewed distribution, the excluded species constituted, on average, less than 0.04% of the basal area in the forests in this study. For identification of typical forest types for visualization and a general understanding of PCA results, clustering was performed on the FIA sites based on PCA results (also using FactoMineR), but no discrete categorizations were used in the actual calculations of common practice carbon stocks.

In order to use these axes of ecological variability to calculate common practice carbon, each site was projected onto the PCA axes using its species composition reported in the publicly available application to the IFM program (see Appendix S1 for details). For this step, sites were restricted to those under private or tribal ownership (not distinguished in publicly available FIA datasets). The restriction of datapoints by ownership class maintained consistency with current common practice calculation methods and prevented potential positive biases that could arise from including publicly-owned forests, which are often harvested less intensely. The carbon stocks of the 5% of FIA sites (137 sites) that are nearest to the project site in the first three axes of PCA space were then averaged, weighted by proximity to the project site in PCA space. Bootstrap resampling of FIA sites with 1000 iterations was used to estimate the 90% confidence interval (CI) for each common practice estimate. The same analysis is then performed using FIA sites on public rather than private lands. This analysis provides insight into how ecosystem type affects carbon-carrying capacity, independent of harvesting practices, which can be expected to have a greater effect on privately owned forestlands.

The recalculated common practice and the public forest reference values were compared with the reported initial aboveground live biomass carbon, official common practice, and official aboveground live biomass baseline (where available). This information is available from the online databases of Climate Action Reserve and American Carbon Registry offset projects (https://thereserve2.apx.com/ and https://acr2.apx.com/myModule/rpt/myrpt.asp?r=111, respectively) and from the forest project spreadsheet made publicly available by Badgley et al. (2022).

To apply these methods, we focused on project sites that were entirely or partially situated in the Southern Cascades supersection and classified into the Mixed Conifer assessment area, for several reasons. The Southern Cascades supersection contains more IFM projects than any other supersection in the California market, and the Mixed Conifer assessment area in this supersection contains sharp ecological gradients and therefore is vulnerable to geographical bias (adverse selection) in common practice calculations (Badgely et al., 2021). Using Badgley et al.'s spreadsheet of IFM projects, we identified Southern Cascades Mixed Conifer IFM project sites that (a) exceeded common practice in their initial reporting period, (b) were located entirely or partially in the California area of this supersection, (c) were classified entirely or partially into the Mixed Conifer assessment area of this supersection, and (d) had self-reported species compositions that left less than 10% of the species composition unspecified in the publicly available documentation. Out of 15 sites in this assessment area that received upfront credits, of which 13 are located on the coastal edge and two are located further inland, we were able to analyze 10 sites, of which nine were located along the edge nearest to the coast and one was located further inland. The additional five sites were excluded because the applications associated with these sites did not provide specific enough species composition information for effectively deploying the methods presented here.

RESULTS

The species composition PCA captured ecological gradients of forests across California. The first three axes identified in the PCA represented 9.3% of the variability of species composition by basal area across all surveyed sites included in this study across the entire State of California. Importantly, this variability contained ecosystem distinctions that were directly applicable to the characterization of conifer forests in the Southern Cascades supersection, thus accounting for more ecological variability than was achieved by averaging these forests together in a Mixed Conifer assessment area. Species and species combinations associated with primary axes of variability included lodgepole pine (Pinus contorta), a combination of ponderosa pine (Pinus ponderosa), sugar pine (Pinus lambertiana), white fir (Abies concolor), and incense cedar (hereafter denoted as PP/SP/WF/IC), redwood (Sequoia sempervirens), and Douglas fir (Pseudotsuga menziesii), with tanoak (Lithocarpus densiflorus) and Pacific madrone (Arbutus menziesii) correlated with both redwood and Douglas fir (Figure 2a). The dominance of Douglas fir forest types across the Southern Cascades supersection was clearly identified by the clustering algorithm (Figures 2 and 3). Depending on exact placement in PCA space, sites in this cluster may either be adjacent to the redwood forest type, or adjacent to the PP/SP/WF/IC forest type, which tended to occur further inland in the Southern Cascades region (Figures 2 and 3). This span represents an ecological continuum of species associations. The ability to distinguish between Douglas fir forests and PP/SP/WF/IC forests, as well as the ability to place Douglas fir forests on a continuum between redwood forests and PP/SP/WF/IC forests, provided a greater amount of ecological and geographical specificity than is provided by the current Mixed Conifer assessment area.

Details are in the caption following the image
(a) Forest Inventory and Analysis (FIA) sites are projected onto the first two principal component analysis (PCA) axes, which account for 6.5% of variability across all 2727 FIA sites and 53 tree species. Labeled vectors symbolize the contribution of tree species, listed here by their common name initials, to the placement of sites across these two axes. The species labels are (clockwise from top left): LP (lodgepole pine, Pinus contorta), ILO (interior live oak, Quercus wislizeni), R (redwood, Sequoia sempervirens), T (tanoak, Lithocarpus densiflorus), PM (Pacific madrone, Arbutus menziesii), DF (Douglas fir, Pseudotsuga menziesii), SP (sugar pine, Pinus lambertiana), IC (incense cedar, Calocedrus decurrens), PP (ponderosa pine, Pinus ponderosa), and WF (white fir, Abies concolor). Forest-type clusters are obtained using the first four PCA dimensions. (b) The geographical locations of California FIA sites are plotted onto a map of California. Colors correspond to clusters obtained from an unsupervised clustering algorithm based on the PCA results and can be used to link species associations in panel (a) to geographical locations in panel (b), and are for visualization purposes only, as the analysis did not rely on discrete classification. For the purposes of this analysis, the most important clusters are the cluster shown in orange, which corresponds with an abundance of ponderosa pine, sugar pine, white fir, and incense cedar (a combination referred to as PP/SP/WF/IC in this paper), the cluster shown in green, which corresponds with an abundance of Douglas fir, and the cluster shown in dark red, which corresponds with an abundance of redwood.
Details are in the caption following the image
(a) White diamonds represent carbon project sites projected into principal component analysis (PCA) space using species composition. Sites are labeled using anonymous site codes in PCA space, whereas colored points represent Forest Inventory and Analysis (FIA) sites as in Figure 1. The alphabetical site codes are assigned based on placement along the first PCA dimension, with Site A having the highest value of dimension 1 and Site J having the lowest value of dimension 1. (b) Carbon project site geographical locations are overlayed onto FIA site forest types, with the Southern Cascades supersection outlined in white. Sites are not labeled to protect landowner anonymity.

When PCA clustering results were plotted in geographical space, clear ecoregions emerged (Figure 2b). These ecoregions appeared to correspond to geographical factors such as distance from the coast, leeward/windward orientation on mountain ranges, and elevation (Figure 2b). The PCA axes therefore implicitly contained information regarding geographical regions, in a way that was arguably more relevant for forest growth than current supersections. For example, the PP/SP/WF/IC forest type was typically located further from the coast in mountainous areas, whereas the redwood forest type was always found on the coast, so the ecological continuum of redwood-adjacent versus PP/SP/WF/IC-adjacent Douglas fir can also be thought of as a spectrum between coastal-type Douglas fir forests versus more mountainous and arid Douglas fir forests (Figure 2b). Because of the clear geographical patterns and the importance of conifer species in the PCA, the PCA results appeared to be more influenced by geography than by management factors such as selective harvest, which would change the ratio of conifers to hardwoods (e.g., Berrill & Boston, 2019).

The Southern Cascades region spanned a broad continuum of these forest types, from those dominated by PP/SP/WF/IC in the east to those dominated by redwood-adjacent Douglas fir closer to the coast, but an ecological and geographical bias exists in the IFM projects that have received upfront credits (Figure 3). Badgley et al. (2022) noted that Southern Cascades Mixed Conifer project sites that have been issued upfront offset credits were overwhelmingly located in Douglas fir-dominated forests on the coastal edge of the Southern Cascades, a finding corroborated by these results (Figure 3). Our analysis revealed that these sites were moreover differentiated from other mixed conifer forests in this region due to the similarity in their ecology to that of a coastal redwood forest, indicated by how closely they fell to the redwood cluster in the species composition space (Figure 3a). Only one site in our sample fell firmly into the PP/SP/WF/IC category, far from the other sites in the species composition space (Figure 3a).

The ecological continuum identified here corresponded with a gradient in average carbon stocks, which was not reflected in the original common practice or baseline carbon values (Figure 4). The IFM projects that fell furthest away from the PP/SP/WF/IC forest type and closest to the coastal redwood forest type in terms of species composition had the highest recalculated common practice values (Figure 4). These values represent the carbon density of privately owned FIA sites with similar species compositions. This pattern is even stronger in the carbon density averages of corresponding publicly rather than privately owned FIA sites, indicating that biophysical potentials rather than industry practices drive this gradient (Figure 4). This gradient could also be seen in initial carbon measurements for IFM projects (Figure 4). However, the official common practice and baseline carbon values did not capture this gradient (Figure 4).

Details are in the caption following the image
Improved forest management project sites are arranged from the site with the greatest value in dimension 1 of the PCA (Site A) to the smallest value in dimension 1 (Site J). The average carbon stocks in standing live aboveground tree biomass of ecologically similar FIA-inventoried privately held forests (recalculated common practice; yellow bars) and publicly held forests (red bars) decrease along this spectrum of project sites, as do the inventoried carbon stocks (gray bars), whereas original common practice and baseline carbon stock values (dark teal and light teal bars, respectively) do not. Error bars represent 90% CIs obtained from a bootstrap with 1000 iterations. Original common practice values that are below the lower limits of the CIs of the corresponding recalculated common practice values are significantly underestimated (p < 0.05). This is the case for all projects except Site J. (Note that not all projects provided numbers for the standing live aboveground biomass component of their baseline carbon stocks).

The establishment of a large majority of upfront-credited Southern Cascades Mixed Conifer IFM projects in forests with ecological characteristics associated with greater carbon sequestration than the official common practice corroborated claims of adverse selection and a significant overestimate of additionality, however unintentional. The ecological characteristics associated with greater expected carbon stocks relative to the official common practice values (that is, coastal-adjacent Douglas fir forest characteristics) were the same ecological characteristics that were over-represented in the Southern Cascades Mixed Conifer IFM projects (Figures 3 and 4). As a result, common practice is significantly underestimated compared with our recalculated values (using a 90% CI of recalculated common practice) at all project sites in our study except for the one site that fell firmly into the PP/SP/WF/IC forest type (labeled Site J in Figures 3-5), at which common practice was overestimated (Figures 4 and 5). Because of the underestimation of common practice at nine out of 10 study sites, net over-crediting occurred over the sites included in this study (Figure 5). In addition, at six out of the 10 study sites, initial project carbon stocks were within or lower than a 90% CI of the recalculated common practice (Figure 5). Thus, at a majority of sites in this study, initial carbon stocks were not significantly greater than the ecologically informed common practice, and could therefore be thought of as nonadditional under this definition of additionality. Together, these results indicate that many credits issued in the initial reporting period for Southern Cascades Mixed Conifer IFM projects reflect carbon stocks that already existed due to geography and associated forest types, not additional carbon stocks generated from improved management.

Details are in the caption following the image
Additional carbon in standing live aboveground tree biomass (gray) is defined as initial inventoried carbon stocks in this carbon pool minus the baseline, or, when the baseline for this individual carbon pool was unavailable, common practice. Recalculated additional carbon (pink) uses the recalculated common practice. Error bars represent the range produced using the 90% CIs of the recalculated common practice, as shown in Figure 3. A 90% CI of recalculated additional carbon that is either above zero or encompasses zero indicates a lack of statistical significance (p < 0.05) additionality above the recalculated common practice. A negative value of recalculated additional carbon indicates that a project contains less carbon than the recalculated common practice. Panel (a) provides values per hectare; panel (b) provides values integrated over the area of each project.

Sites that did exceed the ecologically informed common practice carbon stocks tended to have carbon stocks comparable with those of ecologically similar public forests (Figure 4). If leakage is properly accounted for, and if these forests are maintained for climate resilience, then these carbon stocks can be thought of as additional relative to typical private management.

DISCUSSION AND CONCLUSIONS

Based on a common practice estimation method that takes into account ecological gradients, this study suggests overall over-crediting and adverse selection in the Southern Cascades Mixed Conifer Assessment area. For a majority of sites in this assessment area that received offset credits in their first crediting period, the data suggest that forest management had not significantly increased aboveground live biomass carbon stocks compared with privately owned forests with similar ecology. Even for sites that did contain greater carbon stocks than indicated by ecologically informed common practice, the data suggest additionality was overestimated in most cases. In regions with sharp geographical and ecological gradients, such as the Southern Cascades, a failure to account for these gradients appears to lead to nonrandom bias in common practice carbon stocks, consistent with concerns that current methods for establishing minimum baselines and resulting upfront offset credits do not correspond to net climate benefits. Furthermore, the current situation has resulted in projects being preferentially established in forest types with ecological characteristics associated with greater existing carbon stocks compared with current common practice values, thus not meeting the climate objectives of California's offset program.

Despite over-crediting, management on some sites has allowed carbon stocks to reach levels typical of public forests with comparable ecology, far greater than even the recalculated common practice values. If such results can be achieved without displacing timber production to other areas or increasing the risk of severe disturbance (e.g., wildfire), then such projects would result in additional carbon sequestration. Further research is needed to assess the net carbon benefits of such projects.

Regarding methodology, this work, contrasted with the work of Badgley et al. (2022), indicates that methods that account for geographical and ecological gradients rather than discrete forest-type categorizations may more accurately reflect carbon density differences. Badgley et al. (2022), using discrete forest-type categories that did account for more ecological variability than is captured by current assessment areas, did not estimate 100% over-crediting for any Southern Cascades Mixed Conifer site, in contrast with the results presented here. The methods developed and demonstrated here could most likely be replicated across forest regions, both to assess biases in IFM crediting, and to more accurately estimate common practice and baseline carbon stock values to increase the integrity of crediting associated with new projects going forward. Because a large majority of credited IFM carbon projects, and all projects administered under the California compliance offset market, are located in the USA, FIA data are available to support species-based methods. Therefore, if the methods developed in this paper were mediated through a user-friendly platform, they could be adopted in other regions. Using such methods, additionality can be assessed relative to typical management even in regions where sharp gradients in ecological characteristics pose challenges to common practice assessment, such as the West Coast of North America. We believe that using the methods outlined could significantly reduce over-crediting and adverse selection, which in turn would increase confidence in the use of forest management credits as an effective climate mitigation tool.

Future studies on the use of common practice and baseline carbon stocking values in the California cap-and-trade market may include the analysis of how to account for management techniques that increase disturbance resilience and decrease the risk of a large carbon loss (e.g., from severe wildfire), even if these techniques decrease carbon stocks. When combined with the research presented here, this work may achieve forest carbon projects that represent additional and long-lasting carbon sequestration.

ACKNOWLEDGMENTS

This work was funded by a grant to Environmental Defense Fund from the Bezos Earth Fund.

    CONFLICT OF INTEREST STATEMENT

    The authors declare no conflict of interest.

    DATA AVAILABILITY STATEMENT

    United States Forest Service Forest Inventory and Analysis (FIA) data on tree-level, plot-level, and condition-level for the State of California were used in this analysis, and these data are available via the FIA DataMart at https://apps.fs.usda.gov/fia/datamart/datamart.html by selecting a Data Type and State, and locating the CA_TREE, CA_PLOT, and CA_COND files. The application materials (including species composition, initial carbon stock, and common practice carbon numbers) for all carbon offset projects registered through the Climate Action Reserve can be found at https://thereserve2.apx.com/myModule/rpt/myrpt.asp?r=211, and specific files used in this analysis are the listing applications, initial reporting period reports, and attachments for Project IDs CAR1046, CAR1066, CAR1092, CAR1095, CAR1102, CAR1103, and CAR1174, and these projects can be located via the Advanced Search in the upper right corner of the projects table. The application materials (including species composition, initial carbon stock, and common practice carbon numbers) for all carbon offset projects registered through the American Carbon Registry can be found at https://acr2.apx.com/myModule/rpt/myrpt.asp?r=111, and specific files used in this analysis are the listing applications, initial reporting period reports, and attachments for Project IDs ACR262, ACR 282, ACR377, and ACR378, and these projects can be located via the Advanced Search in the upper right corner of the projects table. The compilation of forest carbon projects that have been credited through the California Air Resources Board is available from Badgely et al. (2021) in Zenodo at https://zenodo.org/record/4630684#.Y91m13bMLIU within the file named forest-offsets-database-v1.0.csv.