香京julia种子在线播放

Badgley et al. (2022b) documented that most ARB projects define their baseline at, or very close to, the minimum level allowed by the protocol. For most projects the minimum allowed baseline is the regional average carbon stock density for the forest type. Badgley et al. found that many participating projects are composed of species with greater carbon stocks than the regional/forest type average as defined by the protocol. Because carbon stocks often change gradually over space but the minimum baseline is defined regionally, there is a strong incentive to enroll lands with naturally higher carbon stocks than the regional average. Badgley et al. estimated that this has led to over-crediting of close to 30% across the study's projects compared to what would have been credited if a more refined method was used to determine the minimum allowed baseline. Coffield et al. (2022) used remote sensing-based datasets to compare the outcomes of 37 California-based ARB IFM offset projects with similar “control” lands. They found lack of evidence that the offset program influenced land management and therefore lack of project additionality. van Kooten et al. (2015) investigated a large VCS IFM project in British Columbia that assumed a “lumber liquidator” counterfactual—that an alternative forest owner would have aggressively logged the forest. van Kooten et al. found that in this case, the chosen baseline created substantially more carbon credits than would have been generated if a more likely sustainable management scenario was used as the baseline.

Qualitative research also has consistently identified problem areas in baseline setting. Several studies identified asymmetric information as a pervasive, inherent problem in baseline setting for IFM projects. Asymmetric information creates uncertainty for the program administrator and third-party verifier but not the project developer, who implements a project with full information (van Kooten et al., 2009; Asante and Armstrong, 2016; Gren and Aklilu, 2016). For example, one study highlighted the trend of pulp timberland acquisitions by real estate investment trusts (REITs) and timber investment management organizations (TIMOs), who aggressively harvest and then sell the land to carbon project developers (Gifford, 2020). The project developers can report a low baseline carbon stocking as a result of the recent harvesting. This is an example of how a complex management history and asymmetric information make accurate baseline-setting difficult.

One study documented how program administrators deflated baselines in order to reduce barriers to entry in IFM projects. The study quoted one project developer stating that “if baselines are set too high, many potential projects will not be viable for participation” (Ruseva et al., 2017). To the best of our knowledge, Anderson et al. (2017) is alone in finding “strong evidence of additionality” of projects under ARB's IFM protocol and suggests that baseline/additionality criteria may be too strict and may impede projects with “multiple desirable features.” However, an expanded discussion by the authors suggested that they based their assessment on their observation that some rather than all projects are likely to be additional (Anderson and Perkins, 2017). Their survey of landowners with IFM projects showed that 5 of 17 (29%) self-reported that they were either not confident or unsure whether the offset credits generated by their projects “represent additional carbon sequestration that would not have happened without the forest offset program.”

Where good data on forest harvest exists, baseline uncertainty can potentially be reduced and conservativeness increased by developing baselines on historical practice, initial carbon stocks, similar lands with “dynamic” baselines, and NPV for landowners where NPV is reasonably predictive with some restrictions.

When NPV is used as the baseline, project developers should describe their capacity to harvest at this level and also the market conditions and mill capacity to absorb this harvest. Project developers wishing to use NPV can justify their case by demonstrating that they have a strong history of harvesting on similar lands, or better yet, can demonstrate a history of NPV harvesting on that project property. For projects that cannot demonstrate NPV-type harvest schedules, NPV is likely inappropriate.

Baselines that reflect current carbon stocking of the participating parcel are usually more conservative than broad regional averages. Such baselines only credit removals through growth.

When past management actions are used as baselines, statistical land use models can be used to provide quantitative estimates on the likelihood of harvest given a project's characteristics (Lewis, 2010). Such models can be used to create credible baselines and importantly, these models can be used to simulate alternative baselines which might reflect different market conditions (Radeloff et al., 2012).

The use of dynamic baselines is similar to control plots in experimental science. In this system, properties similar to the offset property in past management, market conditions, ecosystem, landowner type, etc., can be used as the baseline for offset projects. Matching methods developed for causal inference can be used to create comparison sets (Andam et al., 2008; Ferraro and Hanauer, 2014). Each year, the carbon values of the offset and the baseline properties can be compared, and credits can be issued on the basis of this comparison.

An advantage of dynamic baselines is that by observing similar properties in each year, changing market conditions can be integrated into baselines. For example, consider an offset in an area where mill capacity falls dramatically. Under static baselines, the offset would continue to generate credits, even though in reality there may be no market for timber in the area. Conversely, if a new technology increases the profit of harvesting, more credits could be granted. Dynamic baselines solve this problem by accurately reflecting baseline conditions relative to the project in pre-defined time periods. Such baselines might be particularly useful in areas where markets are in rapid flux, where forest managers cannot show that they have historically managed for NPV, or where land use is rapidly changing.

With all of these options, adverse selection might still lead to over-crediting. Because landowners or project developers will always know more than registries and verifiers about what would have happened without the offset income, adverse selection is a persistent issue. Statistically, adverse selection can be thought of as an unobserved variable that is correlated with the treatment decision (project enrollment) and the outcome (forest harvest). If this unobserved variable is correlated with increased enrollment and decreased forest harvest, the baseline is an overestimate of the true counterfactual. For example, this might be the case where a landowner has a strong conservation ethic and prefers to preserve rather than cut down their trees. A case like this can lead to over-crediting, because such a landowner is unlikely to harvest, even in the absence of the program.

Using historical forest harvest data can help to control conditions that lead to adverse selection, especially if these conditions do not change over time. For example, in the case of a conservation-minded landowner, if they have held similar preferences in the past, a baseline that takes into account their historical harvest levels would not over-credit (even though we cannot measure their land management philosophy). At the same time, a baseline based on regional averages or NPV alone would likely over-credit.

However, while historical baselines can help to account for unobserved variables that do not change over time, they cannot account for cases where the unobserved variable is not static. An example of this could be when a property is inherited or purchased by a new landowner. The application of a historical baseline for a property that had been harvested, but was purchased by a conservation NGO and then later enrolled in an offset program could lead to over-crediting because the true counterfactual for the new landowner is different than from the past landowner.

Dynamic baselines cannot directly account for the problem of adverse selection. To the extent that similar properties also have similar unobserved variables, then matching may reduce the impact of adverse selection. However, there is limited empirical evidence for this. Indeed, using nearby non-enrolled parcels as “control plots” could actually increase the effect of unobserved variables: if some parcels enroll and others do not, then it may precisely be an unobserved variable that is influencing this self-selection, biasing the dynamic baseline in favor of over-crediting.

All protocols have a mechanism for deducting leakage when timber harvesting is lower in a project relative to the baseline. All protocols use a leakage rate that reflects the assumed percent of onsite carbon loss (or gain) from a change in timber harvesting due to the offset projects that are lost (or gained) in other forests to which the harvesting is displaced.

ACR applies a 10% leakage rate if the project reduces harvesting by 5–25% compared to the baseline, and 40% if reduction in harvesting is more than 25% compared to the baseline. In the ARB, CAR-U.S., and CAR-Mexico protocols, leakage is deducted at a constant rate of 20%. Leakage rates used by all of the VCS protocols reviewed vary based on the carbon density, defined as the ratio of merchantable biomass to total biomass, of the forests where the displaced harvesting is assumed to occur compared to the forest enrolled in the carbon project. If harvesting is expected to shift to a forest with a ratio of merchantable biomass more than 15% lower than the project forest, a higher leakage rate (70%) is applied; if the destination forest produces more than 15% more merchantable biomass, relative to the project forest, a lower leakage rate applies (20%); if displacement occurs in a similar forest type, a 40% leakage rate is applied. VCS's extended rotation protocol (VM0003) also prescribes a 10% leakage rate if the rotation extension is < 10 years and the harvest reduction over this time frame is < 25%. VCS protocols exclude international leakage from their deduction formulas and allow for project-specific justifications for the application of a 0% leakage rate.

The academic literature has estimated forest carbon leakage using two general methods. Partial and general equilibrium models are complex optimization models based on economic theory of how markets function and calibrated to real-world data. Behavioral parameters, such as supply and demand elasticities, are drawn from the economic literature. These models are designed to capture the interconnectedness of different markets. General equilibrium models capture all economic flows within an economy, while partial equilibrium models usually focus in more detail on a subsection of the overall economy. Equilibrium models are generally used for ex-ante economic and policy analysis. Causal econometric models, which are an ex-post evaluation methodology that use statistical techniques to evaluate programs, have been utilized to assign causal attribution to leakage from other project types (e.g., Roopsind et al., 2019), but not IFM programs or projects. Challenges in applying causal inference methods to IFM include difficulty in observing a plausible harvesting counterfactual and the challenge of isolating program effects when so many IFM offset programs are currently being implemented with different rules.

Studies estimating leakage rates from reducing harvest activities have found a wide range of plausible leakage rates depending on different locations, spatial scales, time horizons, and methodological approaches. Some studies focused on national IFM programs (primarily in the United States), while others focused on global estimates. Studies in the United States context showed that leakage rates are generally higher than those commonly used in the protocols. In an econometric study of the effects of an 85% reduction in harvest on public lands in the Pacific Northwest of the United States during the 1990's, Wear and Murray (2004) found substantial evidence of output market leakage as softwood lumber prices increased by 15%. They estimated that nearly 84% of the timber harvest restriction shifted to unrestricted areas. Of that 84% leakage, they found that 43% in the region, 15% in other U.S. markets, and an additional 26% in Canadian markets. Using a general equilibrium model, Gan and McCarl (2007) estimated leakage rates from U.S. forest offset programs to be in the 75–78% range, including both domestic and global leakage.

One challenge in applying rates from the published literature to the protocols is that most, rather than quantifying leakage in units of carbon, estimate leakage of another metric like harvested wood products (Wear and Murray, 2004) or economic welfare (Gan and McCarl, 2007). Murray et al. (2004) and Murray et al. (2005) applied modeling frameworks that estimate carbon leakage directly. Murray et al. (2004) showed that domestic leakage rates (ignoring international leakage and focused on carbon instead of timber) for forest offset set-aside programs in the United States can vary from 16 to 68% depending on where the offset occurs in the country and carbon density of the protected forest. Murray et al. (2005) also conducted extensive carbon leakage analysis of forest sector carbon programs but did not focus explicitly on improved forest management is the focus of the protocols reviewed here.

Sun and Sohngen (2009) used a global economic optimization model and found that set-aside programs applied globally, which permanently reduced the land available for forest harvest, resulted in leakage rates of 47–52%, depending on the specific land taken out of production. Several studies in countries other than the United States showed significant variation in IFM leakage rates. Kallio and Solberg (2018) estimated leakage rates of 60–100% from harvest reduction projects in Norway. While the model had a relatively limited temporal and carbon accounting framework, it found that the variation in leakage rates is driven by the degree of harvest reduction, the type of forest product considered (e.g., pulpwood vs. sawlogs), and the forest product supply elasticity. By contrast, Sohngen and Brown (2004), estimated leakage rates of 2–38% for a Bolivian forest set-aside program. The country-to-country differences were likely driven by the country's integration into global wood product markets.

Based on findings from the literature and factors identified in Murray et al. (2004), leakage risk is likely to be highest in tight timber markets with responsive supply and in regions where non-participating land can produce similar timber products. One important caveat is that the economic equilibrium models used in the academic literature assumed that all actors have perfect information and as a result may slightly overestimate leakage risk in practice when markets are slower to adjust. More research is needed to update and refine understanding of leakage in IFM carbon projects. One particularly important area of future research is in leakage from short-term harvest deferrals.

There is variation in how the protocols consider market vs. activity leakage. CAR and ARB do not distinguish between market and activity leakage; any activity leakage is effectively included in the 20% market leakage rate. ACR and VCS monitor activity leakage separately. Under both of these registries, if production declines by more than 5% relative to the baseline, the landowner must demonstrate that no leakage occurs on other lands they manage or operate outside of the offset project. Landowners can demonstrate that no activity leakage occurs with historical harvesting records, or forest management plans prepared at least 2 years prior to the start of the project showing no change in harvesting on non-project lands with the implementation of the offset project. ACR includes a third option where landowners can demonstrate that they are not engaging in activity leakage if all lands owned by the landowner are certified as sustainable, such as by the Forest Stewardship Council (FSC).

These requirements prevent the most flagrant violations of activity leakage, but there are plausible cases when activity leakage might still occur. For example, a landowner could write a forest management plan with increased levels of harvesting and then enroll part of their lands in a carbon project 2 years later. As another example, FSC certification does not prevent any increase in harvesting, and thus activity leakage could easily occur on FSC-certified land. On the other hand, cumbersome activity leakage rules may prevent timberland owners from being able to enroll portions of their forest holdings as carbon projects due to the inability to manage unenrolled lands in response to changing wood product markets.

In addition to market leakage rates, the timing of the leakage deduction can have large effects on the number of credits issued. Prior research found that the ARB and CAR-U.S. protocols tend to greatly over-credit at the start of each project, due to a timing mismatch in the construction of the baseline scenario (Haya, 2019; Haya and Stewart, 2019). Most ARB IFM projects start with carbon stocks far above estimated baseline levels; initial carbon stocks 40–50% higher than baseline levels are typical (Haya, 2019). This is based on the assumption that without the offset program, timber would be aggressively harvested, reducing onsite carbon stocks substantially. This initial onsite carbon above the 100-year-average baseline is credited in the first reporting period, promptly generating a large number of credits without requiring any change in land management.

However, the displacement of harvesting (leakage) associated with that large reduction in harvesting is not all deducted in the project's 1st year, but rather is deducted evenly over the 100-year life of the project. This results in over-crediting at the start of the project, which is gradually paid back over the project life. We are not aware of any academic literature that has examined the correct timing of harvest displacements in timber markets. A conservative approach would apply the leakage deduction in the year that harvest was assumed to occur in the baseline and is credited by the project. Haya (2019) estimated that this correction would reduce the number of credits generated by the ARB protocol by 35%, and if the correction were combined with a higher leakage rate of 40–80%, crediting would be reduced by 51–82%. Levels of over-crediting would be even higher if reversals were not adequately monitored and compensated for after the end of the final reporting period in which credits were issued (Haya, 2019). The CAR-Mexico, ACR, and VCS protocols do not have this timing issue.

Leakage can also result in positive carbon outcomes when the project increases timber harvesting, thus leading to less harvesting elsewhere. None of the protocols account for reverse leakage from increased harvesting compared to the baseline, which is a form of conservativeness built into the protocols. Only the CAR protocols allow for reverse leakage to be counted if cumulative leakage from the project start is positive. While accounting for leakage annually is more conservative, cumulative leakage accounting may create more incentive for forest owners to decrease harvesting temporarily and conduct thinning to enable increases in harvesting later from an older, better managed forest.

Leakage is a complex economic phenomenon that is both hard to quantify and likely varies considerably across many dimensions, including IFM project type, location, and supply and demand conditions. The risk of over-crediting due to leakage would be reduced considerably if baselines were set more conservatively as described above. More conservative baselines that involve no or little difference in harvesting compared to the project would involve lower estimates of leakage, and so uncertainty in the leakage rate would have less impact on the number of credits generated.

ARB and CAR-U.S. protocols, which attribute leakage evenly over 100 years, are likely to over-credit significantly in the 1st year of each project that chooses a baseline lower than initial carbon stocks (which is the case for most projects). This source of over-crediting can be easily removed if leakage were deducted at the same time that the onsite benefits of reducing harvest are credited.

Current literature does not provide much guidance on the appropriate leakage rate to apply in specific contexts. Generally, the literature supports higher leakage rates than are currently used, although there are only a few studies that are mostly decades old and based on national or global economic equilibrium models or statistical evidence from large policy changes. For projects that reduce harvesting permanently, a higher leakage rate than those used by current protocols would be conservative given the large uncertainties. However, there is a risk that large, immediate leakage deductions may discourage extended rotation projects with only temporary leakage risk. This may be partially remedied without over-crediting by assuming leakage plays out over several years. This would strike a balance between the ARB and CAR-U.S. protocols (which average baseline harvesting, and therefore leakage deductions, over 100 years), and ACR, VCS, and CAR-Mexico protocols (which deduct leakage immediately). In addition, assessing leakage cumulatively would better reflect the impact of projects that defer rather than reduce harvesting. Currently only the CAR protocols credit projects for reverse leakage when increased harvesting compared to the baseline is likely to cause less harvesting elsewhere. These credits can be earned if cumulative emissions from leakage over the project lifetime are still positive. Lastly, discretion for projects to choose the leakage rate, as offered by all VCS protocols reviewed, has the potential to lead to under-counting leakage impacts.

Project terms are highly variable across protocols, but even the longest term (100 years) does not constitute a truly permanent offset equivalent to reducing fossil fuel emissions. Forest credits used to offset fossil fuel emissions convert carbon permanently stored as fossil fuels into carbon stored in trees in the short-term carbon cycle. If the end of a project term represents a reversal event, then non-permanent carbon storage (like all IFM projects) can more accurately be understood as delaying, not fully neutralizing, emissions (Herzog et al., 2003). Decisions about the appropriate duration of carbon storage fundamentally depend on assumptions about the future, and academics have called the default choice of 100 years “political” (Archer et al., 2009; Allen et al., 2016). In practice, project terms in IFM projects can range from 1 to 100 years, and there is not yet a widely adopted framework for comparing these different terms. Even taking for granted that these projects do not represent permanent offsets, questions remain about whether the current approach (relying on buffer pools) can achieve the promised durability.

Three key limitations of buffer pools could critically undermine their usefulness. First, none of the reviewed protocols take climate change into account in estimating buffer pool allocations and so may not reflect increasing risks of reversal over decadal time scales. For example, the ARB protocol for U.S.-based projects includes a buffer allocation of 2–4% for fire, 3% for biotic risks, and 3% for “other episodic catastrophic events” (e.g., drought). However, because annual acreage of forest fires in the United States is projected to quadruple by the end of the century even under a moderate emissions scenario (Anderegg et al., 2022), current buffer pool allocations may prove insufficient on the basis of wildfire risk alone. If recent wildfire trends continue in the United States, the entirety of the buffer pool for existing ARB projects will be consumed well before its intended lifetime is up (Badgley et al., 2022a). The ACR and VCS protocols have similarly low buffer allocations for natural disturbances, although no systematic assessment of these buffer pools have been conducted in the academic literature. A proposed VCS risk calculation tool may remedy this by using Climatic Impact Drivers (CIDs) to project increased risk.²

Second, some registries may not have a sufficiently diversified offset portfolio to effectively mitigate risk through the buffer pool mechanism. Such systemic risks may arise when a large proportion of projects in a registry are similar and/or exist in a constrained geographic area or ecological type. For example, the ARB compliance offset pool, which is composed mostly of IFM projects entirely in the United States (Badgley et al., 2022b), may be exposed to systemic forest risks that decrease the efficacy of the buffer pool as a risk mitigation tool.

Third, a buffer pool is defined by the quality of its constituent credits. Buffer pools composed of low quality credits have little value. Extensive work has shown systematic issues with additionality, baselines, leakage, and carbon accounting for land-based offset projects across protocols (e.g., Haya, 2019; West et al., 2020; Badgley et al., 2022b). Further, the ACR protocol allows project developers to put credits into the buffer pool from any ACR project (not just the project under consideration), which creates a perverse incentive to fill the buffer pool with low-value, potentially non-additional credits.

Broadly, climate change is expected to push forest systems toward younger, shorter, less carbon-dense forests (McDowell et al., 2020). These future forests are expected to have higher rates of mortality due to climate-exacerbated disturbances, making the carbon they store less durable (Anderegg et al., 2020; McDowell et al., 2020). Many types of disturbances are expected to increase in both frequency and severity. Offset registries should incorporate these increasing risks into the rules defining buffer pool allocations. If possible, reversal risk should be defined in a spatially explicit way to reflect the fact that different types of risks vary tremendously depending on the location, species composition, and stand structure (Anderegg et al., 2020). Further, existing protocols give minimal incentive to reduce disturbance hazards and could be updated to more actively reward management activities like prescribed burning, species selection, and thinning that increase resistance to reversals (Stephens et al., 2020; Herbert et al., 2022).

New time accounting frameworks have been proposed to clarify the value of shorter project terms. These fall into two broad categories: vertical and horizontal stacking of offset credits. Vertical stacking approaches, which include ton-year accounting like that used by the CAR-Mexico protocol, involve purchasing multiple short-term credits upfront to offset emitted CO₂. The multiple approaches to vertical stacking can have highly varied results depending on which assumptions are made (Levasseur et al., 2012; Groom and Venmans, 2022) and have been criticized for simply postponing climate impacts (Carton et al., 2021). Horizontal stacking, sometimes called offset rental or leasing, involves repeat purchasing of offset credits after they expire or after a reversal occurs (Herzog et al., 2003), which, if adequately enforced, could ameliorate some of the challenges of short durability terms.

The protocols employ standardized approaches to measurement of aboveground carbon stock changes. High level-guidance from the IPCC tends to distinguish between “stock change” vs. “flux” approaches to measuring carbon sources and sinks. While “flux” approaches measure GHG exchanges to and from forested systems, “stock change” approaches quantify carbon stocks across pools as well as the changes in them. The protocols that we reviewed primarily use stock change approaches, which include plot-based inventories with extrapolation to the project area, field measurement of trees, and use of allometric equations (which describe non-linear relationships between a tree's biomass and its more easily measured parameters, such as its height and/or diameter).

The protocols tend to provide appropriately rigorous, high-level guidance on inventory design under a stock change approach that aligns with recommendations from the IPCC (2019). Forest structure and composition (and thus aboveground biomass) can be highly variable. The protocols allow flexibility in carbon accounting such that project developers can adapt methods to local conditions and efficiently conduct monitoring, reporting, and verification. Protocols allow either permanent or temporary sample plots (ACR, ARB) as well as stratified random or systematic random plot designs (CAR-U.S.). Both approaches can produce unbiased and precise estimates of aboveground carbon stocks, but will depend on local forest structure and composition as well as the field inventory design used. IFM projects in regions with fewer relevant datasets may use less appropriate allometric equations and thus less robust estimates of aboveground biomass (Yuen et al., 2016). Depending on the methods used, overestimation of aboveground carbon stocks can occur (Clough et al., 2016), but this is likely to be less consequential to the overall validity of a forest carbon project than other considerations (e.g., baselines and leakage).

Methods for quantifying forest carbon stocks and their changes are rapidly evolving, including through the integration of field-based methods and remote sensing. Although challenges associated with accurately measuring changes in below-canopy forest structure for some remote sensing types (e.g., optical imagery) may limit their application to IFM projects (Asbeck and Frey, 2021), we expect technological advances to improve its future utility. However, a full discussion of these future opportunities is out of scope of this study, and we refer the reader to other reviews of the topic (Goetz and Dubayah, 2011; Xiao et al., 2019).

Belowground biomass refers to living roots, typically comprising 15–25% of total living biomass in a forest (Jackson et al., 1996). The belowground biomass pool does not include soil carbon, microbial carbon, or dead roots (although living roots contribute directly to each of these other pools via complex processes including root death, root exudates, and interactions of mycorrhizal fungi). Belowground biomass estimation models vary widely across protocols. Because empirical measurement of belowground biomass is difficult and time-consuming (requiring excavating, cleaning, sorting, and weighing roots), belowground biomass is estimated indirectly based on aboveground biomass measurements. The IFM protocols estimate belowground biomass using allometric equations or root:shoot ratios, which are inherently unable to capture detailed natural variation and, additionally, may introduce systematic errors by being inappropriately matched to the system in question (Ledo et al., 2018). Root:shoot ratios assume that belowground biomass occurs in a fixed ratio to aboveground biomass, whereas allometric equations allow for non-linear relationships.

VCS protocols tend to provide the greatest flexibility in ratio selection for belowground biomass estimation. VCS establishes basic criteria for eligible models, including peer-review, appropriate parameterization, and consistency with the original scope of the study. Regions with more abundant literature documenting root:shoot ratios enable developers to select estimates that produce the greatest number of credits. For example, VM0003 allows for use of the standard root:shoot ratios cited in Cairns et al. (1997), or any root:shoot value from research literature or national inventories with comparable climate and forest type. VM0012 is more stringent, requiring the use of the Cairns et al. ratios unless project-specific measurements have been taken. VM0010 is the only protocol that excludes belowground biomass entirely.

Both CAR and ARB require that projects in Washington, California, and Oregon use the Cairns et al. ratios. For other contiguous states, CAR and ARB protocols provide region-specific component ratio methods (which further divide aboveground and belowground biomass into subcompartments). ACR requires use of USFS merchantable volume equations tailored for region and species, which are then extrapolated to belowground biomass using ratios in Jenkins et al. (2003).

Because relatively little empirical belowground biomass data exists for validating either the allometric or root:shoot ratio approaches, it is not well-understood which of these approaches is preferable, what magnitude of error they may introduce, and whether they systematically over- or underestimate belowground biomass according to vegetation type, region, or climate regime (Xing et al., 2019). Across protocols, the Cairns et al. (1997) and Jenkins et al. (2003) reviews underpin nearly all belowground biomass estimates in IFM projects. Efforts to “spot-check” the validity of these simple modeling approaches have sometimes revealed large errors: for example, Xing et al. (2019) used empirical data to reveal that a root:shoot ratio approach overestimated belowground biomass in a Canadian poplar forest by between 18 and 42%.

Soils comprise 56% of the carbon stock within managed ecosystems across the United States, and 80% of the terrestrial carbon pool globally (Lal, 2008; Domke et al., 2017). IFM protocols rarely require the measurement or estimation of soil organic carbon (SOC) stocks and fluxes due to the assumption that changes in the soil pool are negligible relative to credit volumes and due to the considerable expense and logistical challenge of measuring the soil carbon stock accurately and comprehensively (Paustian et al., 2019). ACR and VCS IFM protocols fail to account for advances in soil science, and potentially omit declines in SOC caused by certain IFM practices. In some instances this omission could enable over-crediting by neglecting substantial losses in soil organic matter that are likely not recuperated during the crediting period (Johnson and Curtis, 2001; Jandl et al., 2007; Noormets et al., 2015; Johnson and Henderson, 2018). A growing body of literature indicates that site preparation and ongoing management can cause significant disturbance to soil stocks, especially in litter, organic, and topsoil carbon pools, partially eroding the benefits of biomass stock increases (Jandl et al., 2007; Achat et al., 2015). In the crediting context, the primary consideration should be whether soil disturbance and SOC stock declines under IFM exceed the baseline.

Some IFM practices, such as extended rotation and retention of coarse woody material, are unlikely to yield significant or persistent changes in the soil carbon stock, and may prevent SOC losses that may have occurred under the baseline (Mayer et al., 2020). In contrast, mechanical site preparation, such as thinning, planting, removal of brush or shrubbery, or partial harvesting, may have significant and long lasting negative impacts on the SOC pool (Walmsley and Godbold, 2010; Zhang et al., 2018). The CAR and ARB protocols most appropriately and conservatively include these fluxes by requiring that projects with site preparation, harvesting, or treatment (deep ripping, furrowing, or plowing where soil disturbance exceeds 25% of project area or is not done on contours) estimate the loss of soil carbon as a product of biomass removal, mineral soil exposure, and frequency of disturbance. Estimated carbon stocks and losses are calculated using predetermined coefficients, which are determined by the soil order, harvesting intensity, disturbance frequency, site treatment, and tree type composition.

This is aligned with a growing body of evidence demonstrating that harvesting can yield losses between 8 and 11% in the top meter of soil (James and Harrison, 2016). Similarly, thinning and removal of dead biomass reduce organic matter inputs, compact topsoil, mix soil layers, and reduce the total SOC stock (Mayer et al., 2020; Kaarakka et al., 2021). These impacts are most substantial in the organic layer and topsoil (0–10 cm) even under conventional thinning practices, demonstrating losses of ~25 and 5% of total SOC stock 10 years after management, respectively (Achat et al., 2015). SOC stocks are not homogenous and can be considered relatively recalcitrant or labile depending on the degree to which the carbon is mineral-associated or particle-associated organic matter (Lavallee et al., 2020). On average, the top 20 cm of forest soils in the United States contain ~230 tCO₂/ha (Cao et al., 2019), thus a loss of 15% of this stock across only 20% of the project area may reduce total project credits on the order of 7 tCO₂/ha. For context, across the 74 projects reviewed by Badgley et al. (2022b), credit issuances averaged 73 tCO₂/ha, implying an average project could over-credit by 10% or more without violating CAR or ARB SOC stock estimation requirements. However, this is only relevant to crediting outcomes if the SOC stock under IFM declines more substantially than the baseline, which is unlikely in projects that involve a reduction in harvesting.

Only CAR and ARB allow for the inclusion of the SOC pool, and require it if the stock is likely to decline due to site preparation disturbances or other management activities. Appropriately, none of the IFM protocols include an option for additional crediting from increases in SOC. All VCS and ACR IFM protocols presume that impacts on soil carbon would be negligible or positive relative to the baseline. To rigorously incorporate the impact of SOC losses within IFM projects, protocols would need to quantify not only the impact of project management practices, but also the alternative impact to the soil carbon stock under the baseline scenario.

The harvest of biomass for use in wood products is included in all reviewed IFM protocols with the exception of the CAR-Mexico protocol, whose projects are not expected to significantly alter the production of wood products. The ARB, ACR, CAR-U.S., and VCS protocols all offer detailed methodologies for estimating the carbon stock stored in wood products. The methodologies require an estimate of the carbon stock for both baseline and project HWPs. In general, they follow a similar process where project proponents must estimate (a) the volume of timber removed in the project and baseline scenarios, (b) the merchantable carbon in these HWPs, the carbon loss due to mill processing, and (c) the decay of HWP carbon in final products and landfills over a 100 year horizon. This decay rate varies based on the lifetime of the product category.

For example, in ARB and CAR-U.S. projects, carbon in HWPs is annualized across a 100-year decay function to generate a HWP “storage factor.” This means that each year, carbon flowing into the HWP pool is immediately discounted to its 100-year average value. In other words, a large portion of carbon reduced in the forest as a result of harvesting is assumed to instantaneously decay. Since much of that carbon is actually released over decades rather than immediately, for the first 50 years of the project, if the project harvests less than that projected in the baseline scenario carbon, which is the case for most IFM projects, benefits and credits are overestimated. ACR and VCS protocols use similar “storage factor” approaches for estimating carbon in HWPs.

All of the protocols we reviewed exaggerate the emissions associated with the production of HWPs by ignoring their displacement of other fossil-intensive alternatives. Substitution benefits are typically high for construction-based materials, such as steel or concrete (Smyth et al., 2017; Geng et al., 2019) and vary widely for energy products, such as biomass used to generate electricity and heat, based on the product displaced (Cabiyo et al., 2021). Ignoring these benefits results in some over-crediting and also shifts protocol incentives toward projects that reduce harvesting.

The accuracy and precision of estimating forest carbon stocks within IFM protocols should improve over time as measurement technologies, inventories, allometric equations, and root:shoot ratios improve. IFM protocols generally provide appropriate selection criteria for plot distribution, measurement, and carbon stock estimation and distribution methods. The accuracy of a given site's carbon stock estimate is likely to be most significantly impacted by the availability of regionally tailored and species-specific allometric equations and root:shoot ratios to approximate the impact of IFM practices on biomass distribution. Accounting for carbon in harvested wood products is more straightforward than estimating carbon in the ecosystem, and unnecessary over-crediting in the early decades of a project could easily be avoided by modeling HWPs in a temporally realistic way instead of immediately discounting them to their 100-year “storage factor.” Lastly, the protocols should account for potentially significant and lasting losses in soil carbon pools as a result of disruptive site preparation and management methods. While CAR and ARB have already incorporated literature-driven methods to account for reductions in the soil carbon stock of a project, more research is needed to understand how specific practices, species, and soil types respond to interventions.