Brent Sohngen (Ohio State University; sohngen.@osu.edu); Justin Baker (North Carolina State University); Adam Daigneault (University of Maine), and Alice Favero (RTI, International)
Summary
- This article continues a dialogue that began when we wrote a comment to Nature about the CHARM model (our comment is available here), which was proposed by Peng et al. (2023) as a new approach to measure carbon emissions from wood harvesting.
- Tim Searchinger and Steven Berry have subsequently written a response to our comment that does not respond to the problems of their approach, but instead criticizes a different model – the Global Timber Model (GTM). This article addresses their criticisms of GTM (The comment and Searchinger and Berry’s response are both still under review).
- A key reason for the dispute lies in the vastly different world views of the GTM and CHARM modeling groups. We present and discuss these two world views.
- Searchinger and Berry have specific concerns for the structure of GTM, including its forward-looking nature and the fact that the forest management incentives in the model are driven by timber demand. We argue that forestry is inherently a forward-looking sector, which requires models to account for shifting forest rotations and long-term investments. Further, we note that focusing on timber harvesting is the right approach when evaluating carbon neutrality of wood products.
- Searchinger and Berry have specific concerns about the land supply elasticity, relying primarily on work by Ruben Lubowski. We lay out the assumptions that have been used in GTM over the years and the empirical rationale for those assumptions. The response of GTM to rental increases in the United States are not substantially different from the response in Lubowski et al. for a similar rental increase.
- Searchinger and Berry also level concerns about the demand price elasticity. The rationale for the unitary elasticity assumption used in GTM is based on the long-term (i.e., decadal) and aggregate nature of the demand function used in the model. More inelastic demand, on which we have run sensitivity analysis from time to time, induces the expected results: larger price increases over time and larger investments in forest resources. It does not alter the fundamental outcome of the model – a sustained increase in wood demand will encourage investments in wood resources that mitigate or reverse any new emissions from wood harvesting.
- We further illustrate the fallacy of the no human intervention counterfactual suggested by Searchinger and Berry.
Sections in this report
This turned out to be a longer response than we expected, so for easier reading, here are links to the sections in the report. These section headings should allow individuals who want to review our response to a specific issue raised by Searchinger and Berry without reading the entire document. Click here for a pdf of this document.
Approaches to Analyze the Carbon Implications of Wood Harvesting
No human activity counterfactual
Structural Bias Due to Parameter Extrapolation
Introduction
Recently, Tim Searchinger and Steven Berry have published a report through WRI making numerous claims about the carbon implications of wood harvesting. Their article is a response to our comment (available here) on the CHARM model, which we wrote and sent to Nature as a “Matters Arising” submission. That exchange is under review.
In their response to our criticisms of CHARM, Searchinger and Berry make several assertions and claims about the carbon implications of wood harvesting in general and the Global Timber Model (GTM) in particular. While our criticism of CHARM focuses on deficiencies in the approach taken by CHARM, Searchinger and Berry (the latter was not an author on the original CHARM paper, but apparently should have been?) take GTM to task. In this response to their response, we reiterate fundamental points about carbon accounting and forests, and we set the record straight on mistakes they have made in their interpretation of GTM and several of its parameters.
We start with the basic claim in Searchinger and Berry that the question of carbon neutrality in harvested wood products is a purely physical question – having nothing to do with economics. The carbon intensity of any product is a function of the inputs, management decisions, and technologies used to produce those products. The choice of technology, supplier, and other factors are all at their heart economic decisions. In forestry, the carbon intensity of wood is a function of its production process, including how a forest is managed. It just so happens that for forest products, a big chunk of the production process happens in nature and that part has an outsized influence on the carbon intensity of wood products.
We do not believe that there is any way to divorce the question of carbon neutrality in wood products from economics – especially natural resource economics. Models used to conduct the analysis or project future changes in forest sector carbon sinks and emissions sources must integrate natural processes (ecology, atmospheric exchange) with economic systems (markets and resource management dynamics). This brings us to the different world view of GTM (natural resource economics) and Searchinger and Berry’s CHARM model (purely physical). It turns out that this difference in world view has an outsized influence on the outcomes of the two approaches, so we start by describing this difference. We then tackle some of the claims Searchinger and Berry make about GTM.
World View
The easiest way to understand differences in how these two groups view wood consumption and forest management can be illustrated with a figure (figure 1), which displays key big picture differences in the Searchinger and Berry/CHARM world view (Panel A) versus GTM (Panel B). In the non-economic system proposed by Searchinger and Berry, the world begins with a forest stock and cuts through it, taking whatever it wants, whenever and wherever it wants. In their model forests are backfilled in harvest locations with a mechanism that creates forests without costs. But, given biological growth in forests, the new replacement forest has less carbon for much of the future (hence it is less green) and generates a net emission, as evidenced by the large carbon emission they derive from their calculations (Peng et al., 2023).
The GTM world in Panel B is more complicated and nuanced. There, the forest system is malleable over time. People and policy makers can change it to achieve different outcomes. People respond to market and policy incentives by making economically rational decisions considering timber, agricultural (land rents), and in some scenarios, biomass or carbon markets, or both. Several forest management systems are modeled to reflect different levels of management or production intensity. There are hundreds of management systems that could be modeled in the global forestry system, however, GTM puts them into four broad categories.
GTM defines Inaccessible land as forestland that is currently beyond the economic accessibility margin, as dictated by current forest or agricultural rents and access costs (dark green). Managed land reflects forest that has been harvested previously so has been accessed and thus has lower access and harvesting costs than inaccessible land. An example might include mixed hardwood systems in the Eastern U.S. that were historically harvested for timber production but are currently managed for multiple amenities and ecosystem services. Some converted inaccessible land will shift into agriculture if the relative rents in agricultural production outweigh long-term returns to forestland (the white area at the bottom left) and some will become managed (the lighter green area that encompasses an increasing amount of inaccessible forest over time).
Some managed forests will be more intensively managed over time through higher levels of investment or input use, leading to more timber or more carbon storage or more of both. We have shown the lighter green area as getting a darker shade over time to represent this process. The peach area represents plantation forest. These are heavily managed monoculture forest systems, planted with improved varieties over time, with both competition suppression and fertilization. They are intensively managed for wood products, and their shading darkens over time to represent increasing intensity of management and output.
Plantations in the southern U.S. and other parts of the world can strengthen regional forest carbon fluxes through more rapid growth (contributing to the annual sink), by relaxing pressure to harvest naturally regenerated forests, and by increasing C storage in HWP carbon pools (Puls et al., 2024a, 2024b; Tian et al., 2018). Plantations are an increasingly critical component of the global wood supply chain – in addition to productivity improvements relative to natural forests, plantations allow trees to be grown to a particular diameter class or size specification consistent with forest product manufacturing infrastructure.
When faced with increased demand for wood, the Searchinger and Berry model can do only one thing, take more wood. Without any economic rationale, they apply arbitrary accounting rules to dictate which forests this wood is taken from. Because the new forest has less carbon, more carbon emissions are the inevitable result from harvesting, which is the key result from Peng et al. (2023).
GTM in contrast adjusts harvests and production along multiple margins. First, many forests used for timber are managed in economic rotations, and this economic rotation is often shorter than the rotation that maximizes supply from the site (the Maximum Sustainable Yield rotation). If there is a long-term increase in wood demand that increases price growth, or a change in preferences that increases the relative demand of sawtimber (typically produced from longer rotations), GTM will start moving some forests from their economically optimal rotation under business-as-usual market conditions to an older economically optimal rotation so as to increase the supply of wood from the same site over time. European foresters figured this out in the 1700s as the industrial revolution, ship-building needs for growing empires, and deforestation to create agricultural land put stress on wood supplies (Heske, 1938; Rackham, 2020; Scorgie and Kennedy, 1996; Viitala, 2016). For a simple, single region model of the Southern US pine plantations, Sohngen and Sedjo (1998) showed this effect could be considerable for the supply of wood and the storage of carbon because older trees on average contain more carbon per hectare.
Second, GTM intensifies regeneration of existing forests. Foresters can increase the yield of timber by switching to varieties or species that grow quicker, suppressing competition, thinning, and fertilizing, to name a few activities. Such intensification of management will not have an immediate impact if demand increases today, but these productivity improvements can have a meaningful impact upon carbon fluxes and stocks over time. The shorter the rotation age, the faster the impact, and with many plantations achieving rotation ages of 10 years or less in the tropics, the impact on regional production possibilities and carbon can be quick.
Third, GTM can increase the forest area by planting or naturally regenerating new forests. Realistically, new forests can only be established in forest types that are managed. This activity in GTM leads to one of the big complaints by Searchinger and Berry, believing that we have over-estimated potential changes in overall forestland area in response to increased timber demand. More on that later.
Fourth, GTM can harvest old growth, inaccessible forests if economic conditions are such that the relative benefits of accessing and harvesting these forests outweighs the high costs. Unlike the other three actions, this one will lead to net carbon emissions, especially in the short-term, and this outcome over these harvests in GTM is most similar to the outcome for all forests in Searchinger and Berry. In GTM, this harvesting activity is cost-prohibitive and represents only a small portion of the world’s total annual timber harvesting. In the tropics, most inaccessible harvests cause the land to shift to agriculture, while in boreal and temperate regions, we assume this land will reforest.
Finally, GTM can also substitute products, switching between sawtimber and pulpwood/biomass products. GTM only keeps track of sawnwood, pulpwood, and wood for biomass energy, so substitutions happen at an aggregate level. Less or more consumption is also possible, depending on prices.
| Figure 1: Comparison of competing forest world views.
|
| Panel A: Searchinger and Berry
|
| Panel B: Global Timber Model
|
Approaches to Analyze the Carbon Implications of Wood Harvesting
Stand Level Approach: Although CHARM’s world view includes all the world’s forests (Figure 1), it treats each stand that is harvested as an independent event that is analyzed outside of the forest system in which that cut happened. CHARM then assesses the carbon flux after the wood removal on that site. Walker et al (2010) already showed us that this approach to measure carbon emission would always result in an estimate of an emission. CHARM simply counted harvests all over the world and added discounting to make the effect bigger.
Forest Level Approach: Other approaches based in sustainability science treat individual wood harvesting operations as part of a system where many hectares are managed over a large area to provide a consistent supply of wood over a long period of time. European foresters worked this out in the 17th to 19th centuries. Foresters in the US figured this out over the 20th century, as have Brazilians, Chileans, and countless others around the world. When the forest is viewed as a system, the carbon flux from harvesting is analyzed across a contiguous forest area that is managed for timber, not each site individually.
Harvests in most intensively managed plantation systems will turn out to be carbon neutral because growth generally equals or exceeds harvest emissions across a productive region. Less intensively managed systems around the world will have mixed results, with carbon increasing in some systems and declining in others. Cutting old growth forests will in probably all cases result in carbon emissions.
An additionality question is embedded in the forest-level approach. It counts current growth in forests that will be harvested in the future against those future harvests. This may be a safe assumption for certain systems –softwood or eucalyptus plantations in northern Europe, the US, Brazil, Chile, Australia, and New Zealand to name a few – because those systems are heavily managed for wood. Searchinger and Berry worry that if the boundaries are drawn wide enough, wood harvesting will always look carbon neutral if only because forests around the world are accumulating carbon despite wood harvests and deforestation (Friedlingstein et al., 2022). They have a point.
The forest level approach can identify forest systems that are likely contributing to net emissions (old growth harvests), and systems that are likely contributing to net sequestration (most managed forests), but the approach has some drawbacks.
Counterfactual approach: This method involves projecting future wood consumption in a baseline, and then assessing a deviation around that consumption pathway (e.g., increased demand, decreased demand, zero harvests). The baseline scenario is a projection of equilibrium wood quantities consumed, prices, and land use assuming income growth, and region-specific shifts in land rents consistent with regional assumptions about the demand for land in other uses, like agriculture and urban uses. When demand is increased or decreased marginally, the difference in carbon is the impact of those marginal units of increased/decreased wood consumption on carbon fluxes in the global forestry system – including changes in harvest emissions, slash carbon fluxes, HWP sinks, and changes to annual carbon sinks due to demand-induced shifts in harvests across regions, forest types, and age classes. This approach only attributes carbon changes to the harvest changes caused by the increase or decrease in demand, but not other changes in the system, such as carbon fertilization or forest aging in inaccessible forests (which are captured in the baseline). Notably, this approach does not raise the same additionality concern because the initial carbon is differenced from the estimates. This is the approach used in GTM.
Model Structure
In their response, Searchinger and Berry state “These structural assumptions dictate the model’s finding that increasing wood demand causes limited reductions in forest carbon.” The structural assumption they are referring to is the optimal control nature of GTM with rational expectations in the timber market. There are two parts of the structural elements of GTM that Searchinger and Berry do not like. First, they do not agree with the idea that forward looking actors will start adapting the current forest to future demand today by adjusting the age at which current trees are harvested (lengthening rotations to get more stock into future higher value periods), by cutting some old growth trees to move that land from no growth to positive growth phases, by planting trees, and by intensifying the management of those trees.
This complaint that GTM assumes people are forward-looking is levelled frequently at GTM. The GTM approach derives from extensive natural resource economics literature (e.g., Berck, 1979; Brazee and Mendelsohn, 1990, 1988; Cummings and Burt, 1969; Hotelling, 1931; Lyon and Sedjo, 1983; Mitra and Wan Jr, 1986; Salo and Tahvonen, 2002). The question Searchinger and Berry ask revolves around whether there is enough evidence that people behave this way to support using a model based on this assumption.
One piece of evidence derives from timber prices, which followed a Hotelling path by increasing 3-5% per year for over 150 years in the United States. Using the efficient market hypothesis, Johnson and Libecap (1980) showed that timber prices rose at an expected 4-5% annual rate of return from the mid-1800s to the early 1900s – a rate consistent with the theory of old growth depletion. They also presented evidence of landowners withholding timber from harvest to reserve timber for future production with higher prices. In the 20th century, prices continued to rise (Haynes, 2008) as remaining old growth forests on public lands in the western US were cut. Although prices have fluctuated substantially since the 1970s, including after a large reduction in federal timber supplies happened in the 1990s, real price growth has largely dissipated as replanted forests have become a more important part of the market (Berck, 1979; Mendelsohn and Sohngen, 2019).
Planting itself, we argue, is evidence of forward-looking behavior in forestry. FAO (2020) reports 293 million hectares of planted forests, equaling 7.1% of the world’s forest area. Since 1990, planted area has increased 4.1 million hectares per year, a considerable investment in future forests. Tree planting in the United States increased throughout the 20th century, and the area of planted forests continues to grow, albeit at a more modest rate in recent years (Oswalt et al., 2019).
However, Searchinger and Berry’s main worry with the forward-looking aspect of GTM is that if wood demand increases, the model is pre-destined to increase wood supply. This certainly is the response the model undertakes, consistent with all market models. Timber supply in GTM increases through regional harvest reallocations, or through additional harvesting of old growth, shifting to more productive species, changing rotation ages in planted forests, increasing plantation area, and increasing intensity of management in planted forests.
However, even as the supply of timber increases in GTM to meet rising demand, the carbon stock in forests and HWP pools does not always increase (Favero et al., 2023a). It turns out that the total area of forests and the C stock don’t always increase on net in every type of demand scenario. In all cases more wood demand will lead to increased investments in forests (a reasonable response to higher prices), but this does not always result in more carbon – and understanding the policy, environmental, market conditions that support carbon stock growth is the subject of many GTM papers (and published manuscripts using other, complementary forest sector models).
Second, Searchinger and Berry worry that GTM only focuses on wood product markets and not other reasons why people hold forests. They are correct. We could ignore more forests in our model that do not contribute to wood supply, but we model all forests using standard ecosystem approaches. GTM holds many of these forests in inaccessible types and uses rental functions and access cost functions to determine whether timber market conditions dictate that those forests will be accessed and potentially harvested. Under most demand scenarios, a large share of the currently inaccessible forests in GTM remain inaccessible even after 100 years of demand growth, in part because foresters have intensified management in forests that are managed. That is, the costs of increasing timber output through forest management intensification are lower than shifting harvests to inaccessible natural forests (the extensive production margin). This result mimics forest sector transitions that we observe in many regions of the world (the U.S., China, Brazil, Chile, Vietnam, to name a few) in which plantation investments and management intensification has expanded production possibilities for certain forest product types.
Developers of GTM recognize the many reasons forests are held by people, management institutions, and governments all over the world. Fortunately, many of these reasons do not lead to carbon emissions when people engage in them (e.g., harvesting non-timber forest products, hunting, hiking, or simply providing habitat). We agree that these motivations often create compelling reasons for people not to harvest forests that otherwise would be harvested if the timber criteria were all that mattered. In GTM, we control for this by calibrating our initial timber flows and timber prices to match FAO data and adjusting our accessible and inaccessible forest stocks.
If the research question is “how do current and future stand conditions influence local deer populations”, then GTM is unlikely to be deployed at all, or would have to undergo significant changes. However, GTM is well suited to address the question on carbon implications of harvesting wood products. It is just simply not clear how our model’s focus on the demand for and supply of wood is a weakness with respect to measuring carbon fluxes associated with timber harvesting. Lots of forests are never touched in GTM, so carbon fluxes there would not factor into an analysis of carbon flux from wood demand, nor would they factor into an analysis of how an increase in demand would affect fluxes.
Land Supply Elasticity
The biggest criticism Searchinger and Berry level at GTM is the land supply specification, in particular the choice of land supply elasticity parameter. The current choice of land supply elasticity parameter is inelastic, 0.3, meaning a 1% increase in forest rents will lead to an increase in forests of 0.3%. Searchinger and Berry claim this parameter was chosen incorrectly from Lubowski et al. (2006).
The original “GTM” published in 1999 (Sohngen et al., 1999) assumed land supply was perfectly inelastic. When we developed the model to integrate with DICE in order to conduct carbon sequestration analysis, we incorporated linear land supply functions to be able to model avoided deforestation, afforestation, and shifts in forest types (Sohngen and Mendelsohn, 2003). For the US, those land supply functions were calibrated to an initial elasticity of 0.25, based on several studies and discussions with some of those authors (Hardie and Parks, 1997; Plantinga et al., 1999; Stavins, 1999). We then adopted constant elasticity rental functions with the 0.25 elasticity estimate, which were used for a series of papers (Daigneault et al., 2012; Favero et al., 2020, 2017; Sohngen and Sedjo, 2006; Tavoni et al., 2007).
In 2011, we published a version of the model with an agricultural sector where we deployed a CET function to manage land competition within regions (S. Choi et al., 2011). To choose the elasticity parameter for the CET function, we relied on Ahmed et al. (2008) and Lubowski et al. (2006). Specifically, we used the revenue weighted CET calibrations in Ahmed et al., which presented elasticity values ranging from 0 to >1 depending on the period of analysis. The estimate was 0.3 for 10 years, and since the unit of time in the Choi et al. model was 10, we used 0.3.
One issue we have long identified with the original constant elasticity land supply specification in GTM was that we had no agricultural market. Land could be taken from agriculture with no penalty on agricultural commodity prices, and thus no adjustment in the real rents faced by forests. This issue was particularly significant under future scenarios with high demand for wood biomass for the energy sector as identified in Favero et al. (2020, 2017) and Favero and Mendelsohn (2014) To accommodate some elements of an agricultural market in GTM, without specifying the entire agricultural sector, we implemented a function in the model that shifted all rent functions upward as the aggregate area of land in forests increased. For this, we adopted an elasticity estimate of 0.3, which was consistent with the CET parameter we used in the forest and agricultural model. We also used the 0.3 parameter for the global rental function shifter. Under scenarios that increase rents for all forest types at once, the land supply elasticity for any land class in GTM is effectively smaller than 0.3, and smaller than the original 0.25 assumption.
This approach and assumption about the land supply elasticity parameter has been used in numerous papers, including (Baker et al., 2019; Daigneault et al., 2022; Favero et al., 2023b, 2023a; Kim et al., 2018; Sohngen et al., 2019). The Sohngen et al. (2019) paper in particular conducts a sensitivity analysis on the land supply parameter, with results indicating that model projections are more sensitive to forest growth parameters for key forest types than for the land supply elasticity.
In their assessment of this choice of land supply, Searchinger and Berry suggest that a land supply elasticity estimate of 0.3 is too high given the results of Lubowski et al. (2006). It is interesting to note that Searchinger and Berry do not make the complaint they level at GTM about timber as the only output on forestland in Lubowski et al., although timber is the only factor Lubowski et al. consider when calculating forest rents in their model. The outcome in Lubowski’s paper, however, does not square with the claim in Searchinger and Berry about the strongly inelastic estimate. For example, Lubowski et al. report an increase in forest area of 349 million acres in the United States, an 86% increase, for a $100 per acre subsidy. Assuming their subsidy is annual (Lubowski et al. do not state whether it’s one time or annual) and given initial forest rents of $17 per acre per year, this 488% increase in rents leads to an 86% increase in land area, implying an elasticity closer to 0.2, not the 0.004 that Searchinger and Berry claim as the truth.
We interpret the low elasticity estimate in Lubowski as an instantaneous, or short-term elasticity – ie., if forest rents double at the beginning of the year, how much land will switch to forests by the end of the year? It makes sense that landowners don’t switch land immediately when prices change given that short-term price fluctuations in forestry are often significant (see demand elasticity section below). A different question is what happens to land use if rents increase, and remain increased for the foreseeable future? This is the question GTM asks with its decadal time steps and rents calculated along a forward-looking pathway (i.e., the prices a forest harvested in 30 or 60 years will face). Although Lubowski et al. conduct a short-run estimate, they calculate a longer-run response that is more elastic and similar to the long-run response in GTM.
Two other analyses from the US have informed the choice of land supply parameter (S.-W. Choi et al., 2011; Sohngen and Brown, 2006). The Sohngen and Brown study is from a three-state region in the US South and breaks forests into planted pine, natural pine and upland hardwoods. If just planted pine rents increase 1%, the area of planted pine is predicted to increase 5.5%, a really high elasticity. If all forest rents increase 1%, then the average elasticity across the three types is 0.46. This higher elasticity is based on use of USDA Forest Inventory and Analysis data for forest areas rather than the USDA NRI data used in Lubowski et al. During the 1980s and 1990s, which is the period covered in the analysis, the USDA Forest Service estimates that 2.7 million acres per year were planted in the US compared to 1.5 million acres per year in the previous 20 -year period (Oswalt et al., 2019). Lubowski et al. of course do not separately account for planted forests and thus miss the planting response to higher rents.
One of the articles Searchinger and Berry take to task is Favero et al. (2020), which used a constant elasticity of land supply of 0.25, but did not use the global shifter described above. Searchinger and Berry seem aghast at the large, predicted increase in forest area – as if a billion more hectares of forest would be such a bad thing – so let’s have a closer look at the results. The RCP1.9 scenario in that paper represents a transformative global scale decarbonization effort. Favero et al. assume that forest biomass is used as part of the solution and make a prediction of the resulting demand for such biomass. Under their RCP1.9 scenario, forest rents in the US increase 415% relative to the baseline by 2100, and forest area increases 88%. This is exactly what Lubowski predicted would happen in the United States for nearly the same predicted increase in rents.
Globally, Favero et al. (2020) predict a billion new hectares in forests under the RCP1.9 scenario, which amounts to a 31% increase in forest area for the 415% increase in forest rents. This is not out of line with results for RCP1.9 in many of the IPCC scenarios at the time (IPCC, 2018), nor is it out of line with some of the studies examining the physical potential to increase forests globally (Bastin et al., 2019). The upshot of that scenario is that without substantial improvements in other technologies, if the world wants to meet an RCP1.9, we may have to find a way to live with a billion more hectares of forests and their carbon storage. Worse fates are probably possible.
Demand Elasticity
Searchinger and Berry also worry that GTM has assumed wood demand that is more elastic than it actually is. We assume demand elasticity is -1.0. We certainly are aware of the considerable analysis indicating that wood demand is more inelastic. We have even participated in publishing demand elasticity estimates of our own, in papers that have both extensive reviews of the literature and primary analysis of US regional demand elasticity (Daigneault et al., 2016). The range in the literature is wide, and it is accurate that the -1.0 used in GTM is at the top end of the range of estimates of demand elasticity in regional markets. In the Daigneault et al paper, we estimated short run demand elasticity of -0.25, and long run elasticity of -0.5.
So why do we persist using -1.0 in GTM when it would clearly be convenient to use a more inelastic estimate? We do it because our structural model of long-term global timber markets assumes a global aggregate demand function summed over 10 years. Nearly all of the existing estimates are for individual forest types or timber products in individual regions. Our assumption is that at an aggregate level, there is more potential over a decade for consumers to substitute wood supplied from the various forest types around the world than a single regional model would estimate for a single year.
One can imagine that forest product mills in a region have a limited ability to substitute across timber types in the short run. When faced with supply disruptions, they will accept substantially higher prices for logs delivered from more and more distant places to keep the mill running, such as to meet existing supply contracts. Yet over time, mills facing higher prices for inputs will adjust their production process to substitute other, cheaper inputs. Thus, even within a single region, or wood basket, demand will be more elastic if a longer period is considered.
For example, when the Northern Spotted Owl incident happened in the United States, wood supply in the US fell 15% (Wear and Murray, 2004), and stumpage prices in the Pacific Northwest shot up over 100% between 1991 and 1993 — the epitome of inelastic short-run demand. Yet ten years later, by 2002, real stumpage prices were lower than they had been in 1992. Within a decade, American consumers had substituted Canadian imports and Southern pine for Douglas-fir from the federal forests of the Pacific Northwest.
GTM does not try to capture these short-term adjustments in markets that will happen within a year or two of a supply disruption. Instead, GTM is designed to focus on longer-term adjustments that evolve over time in response to long-term demand or supply stimulus. Demand stimulus can be growth in biomass energy production, new products like cross-laminated timber, or large increases in income like happened in China in the early part of the 21st century. Supply stimulus can include climate change, carbon fertilization, forest investments, lower access or harvesting costs, etc.
Searchinger and Berry assume this more elastic demand function drives some of the results, in particular, that timber and pulpwood wood harvests decline 70% or so in some scenarios as those flows are shifted to biomass energy (Favero et al., 2023a). We are not clear what their point related to carbon costs is on this one because we count biomass energy production as an emission in GTM — it’s an emission from the forest to the atmosphere unless coupled with carbon capture and storage systems.
No human activity counterfactual
A key claim in Searchinger and Berry is that the reason deforestation counts as a carbon emission is because the deforested site is compared to the no-harvest counterfactual on the site. This is not correct. The UNFCCC counts emissions from deforestation when deforestation happens because an actual emission happened due to human actions. It is an emission. They don’t do this because the counterfactual over the next 5 or 10 years would be a standing forest. The next year, if the land remains in agriculture, only the agricultural emissions are counted. On the other hand, if the land clearing was a clear cut and forests start growing back, the UNFCCC counts the forests growing back as a sink because the forests are drawing carbon from the atmosphere as they grow. The rationale in national carbon accounts is as simple as that, no counterfactual needed.
Despite the simplicity, the emission and sequestration element of forestry does confuse many people in part because what society cares about is the net effect between emission from disturbances like land clearing for agriculture and timber harvesting and sequestration from regrowth. One can measure this net effect by measuring stocks at two different time periods and taking the difference, which is what countries that have managed forests in their UNFCCC accounts do.
The car example Searchinger and Berry propose is odd because it seems straightforward to measure emissions from a range of car types to assess the climate impact of driving. If a car is parked forever, then it does not generate any emissions. Forests, as mentioned, are more complicated because disturbances lead both to emissions and to sequestration. Disturbances can be anthropogenic (harvesting) or non-anthropogenic (fires). Regeneration can also be anthropogenic or non-anthropogenic. Establishing plantations, replanting after harvest, or planning for regeneration when harvesting are all forms of anthropogenic regeneration. Old-field succession as happened after widespread agricultural use in Europe, the US, and now Latin America is also arguably anthropogenic. Carbon fertilization and climate change are in this context considered non-anthropogenic stimuli. In UNFCCC accounts we care mostly about the anthropogenic factors affecting forests, while acknowledging the difficulties sometimes of separating the two.
Structural Bias Due to Parameter Extrapolation
Searchinger and Berry claim that GTM is not credible because we do not have parameter estimates for the entire world. They want not only “credible” econometric estimates for all parameters in the model, but more parameters to be able to separately model all classes of products – even toilet paper consumption in Germany.
We are absolutely in favor of developing econometric models of land supply for other regions as well as other parameters in the model. Any work along those lines can only make our work stronger. For instance, Jonah Busch and colleagues have built formidable models that do some of that work (Busch et al., 2015, 2012; Busch and Engelmann, 2017). We’ve compared our results and see some similarities and differences, with definite opportunities to improve (Roe et al., 2021).
However, Searchinger and Berry are setting an impossible standard for everyone on this one, including hundreds of models used by climate change scientists all over the world. We agree more empirical work needs to be done, and when empirical data are lacking, more model testing needs to happen. Further, model inter-comparison needs to happen. That type of work in forestry has been underway for decades (Kallio et al., 1987), and continues with more recent model inter-comparison efforts (Daigneault et al., 2022). The CHARM modelers would do well to join these efforts.
interacting Factors
Searchinger and Berry make an interesting point about climate change. They imply that most of the observed increase in carbon storage in forests globally (Friedlingstein et al., 2022) is due to carbon fertilization and climate change. We agree that carbon fertilization and climate change have had important impacts on forests (we’ve even modeled it extensively with GTM, see Favero et al., 2022, 2018; Sohngen et al., 2001; Tian et al., 2018, 2016). We also agree that it is important that more research be done to disentangle carbon fertilization, climate change, and forest management from observed wood volume and carbon stocks. On the management front, even more research is needed to better attribute advances in silviculture and fertilization, genetic improvement, and other interventions on productivity improvements over time.
In the United States, such work is underway. Davis et al. (2022) showed the carbon fertilization effect in US forests empirically across forests. In that study, planted stands and natural stands of the same species experienced the same proportional impact of carbon fertilization on wood volume. However, because of management, managed stands had more volume per hectare and thus received a larger absolute effect of carbon fertilization on wood volume. Thus, carbon fertilization has a much larger effect on a hectare of planted stands than it has on a hectare of natural stands. Between 1970 and 2015, Davis et al. calculated that natural loblolly stand would have gained 21.4 m3/ha while the planted stand would have gained 34.6 m3/ha, a 62% larger gain in volume.
It is difficult to cleanly disentangle the effects of management and carbon fertilization. At the margin, carbon fertilization provides an incentive for individual landowners to spend more resources than otherwise to increase their growing stock. At the same time, managed stands gain more from carbon fertilization.
The United States has long had a strong carbon sink in forests. Private stands in the southern US, a large share of them planted pine, have become the backbone of the carbon storage system in the US, even as it remains the wood basket. Forests remaining forests in the south stored 340 million tCO2 per year over the last decade, 68% of the US total (Domke et al., 2022), even as 60-90 million tCO2 were removed every year in the form of wood products (Oswalt et al., 2019). The reason for this is straightforward, intensive management has allowed the planted softwood system to supply large quantities of wood, while maintaining a strong balance sheet on carbon.
A key reason for this is that carbon fertilization has particularly strong effects on younger, improved stands that are managed. Proportionally, the effect of carbon fertilization is greater in younger stands than older ones, as well as in stands planted with improved varieties and managed to achieve larger stocks. Carbon fertilization will have a larger per hectare effect on biomass in young, planted stands than older natural stands.
So yes, carbon fertilization has helped the balance of carbon in the United States and elsewhere, in large part because it has such a potent effect on the types of forests we have today.
Conclusion
The question of whether wood-based materials are carbon neutral is a critical question for climate change policy and sustainability science. We argue that that models that integrate biology with economics are well suited to the task of determining whether wood produced in various locations and under different management regimes is carbon neutral. The CHARM model is not one of these models, as outlined in a comment on the original CHARM article submitted to Nature and under review. Many other models are better suited to the question of carbon neutrality, such as the Global Timber Model (GTM).
Tim Searchinger and Steven Berry raise numerous criticisms of GTM in their response to our proposed comment in Nature. Here, we address those comments and point out that their concerns are either based on an incorrect understanding of forestry, an incorrect understanding of the literature, an incorrect understanding of GTM, an incorrect understanding of resource economics, or all the above.
References
Ahmed, S.A., Hertel, T.W., Lubowski, R., 2008. Calibration of a land cover supply function using transition probabilities. GTAP Res. Memo. 14.
Baker, J.S., Wade, C.M., Sohngen, B.L., Ohrel, S., Fawcett, A.A., 2019. Potential complementarity between forest carbon sequestration incentives and biomass energy expansion. Energy Policy 126, 391–401.
Bastin, J.-F., Finegold, Y., Garcia, C., Mollicone, D., Rezende, M., Routh, D., Zohner, C.M., Crowther, T.W., 2019. The global tree restoration potential. Science 365, 76–79.
Berck, P., 1979. The economics of timber: a renewable resource in the long run. Bell J. Econ. 447–462.
Brazee, R., Mendelsohn, R., 1990. A dynamic model of timber markets. For. Sci. 36, 255–264.
Brazee, R., Mendelsohn, R., 1988. Timber harvesting with fluctuating prices. For. Sci. 34, 359–372.
Busch, J., Engelmann, J., 2017. Cost-effectiveness of reducing emissions from tropical deforestation, 2016–2050. Environ. Res. Lett. 13, 015001.
Busch, J., Ferretti-Gallon, K., Engelmann, J., Wright, M., Austin, K.G., Stolle, F., Turubanova, S., Potapov, P.V., Margono, B., Hansen, M.C., 2015. Reductions in emissions from deforestation from Indonesia’s moratorium on new oil palm, timber, and logging concessions. Proc. Natl. Acad. Sci. 112, 1328–1333.
Busch, J., Lubowski, R.N., Godoy, F., Steininger, M., Yusuf, A.A., Austin, K., Hewson, J., Juhn, D., Farid, M., Boltz, F., 2012. Structuring economic incentives to reduce emissions from deforestation within Indonesia. Proc. Natl. Acad. Sci. 109, 1062–1067.
Choi, S., Sohngen, B., Rose, S., Hertel, T., Golub, A., 2011. Total factor productivity change in agriculture and emissions from deforestation. Am. J. Agric. Econ. 93, 349–355.
Choi, S.-W., Sohngen, B., Alig, R., 2011. An assessment of the influence of bioenergy and marketed land amenity values on land uses in the Midwestern US. Ecol. Econ. 70, 713–720.
Cummings, R.G., Burt, O.R., 1969. The economics of production from natural resources: Note. Am. Econ. Rev. 59, 985–990.
Daigneault, A., Baker, J.S., Guo, J., Lauri, P., Favero, A., Forsell, N., Johnston, C., Ohrel, S.B., Sohngen, B., 2022. How the future of the global forest sink depends on timber demand, forest management, and carbon policies. Glob. Environ. Change 76, 102582.
Daigneault, A., Sohngen, B., Sedjo, R., 2012. Economic approach to assess the forest carbon implications of biomass energy. Environ. Sci. Technol. 46, 5664–5671.
Daigneault, A.J., Sohngen, B., Kim, S.J., 2016. Estimating welfare effects from supply shocks with dynamic factor demand models. For. Policy Econ. 73, 41–51.
Davis, E.C., Sohngen, B., Lewis, D.J., 2022. The effect of carbon fertilization on naturally regenerated and planted US forests. Nat. Commun. 13, 1–11.
Domke, G.M., Walters, B.F., Nowak, D.J., Smith, J., Ogle, S.M., Coulston, J.W., Greenfield, D.J., Nichols, M.C., Wirth, T.C., 2022. Greenhouse gas emissions and removals from forest land, woodlands, urban trees, and harvested wood products in the United States, 1990–2020. (No. FS-382), Resource Update. US Department of Agriculture: Forest Service, Northern Research Station, Madison, WI.
FAO, 2020. Global Forest Resources Assessment 2020 Main Report. United Nations Food and Agricultural Organization, Rome.
Favero, A., Baker, J., Sohngen, B., Daigneault, A., 2023a. Economic factors influence net carbon emissions of forest bioenergy expansion. Commun. Earth Environ. 4, 41.
Favero, A., Daigneault, A., Sohngen, B., 2020. Forests: Carbon sequestration, biomass energy, or both? Sci. Adv. 6, eaay6792.
Favero, A., Daigneault, A., Sohngen, B., Baker, J., 2023b. A system-wide assessment of forest biomass production, markets, and carbon. GCB Bioenergy 15, 154–165.
Favero, A., Mendelsohn, R., 2014. Using markets for woody biomass energy to sequester carbon in forests. J. Assoc. Environ. Resour. Econ. 1, 75–95.
Favero, A., Mendelsohn, R., Sohngen, B., 2018. Can the Global Forest Sector Survive 11° C Warming? Agric. Resour. Econ. Rev. 47, 388–413.
Favero, A., Mendelsohn, R., Sohngen, B., 2017. Using forests for climate mitigation: sequester carbon or produce woody biomass? Clim. Change 144, 195–206.
Favero, A., Sohngen, B., Hamilton, W.P., 2022. Climate change and timber in Latin America: Will the forestry sector flourish under climate change? For. Policy Econ. 135, 102657.
Friedlingstein, P., Jones, M.W., O’Sullivan, M., Andrew, R.M., Bakker, D.C., Hauck, J., Le Quéré, C., Peters, G.P., Peters, W., Pongratz, J., 2022. Global carbon budget 2021. Earth Syst. Sci. Data 14, 1917–2005.
Hardie, I.W., Parks, P.J., 1997. Land use with heterogeneous land quality: an application of an area base model. Am. J. Agric. Econ. 79, 299–310.
Haynes, R.W., 2008. Emergent lessons from a century of experience with Pacific Northwest timber markets, General Technical Report. US Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland.
Heske, F., 1938. German Forestry. Oxford University Press, Oxford.
Hotelling, H., 1931. The Economics of Exhaustible Resources. J. Polit. Econ. 39, 137–175. https://doi.org/10.1086/254195
IPCC, 2018. Global Warming of 1.5° C: An IPCC Special Report on the Impacts of Global Warming of 1.5° C Above Pre-industrial Levels and Related Global Greenhouse Gas Emission Pathways, in the Context of Strengthening the Global Response to the Threat of Climate Change, Sustainable Development, and Efforts to Eradicate Poverty. Intergovernmental Panel on Climate Change.
Johnson, R.N., Libecap, G.D., 1980. Efficient markets and Great Lakes timber: A conservation issue reexamined. Explor. Econ. Hist. 17, 372–385.
Kallio, M.J., Dykstra, D.P., Binkley, C.S., 1987. The global forest sector: an analytical perspective. John Wiley & Sons, Chichester.
Kim, S.J., Baker, J.S., Sohngen, B.L., Shell, M., 2018. Cumulative global forest carbon implications of regional bioenergy expansion policies. Resour. Energy Econ. 53, 198–219.
Lubowski, R.N., Plantinga, A.J., Stavins, R.N., 2006. Land-use change and carbon sinks: econometric estimation of the carbon sequestration supply function. J. Environ. Econ. Manag. 51, 135–152.
Lyon, K., Sedjo, R., 1983. An optimal control theory model to estimate the regional long-term supply of timber. For. Sci. 29, 798–812.
Mendelsohn, R., Sohngen, B., 2019. The Net Carbon Emissions from Historic Land Use and Land Use Change. J. For. Econ. 34.
Mitra, T., Wan Jr, H.Y., 1986. On the Faustmann solution to the forest management problem. J. Econ. Theory 40, 229–249.
Oswalt, S.N., Smith, W.B., Miles, P.D., Pugh, S.A., 2019. Forest resources of the United States, 2017: A technical document supporting the Forest Service 2020 RPA Assessment. Gen Tech Rep WO-97 Wash. DC US Dep. Agric. For. Serv. Wash. Off. 97.
Plantinga, A.J., Mauldin, T., Miller, D.J., 1999. An econometric analysis of the costs of sequestering carbon in forests. Am. J. Agric. Econ. 81, 812–824.
Puls, S.J., Cook, R.L., Baker, J.S., Rakestraw, J.L., 2024a. A Range of Management Strategies for Planted Pine Systems Yields Net Climate Benefits.
Puls, S.J., Cook, R.L., Baker, J.S., Rakestraw, J.L., Trlica, A., 2024b. Modeling wood product carbon flows in southern us pine plantations: implications for carbon storage. Carbon Balance Manag. 19, 8. https://doi.org/10.1186/s13021-024-00254-4
Rackham, O., 2020. Trees and Woodland in the British Landscape. Weidenfeld and Nicolson, London.
Roe, S., Streck, C., Beach, R., Busch, J., Chapman, M., Daioglou, V., Deppermann, A., Doelman, J., Emmet-Booth, J., Engelmann, J., 2021. Land-based measures to mitigate climate change: Potential and feasibility by country. Glob. Change Biol. 27, 6025–6058.
Salo, S., Tahvonen, O., 2002. On equilibrium cycles and normal forests in optimal harvesting of tree vintages. J. Environ. Econ. Manag. 44, 1–22.
Scorgie, M., Kennedy, J., 1996. Who discovered the Faustmann condition? Hist. Polit. Econ. 28, 77–80.
Sohngen, B., Brown, S., 2006. The influence of conversion of forest types on carbon sequestration and other ecosystem services in the South Central United States. Ecol. Econ. 57, 698–708.
Sohngen, B., Mendelsohn, R., 2003. An Optimal Control Model of Forest Carbon Sequestration. Am. J. Agric. Econ. 85, 448–457. https://doi.org/10.1111/1467-8276.00133
Sohngen, B., Mendelsohn, R., Sedjo, R., 2001. A global model of climate change impacts on timber markets. J. Agric. Resour. Econ. 326–343.
Sohngen, B., Mendelsohn, R., Sedjo, R., 1999. Forest Management, Conservation, and Global Timber Markets. Am. J. Agric. Econ. 81, 1–13. https://doi.org/10.2307/1244446
Sohngen, B., Salem, M.E., Baker, J.S., Shell, M.J., Kim, S.J., 2019. The Influence of Parametric Uncertainty on Projections of Forest Land Use, Carbon, and Markets. J. For. Econ. 34, 129–158.
Sohngen, B., Sedjo, R., 2006. Carbon Sequestration in Global Forests Under Different Carbon Price Regimes. Energy J. 27, 109–126.
Sohngen, B., Sedjo, R., 1998. A comparison of timber market models: Static simulation and optimal control approaches. For. Sci. 44, 24–36.
Stavins, R.N., 1999. The costs of carbon sequestration: a revealed-preference approach. Am. Econ. Rev. 89, 994–1009.
Tavoni, M., Sohngen, B., Bosetti, V., 2007. Forestry and the carbon market response to stabilize climate. Energy Policy 35, 5346–5353.
Tian, X., Sohngen, B., Baker, J., Ohrel, S., Fawcett, A.A., 2018. Will US forests continue to be a carbon sink? Land Econ. 94, 97–113.
Tian, X., Sohngen, B., Kim, J.B., Ohrel, S., Cole, J., 2016. Global climate change impacts on forests and markets. Environ. Res. Lett. 11, 035011.
Viitala, E.-J., 2016. Faustmann formula before Faustmann in German territorial states. For. Policy Econ. 65, 47–58.
Walker, T., Cardellichio, P., Colnes, A., Gunn, J., Kittler, B., Perschel, B., Guild, F., Recchia, C., Saah, D. and Initiative, N.C., 2010. Biomass sustainability and carbon policy study. Manomet Center for Conservation Sciences.
Wear, D.N., Murray, B.C., 2004. Federal timber restrictions, interregional spillovers, and the impact on US softwood markets. J. Environ. Econ. Manag. 47, 307–330.