Evergrow Questions 9-12: Forestry Projects & Land

Background: I’m exploring an idea called Evergrow, and wrote up a list of 30 questions relating to the idea. I’m now going through and answering each of these questions. Answers to questions 1-3 are here. Answers to questions 4-8 are here. Answers to questions 9-12 are below.

Q9: How do reforestation projects create offsets? What is the process of converting a parcel of barren land into a unit of carbon offset? What is the relative financial efficiency of this method vs other methods of generating carbon offsets? Do we need to certify each defined tract of land separately? What is the frequency of which a carbon-offset-producing land asset can produce a unit of carbon offset? What happens if there’s a fire or other natural disaster? To what degree is eminent domain a factor?

Generally, forestry projects generate offsets in one of two ways: by (a) planting new trees, or (b) preventing deforestation. There are a number of independent certification and verification agencies that evaluate offset projects for both the compliance and voluntary markets. While each protocol is different, in general, a forestry project generates offsets in proportion to the amount of carbon the trees (newly planted or saved from deforestation) and other plants store. Because different species of plants store different amounts of carbon and grow at different rates, there can be large differences in the number of offsets generated by a given parcel of land.

There are two important concepts for evaluating offset projects in general, and forestry projects in particular: additionality, and permanence. Additionality is the idea that any carbon “offset” by a given project must be in addition to the carbon that would have ordinarily been sequestered by the project in a “business as usual” or baseline scenario. For example, if a plot of land already has trees on it, the forest’s owner can’t claim credit for the carbon that those trees might sequester in the future because this will naturally happen without any intervention. While this may seem obvious, its practical application can be difficult, particularly when it comes to avoided deforestation projects. For example, suppose a timber plantation owns 100 trees and has plans to harvest all of them next year. Instead, they commit to cutting down only 80 of them. Should they be able to get offsets for the remaining 20? Does this create a perverse incentive for landowners to earmark more trees for removal so as to then create a higher baseline for the award of offsets?

Permanence is the idea that any carbon offset by a project must be stored permanently. In the case of forestry projects where the trees will eventually be harvested, this means looking at the intended use of the harvested wood. Some wood products have a long or indefinite lifespan – e.g., the wood used in housing stock. Others, such as pulp and paper, have very short lifespans. Generally, only carbon that is captured in addition to the baseline model and stored permanently is eligible for a carbon offset under most protocols.

As discussed in a previous question, forestry projects represent the overwhelming majority of offsets awarded in the California market, and a large proportion of offsets in other markets as well. I haven’t seen a detailed comparison of the unit costs per offset of forestry projects compared to, e.g., livestock projects. That said, I believe the reason forestry projects contribute such a large share of offsets is because of their economies of scale. Beyond the one-time, upfront costs of land acquisition and planting, forestry projects are highly efficient, and their maintenance and verification costs can be spread over large areas of forest, resulting in a lower cost per unit of offset. For reference, there are forestry projects in the voluntary markets selling offsets for $1/mtCO2e, compared to, e.g., direct air capture, which isn’t profitable at even $100/mtCO2e at the time of writing.

In California, the ARB has published a detailed Forestry Projects protocol specifying the types of forestry projects eligible for generating offsets that can be used in the California market. There are three types of projects eligible for certification under the protocol:

  1. Reforestation: A reforestation project involves restoring tree cover on land that meets certain conditions laid out in the protocol. In particular, the land has to have had less than 10% tree cover for at least the last 10 years, or been subject to a “significant disturbance” (e.g., a fire) that has removed at least 20% of that land’s biomass. With few, limited exceptions (e.g., to prevent disease), reforestation projects may not engage in a harvest of reforested trees during the first 30 years of the project.
  2. Improved Forestry Management (“IFM”): An IFM project involves taking existing forests and introducing “natural forest management practices” that result in the sequestration of more carbon than usual. Examples of such practices include increasing the overall age of the forest by using longer rotation periods for harvests; increasing forest productivity by more active forest management; and increasing stocks of trees on understocked areas.
  3. Avoided Conversion: Avoided conversion projects involve the prevention of the conversion of forests into non-forest land use. For example, taking forestland earmarked for housing development and preserving it as forestland.

While each project type has a number of project-specific requirements, particularly around the calculation methodologies for carbon, they all share the same general requirements regarding crediting periods, reporting and verification, etc. In general, all projects must be located in the US, can be spread across multiple tracts of land, and require a project owner to commit to the protocol for 100-125 years. The lengthy commitment reflects the ARB’s desire to ensure that forestry offsets sequester carbon permanently. While 100-125 years is not literally “permanent”, this represents a compromise between the environmental imperative to sequester carbon literally forever and the commercial realities of project development and maintenance.

Because of their long life, forestry projects are at risk of “reversals” – i.e., the release of stored carbon back into the atmosphere. The ARB forestry protocol describes two kinds of reversals: unintentional, and intentional. Unintentional reversals are those that were not purposefully done – e.g., a fire or disease. To compensate for unintentional reversals, the ARB maintains a reserve account of offsets. Whenever a project experiences an unintentional reversal, the ARB “withdraws” a corresponding number of offsets from its reserve account and retires them. Project owners are responsible for contributing a portion of their offsets to the reserve account when their projects are first approved. The contribution amount varies by project depending on the risk of each project. Project owners are responsible for compensating ARB in the event of intentional reversals (e.g., harvests that are not allowed or go beyond the amount allowed under the terms of the protocol).

It’s unclear how eminent domain – or future land ownership transfers more broadly – might jeopardize offsets that have already been issued and/or retired. At a high level, using land to generate offsets and sequester carbon is unlikely to be considered a “blight” or unproductive under eminent domain doctrine (or at least what little of it I can remember from law school). However, the question of “who holds the bag on things that happen way out into the future” is an interesting one. In the case of a reversal or invalidation in the distant future, the ARB would likely look to either (or both) the buyer/user of a CCO and the current owner of a forest to make ARB whole. This strengthens the case for an ownership model that can credibly say it will be around to service the 100-125 year timeline of forestry projects.

Q10: Land acquisition: Does the land exist to do this? Where is it? How much does it cost to buy it? Where is this land located, and why? How does land in these locations compare to land located elsewhere in the US/world? What are the factors that determine how much carbon can be sequestered by trees on a given plot of land?

Below are the requirements for land for Evergrow. We need land that:

  1. Is located in the United States, in order to comply with the ARB Forest Projects protocol;
  2. Is not currently forested, i.e., has less than 10% tree cover, in order to use the ARB reforestation project type;
  3. Could naturally sustain a forest with a certain mix of species, i.e., the land and its climate is capable of sustaining a forest naturally once planted, and that such forest could contain a mix of certain species of trees that sequester an appropriate amount of carbon and yield marketable timber;
  4. Has reasonably foreseeable growth and survivability rates, i.e., that the land is of a type and quality that we can reasonably estimate how quickly the trees we plant will grow and how many will survive over time; and
  5. Is available at an appropriate cost.

Nice to haves:

  • Landowner willing to take a lease or provide seller financing, so that up-front land acquisition costs are minimized;
  • Land is not a candidate for housing development, so that value of the land is lower and competing bids on and uses for the land are few (if any);
  • Land has easy access for planting and maintenance, so that up-front land planting and ongoing land maintenance costs are lower, as well as the cost of eventual harvest; and
  • Larger parcels preferred, because of economies of scale.

There’s a lot here so let’s unpack it. For starters, identifying plots of land in the United States that are not currently forested but could theoretically support forests is doable via satellite imagery and machine learning. Most notably, a team of scientists published a now-famous paper in Science estimating the world’s tree-carrying potential using satellite imagery and machine learning models trained on conservation areas. The paper found that the United States could support an additional 103 million hectares of new forests, which is ample capacity for Evergrow.

Taking this model down a level is more difficult and to my knowledge, has not been done yet academically or commercially. Specifically, it would be useful to be able to take a satellite image of a plot of land and know the values for requirements #3 and #4 above, even if within a range or a low-ish confidence interval. My sense is this process is currently done manually by foresters and forestry consultants when evaluating specific projects. Being able to do it programatically would thus unlock the possibility of proactively going out to landowners who may be unaware of the value of their bare land as potential carbon sinks when converted into forests.

Ideally, this model would also incorporate local land sale data to estimate price, allowing a user to quickly calculate an appropriate trade-off between land quality and acquisition price. For example, Evergrow would happily pay more for land that is well suited for forest growth and survivability. I do not believe the market currently values bare land for its ability to sustain forests, meaning that Evergrow’s valuations may be much higher or lower than the market price of any given plot. For this reason, I believe significant arbitrage opportunities may exist within those 103 million ha that could sustain forests – i.e., there *must* be some plots of land that the market does not value highly (e.g., very remote, not desirable for housing, etc) but Evergrow does (e.g., very well suited for trees, great access for planting and maintenance, etc).

I haven’t built the model yet, so I can’t put numbers on these questions. However, here is how I plan to go about building it. To start, I think we need to supplement the model from the Nature article with specifics on tree species and local climates in the United States, ideally on a per hectare basis (or as granularly as possible). Ideally, we would use one of the methodologies accepted by ARB when estimating rates of tree growth for these purposes. This would give us a map of the United States, with each area marked for how much we would be willing to pay for a given plot of land within it. We could then cross-reference this against market data within each local area – e.g., pulled from state or county records – to determine where the most promising opportunities are. A final step would then be to contact individual landowners who own those plots to negotiate a sale or lease.

Land in the United States is generally going to be more expensive than land in the developing world, perhaps by an order of magnitude. So why then locate projects in the United States? The first and obvious reason is the relatively high price on carbon set by the California cap-and-trade system. The bet is that this high price more than pays for the higher costs. However, suppose theoretically that the voluntary markets had the same or higher price on carbon. Would this change our answer? In short, yes, but not as much as one might think. The other advantage provided by locating projects in the United States is the fact that the US has a strong legal system with well-established property rights, title ownership, and dispute resolution. By contrast, most developing nations have struggled to institute systems of property ownership, especially in rural or forested areas, leading to conflicts between investors, owners, operators, and local communities. Given that offset projects take place over indefinitely long timelines (ideally “forever” to ensure permanent sequestration), locating projects in jurisdictions with weak property rights systems seems like a recipe for disaster, as over a long enough time frame, the likelihood of “something” happening is high. So to answer the question – why the US? Because of the high price on carbon in California. But even if that constraint was lifted, we would still likely limit our development only to those jurisdictions with strong legal systems.

Q11: Land planting and maintenance: What does it cost to plant a plot of land? What is the “carrying”/maintenance cost of forestry? What risks are involved, controllable or otherwise, and how do we mitigate them? Who manages the forests – employees or contractors? Are there uncontrollable risks like stunted growth, forest fires, blight? Do we need to buy insurance, and if so, what does this cost?

Here is a link to a 2019 survey of small- and medium-sized foresters breaking out costs by activity on their forests. The major categories of costs are as follows:

  • Site preparation, at $370/ha;
  • Planting, at $680/ha; and
  • Seedlings, at $0.50/seedling, which at a density of 1,000 trees/ha is $500/ha

In addition, over time a site may need brush control and thinning, which cost $210/ha and $456/ha respectively and occur once every 3-10 years depending on site location and quality. Finally, the survey respondents reported an average cost of around $80/ha for general administration and maintenance. Assuming that a site requires no additional thinning or brush control beyond the 10-year mark prior to harvest, then the costs are as follows:

  • $1,550/ha upfront to prepare land, acquire seedlings, and plant;
  • $666/ha every 5 years for brush control and thinning; plus
  • $80/ha every year in general administrative and maintenance costs.

These numbers are directionally aligned with those in this US Forestry Service study which found tree establishment costs in the $1,000-2,000/ha range, albeit with significant variance in the data. This variance is common in the literature I’ve seen and is largely due to the factors listed in Q10 above. For example, the cost of planting is impacted by the cost of seedling (i.e., species of tree) being planted, the cost of preparing the site for planting (e.g., rocky sites need more pre-clearing than flat, bare land), and the accessibility of the site from main highways and logistics centers. To make matters more complicated, these variables interact with each other! For example, selecting an expensive seedling variety or buying saplings from a nursery will increase up-front costs but lower total costs over time if the trees end up being more suited to the environment in which they are planted, and thus grow faster, live longer, and/or require less maintenance. The fact that many of these risks can be known and quantified upfront puts an even greater emphasis on proper land and tree selection. Better to avoid a risk entirely than mitigate it! Over time, as Evergrow’s balance sheet gets bigger and operations get tighter, we can take more risk, and so the amount of land that becomes investable increases as well.

Once planted, the main risks to new forests are (a) failure of trees to grow at the rates expected, (b) failure of trees to survive, and/or (c) loss due to fire or disease. Survival rates of trees can be as low as 50% in some data that I’ve seen. I think the mitigation of both (a) and (b) comes down to effective site and species selection (again) and being conservative with modeling. On (c), most of the large timber owners self-insure; Weyerhauser’s 10-K even says that self-insurance is the norm in the forestry industry. I suspect this is because for balance sheets as large as Weyerhauser’s, their pool of forests is so large and geographically diverse that self-insuring makes sense. I do think obtaining some measure of standing timber insurance to guard against (c) makes sense for Evergrow, and costs appear to range from 0.5-2% of the value of the timber itself. Interestingly, losses due to (c) would likely be “unintentional reversals” under the ARB Forest Projects protocol, meaning that in such cases the prior offsets generated from the project would not be invalidated.

The physical on-site preparation, planting, and management of forests is done by trained foresters and laborers. Timber plantation owners seem to both outsource this to third-parties and in some cases employ their own teams in house. There are many forestry consulting companies that can manage forests on an outsourced basis, however, most of them appear to operate locally/sub-scale. In addition, I would expect these firms to be concentrated in areas of existing timber plantations, which may or may not be in the areas we end up developing into new forests.

Q12: Yield on land: How many offsets can we generate for a given plot of land? How much timber does that represent, and how much is it worth over time? What else can we do with the land that generates cash without harming the trees? E.g., water rights? People getting to buy individual trees and name them for 100 years? Recreational use/people buying memberships to go visit the forests?

These items are also sensitive to the factors listed in Q10 and Q11 above – namely, site location, species of trees, tree survival rate, etc. Directionally, here is what I have seen:

  • A hectare of forest land can support a density of anywhere from 500 to 2,000 trees depending on species and location. Given that CA requires reforestation projects to be naturally managed, I would expect our density to be toward the lower end of this range – e.g., in the 1,000 trees/ha range.
  • A Douglas Fir tree sequesters 14 imperial tons of carbon over its first 100 years of life. Assuming this happens linearly, that would imply 0.14 imperial tons per year, which is 0.142 mtCO2e/year.
  • At 1,000 trees/ha, that becomes roughly 140 mtCO2e/ha/year.
  • Regardless of actual carbon sequestered by the trees we plant, the ARB protocol provides for a detailed methodology for calculating how many offset credits are awarded for reforestation projects. At a high level, the methodology involves comparing the actual onsite carbon stored in the project against a baseline case. It is unclear to me whether or not this methodology is favorable or unfavorable to Evergrow, and I am reaching out to the ARB and consultants in the space to nail this down.
  • In addition, ARB requires that forest offset projects contribute up to 19% of their offsets to a buffer pool that the ARB maintains to compensate for unintended reversals. In practice, this can be much lower based on the actual risk profile of the project.
  • Putting it all together, I am conservatively estimating generating 50-100 CCOs/ha/year for Evergrow for now.

The value of the timber itself is a function of 4 variables:

  • The species of timber;
  • The size of the tree;
  • The quality of the timber; and
  • The prevailing price for timber of that type.

At the time of writing, Douglas Fir trees trade at around $700/thousand board feet (mbf), and yields are in the 30-40,000mbf/ha range. Taking the midpoint of this range, we would expect an Evergrow plot of 10,000 ha to have timber worth $24.5M in today’s dollars at harvest (50-75 years from planting). This equals roughly just $2,500/ha. Assuming the costs listed above in Q11, the rates of return on the investment are barely positive using timber alone, which I suspect is the primary reason commercial timber investors have not heavily invested in afforestation in the past. Put simply: it appears that what makes the model viable is getting paid both carbon sequestration while waiting for the trees to grow to maturity.

I think the concept of finding ancillary revenue streams from the forest is interesting and financially attractive. The challenge is in blending this into the Evergrow model, which seems to rely on identifying otherwise not valuable land. For example, BetterPlace offers consumers the opportunity to pay to be buried in a forest, and their forests are all driveable from where I live in San Francisco. I suspect that if these forests were located far away from where consumers live, they would be less attractive as burial plots, but likely more attractive to Evergrow (given that they would likely be cheaper to acquire). Water rights licensing is similarly double-edged, as my impression is that water rights investing is fairly mature now (see, e.g., the Texas Pacific Land Trust, a $3BN land trust that invests heavily in water rights) and so any excess yield might not compensate for the increased acquisition costs upfront. That said, I do think this is a worthwhile topic of further exploration in the context of diving deep into sites already identified by the Evergrow model as promising for development into forests (vs incorporating them into a model a priori as a reliable source of revenue). In particular, I think non-timber forestry products could be a meaningful source of near-term cashflow.

Another way to generate near-term cash flow from Evergrow (in addition to offset sales) is using its assets to sell and back financial products, specifically carbon insurance and timber futures. I’ve spoken with someone who once ran a large investment fund that invested in GHG abatement projects in the late 90s and early 00s. One challenge was the assumption of project-level risk; i.e., buyers of the GHG reduction credits were at risk that those credits may not materialize if the project failed, leaving them exposed to the spot price for those credits. Given that Evergrow would have a steady stream of new offsets, we could, in theory, sell insurance policies for abatement project developers. If the project succeeds, the developer gets their abatement and we keep the premium. If the project fails, we deliver offsets equal to the missing abatement; so long as we can evaluate project risk (a big “if”), then we are monetizing the spread between our cost of generating offsets and the risk premium earned by taking project risk off the table for developers. This has the added effect of giving Evergrow a wedge into the abatement finance market. Similarly, selling long-duration timber futures backed by real assets seems like it might be attractive, especially for industrial buyers of timber who value sustainable sourcing.

Sources & further reading:

Open questions/follow-up items:

  • Need to build a model for land acquisition
  • Need to better lock in the costs and yields on plots of land and fold into model
  • Need to do a deep dive into the US Forestry Service report including appendix
  • Need to reach out to carbon consultants and the authors of the Global Tree Restoration Potential article
  • Need to understand cash flow profile and crediting under ARB protocol, and explore ways to call it forward (e.g., securitization)