Measuring Regeneration

Volume IV — Production and Regeneration

THE PLATE

A watercolour of a single tree on orange ground, a young tree growing beside it. — Plate IV.10The Same Hole, Twice.Everything expensive about proving that land is recovering comes from not knowing exactly where you stood the first time.

THE LETTER

You have read, by now, a good deal about why regeneration works. This chapter is about what it costs to prove it, and the answer is a number most of the literature declines to print.

The reason it declines is not evasion. It is that the cost of verification is not a property of the practice — it is a property of the ratio between the signal you are trying to see and the noise it is buried in, and that ratio is brutal in soil. A field that is genuinely recovering gains carbon at a rate of under one percent of its standing stock per year, in a medium whose point-to-point variation within a single hectare is routinely twenty percent. You are trying to hear a whisper in a room where everybody is talking.

So this chapter treats measurement as an engineering problem with a budget. There is a formula for how many cores you need, a formula for how often you may usefully take them, and a single ratio — verification cost over the value of the thing verified — that decides whether a claim is worth making at all. We will call that ratio kappa, we will compute it, and we will find that for the majority of the world's farms it lands somewhere between thirty and three hundred. That is not a rounding problem. It means a great many honest regeneration claims cannot be proven at any honest price by the party making them.

Then we will do the useful part, which is to show the three structures that move kappa, what each one actually moves, and where each one stops. All three exist already and all three have precedents you can look up.

Chapter II.10 established that a measure is an intervention rather than an observation, and everything it says about gaming applies here in full. This chapter takes the prior step: before you can worry about what a measure does to behaviour, you have to be able to afford the measure. Chapter IV.03 carries the agronomy of soil carbon itself. This one carries the instrument.

— The Editors

DISCOVERY

What is already working

Verification has a better record than its reputation, and the best examples are the ones that solved the cost problem decades before anyone framed it as one.

Rothamsted kept the samples. The Classical Experiments at Rothamsted have run continuously on the same plots since 1843, and from the beginning the station archived physical soil and crop samples rather than only the numbers derived from them. That decision, taken by people who could not have known what would later be measurable, means that questions nobody had thought to ask in 1843 — atmospheric deposition, radionuclides, organic pollutants, carbon fractions — can be asked of nineteenth-century soil today. The archive is the cheapest scientific instrument ever built, because its marginal cost per future question is approximately zero. The lesson generalises exactly: the expensive part of measurement is the visit, and the cheap part is keeping what the visit produced.

Europe built one network instead of twenty-seven. LUCAS Soil is a harmonised, repeated topsoil survey across the European Union, sampled on a common protocol at points that are revisited on a multi-year cycle, with the analyses run to one method rather than to each member state's own. A national soil inventory built country by country would have cost more and compared to nothing. One network, sampled sparsely but identically, compares to everything. It is aggregation applied to measurement itself, and it is the direct ancestor of the structure this chapter ends on.

A regulator accepted a molecule as legal evidence. In England, the survey standard for the presence of great crested newts — a strictly protected species whose presence can stop a development — is now an environmental DNA water sample, following the validation work of Biggs and colleagues published in 2015. Before that, presence was established by repeated night torch surveys and bottle trapping across a season by a licensed ecologist. A statutory authority replaced a month of skilled field labour with a bottle of pond water, and did so by first quantifying the detection probability of the new method against the old. That sequence — quantify the error, then swap the instrument — is the whole of tiered assurance, arrived at in one species.

A satellite measures forests for nothing. GEDI, the Global Ecosystem Dynamics Investigation, has been firing lidar at the Earth's vegetation from the International Space Station since 2018 and publishing footprint-level and gridded biomass products openly. Its mission requirement is stated as an error bound rather than a value — one-kilometre cell mean aboveground biomass density within 20 Mg per hectare or 20 percent — which is the honest way to publish a measurement and is rarer than it should be. Anybody with a laptop now has a biomass instrument whose marginal cost per hectare is zero and whose error bar is written on the tin.

Sixty thousand smallholders were verified as one. The Kenya Agricultural Carbon Project, developed with Vi Agroforestry under the World Bank's BioCarbon Fund, brought smallholder farmers into a single registered activity on sustainable agricultural land management, with the sampling, the modelling and the third-party verification carried at the level of the pool rather than the farm. No individual holding in that project could have borne a verification cost. The pool could. Plan Vivo has run the same logic for community-scale projects since the 1990s, and the Verra standard's grouped-project provisions now make the structure available generically.

Five cases, one pattern, and it is the pattern the arithmetic below will produce independently: the cost of a credible claim is dominated by fixed costs and by the precision of relocation, and both of those are shared goods. Nobody in the list above made measurement cheaper by inventing a better sensor. They made it cheaper by returning to the same place, by harmonising the method, by publishing the error bar, and by spreading one audit over many holdings.

THE ARITHMETIC

What works, what does not, and where the line sits

Everything in this movement is computed in lib/verify/IV_10.py and can be re-run.

First, the signal and the noise.

Take a temperate cropland soil holding 40 t C/ha in the top thirty centimetres, with a within-field coefficient of variation of 0.20, so a standard deviation of 8.00 t C/ha. Improved management accrues carbon at around 0.30 t C/ha/yr. Over a five-year claim period that is 1.50 t C/ha — an effect equal to 3.75 percent of the standing stock, to be detected against a standard deviation five times its own size.

Second, how many cores. For a two-sided test at 0.05 with 80 percent power, k = (z₀.₉₇₅ + z₀.₈₀)² = 7.848880. Sampling independently at each date:

  n = 2 k σ² / Δ²  =  2 × 7.848880 × 8.00² / 1.50²  =  446.51  →  447 cores

Four hundred and forty-seven cores. Per date. That is the number that has quietly killed more soil carbon projects than any policy.

Now sample the same georeferenced points at both dates. The variance of the paired difference is σ_d² = 2σ²(1−ρ), where ρ is the correlation between the two visits — high in soil, because most of the variance in a field is spatial rather than temporal and a relocated point carries its own spatial term forward.

  ρ = 0.50   σ_d² = 64.00   n = 223.26  →  224 cores
  ρ = 0.70   σ_d² = 38.40   n = 133.95  →  134 cores
  ρ = 0.90   σ_d² = 12.80   n =  44.65  →   45 cores
  ρ = 0.95   σ_d²  =  6.40  n =  22.33  →   23 cores

At ρ = 0.90, forty-five cores, which is 9.93 times fewer than the unpaired design. This is the chapter's instrument and its first half.

Third, how often. Fix n at 45 and invert. The minimum detectable difference is

  MDD = (z₀.₉₇₅ + z₀.₈₀) × √(σ_d²/n) = 2.801585 × √(12.80/45) = 1.4942 t C/ha

and the minimum useful resampling interval is MDD ÷ rate = 1.4942 ÷ 0.30 = 4.98, which rounds to five years. At one, two and three years the accrual is 0.30, 0.60 and 0.90 t C/ha and sits inside the noise band. Annual soil carbon sampling on this shape buys noise at full price, and it is sold routinely.

Fourth, the cost. Field extraction and compositing at $18.00 a core, dry combustion at $25.00 and bulk density at $12.00 gives $55.00 per sample. Mobilisation and relocation run $2,000.00 an event; design and statistical reporting $3,000.00; third-party validation and verification $12,000.00; registry and issuance $2,000.00 — $17,000.00 fixed per claim.

  paired event    2,000 +  45 × 55  =  $  4,475.00
  unpaired event  2,000 + 447 × 55  =  $ 26,585.00

  PAIRED   claim  17,000 + 2 × 4,475   =  $ 25,950.00
  UNPAIRED claim  17,000 + 2 × 26,585  =  $ 70,170.00

Here is the cut, and it is the only one this chapter makes. The difference between those two totals is $44,220.00. The thing that buys it is forty-five permanent georeferenced markers at $3.00 each — $135.00 of steel, returning 327.6 times its cost. Not a better laboratory. Not a satellite. Not a model. A stake in the ground and a number stamped on it. The correlation coefficient ρ is not a statistical property of soil; it is a purchasable good, and it is the cheapest thing in the entire field of environmental verification. The single most expensive decision in soil measurement is taken by whoever did not write down where they stood.

Fifth, what the claim is worth. 1.50 t C/ha × 44/12 = 5.500 t CO₂e/ha gross. Apply a 20 percent permanence buffer and a 10 percent uncertainty deduction and 3.960 t CO₂e/ha is saleable.

  $  5/t →  $  19.80 /ha      $  50/t →  $ 198.00 /ha
  $ 10/t →  $  39.60 /ha      $ 100/t →  $ 396.00 /ha
  $ 20/t →  $  79.20 /ha

Sixth, kappa. Define

  κ  =  total verification cost  ÷  value of the claim verified

A claim with κ > 1 costs more to prove than it is worth. At $20/t, one stratum per 200 ha, each stratum carrying its own 45 paired points:

   hectares  strata  cores      cost $     value $    kappa
          2       1     45      25,950         158  163.826
         50       1     45      25,950       3,960    6.553
        200       1     45      25,950      15,840    1.638
      1,000       5    225      45,750      79,200    0.578
      5,000      25  1,125     144,750     396,000    0.366
     20,000     100  4,500     516,000   1,584,000    0.326

And here is the honest negative, in two parts.

The first part is the smallholding. Eighty-four percent of the world's farms are under two hectares, and they hold about twelve percent of its farmland. A two-hectare holding making a five-year soil carbon claim at $20/t is claiming $158.40 and must spend $25,950.00 to prove it: κ = 163.8. At $50/t it is still 65.5. Even a stripped audit of $5,000.00 — which would not be credible — leaves κ at 31.6. For the median farm on Earth, verification costs between thirty and three hundred times the thing verified. Most regeneration claims on most farms are unverifiable at any honest price, and the correct response is to say so rather than to certify them.

The second part is worse, and it is the one aggregation does not fix. Look at the column that stops falling. The marginal verification cost is 2 × 45 × 55 ÷ 200 = $24.75 per hectare, and that term does not shrink with scale, because every new stratum brings its own forty-five cores. So kappa has a floor:

  at $  5/t   κ_floor = 24.75 /  19.80 = 1.250
  at $ 10/t   κ_floor = 24.75 /  39.60 = 0.625
  at $ 20/t   κ_floor = 24.75 /  79.20 = 0.312
  at $ 50/t   κ_floor = 24.75 / 198.00 = 0.125

Solve the minimum credible plot size A = C_fixed ÷ (κ_max·V_ha − c_ha) with C_fixed = $21,000.00 and c_ha = $24.75, and at $20/t with a 25 percent tolerance no area clears at all — the per-hectare cost alone, $24.75, exceeds the whole tolerated budget of $19.80. At $50/t, A is 848.5 ha. At $20/t you must relax the tolerance to 50 percent, and then A* is 1,414.1 ha. Direct measurement, done properly, has a minimum credible plot size in the high hundreds of hectares and a price floor beneath which it never works.

Seventh, the proxies, at their real error bars. Visible–near-infrared spectroscopy predicts soil carbon with a relative RMSE of roughly 25 percent, which on a 40 t C/ha stock is 10.00 t C/ha — 6.67 times the 1.50 t C/ha it is being asked to resolve. It measures the stock well and the change not at all, until averaged over many points.

GEDI is more interesting because its geometry is public. Footprints are 25 m across — 0.049087 ha — with shots every 60 m along track and beams separated by about 600 m, so one shot serves 3.60 ha of land. To detect a 3.00 Mg/ha biomass gain against a 20 Mg/ha footprint error with ρ = 0.70:

  σ_d² = 2 × 20² × 0.30 = 240.00      n = 7.848880 × 240 ÷ 9 = 209.30 → 210 shots
  implied area = 210 × 3.60 = 756.0 ha

Seven hundred and fifty-six hectares — arrived at from orbital geometry and an error bar, sharing no assumption with the soil calculation, and landing within one percent of the 755.4 ha that the proxy-based A* produces at $20/t. The binding constraint is the ratio of error bar to effect, and it is not bought off with a cheaper sensor.

Eighth, biodiversity, which is dearer still. With per-replicate eDNA detection probability p, replicates for 95 percent cumulative detection are ln(0.05)/ln(1−p): nine at p = 0.30, five at p = 0.50, three at p = 0.70. To detect an occupancy shift from 0.30 to 0.45 requires 160 sites — 7.848880 × 0.4575 ÷ 0.0225 = 159.59 — at five replicates and $55.00 a replicate, so $44,000.00 an event and $88,000.00 for the two-date claim: 3.39 times the soil claim. For κ = 0.25 that claim must carry $352,000.00 of value, which over 756 ha is $465.61 per hectare. Only a statutory market prices habitat that high; England's biodiversity net gain duty under the Environment Act 2021 is the working example of one, and outside such a duty a standalone biodiversity claim is not fundable by its own proceeds.

DREAM

What becomes ordinary

In the version of this that has already happened, nobody sells an unverified regeneration claim and nobody pretends a small farm can verify one alone.

Every parcel under a regeneration agreement carries a set of permanent markers with stamped identifiers, installed once, mapped once, and treated as infrastructure in the same way a water meter is. When the land changes hands the markers convey with it, and the incoming farmer inherits a baseline rather than starting one. The single largest cost in the old system — the cost of not knowing where the last person stood — has been designed out by an object that costs three dollars.

Claims are made by pools rather than by parcels. A cooperative, a processor, a catchment trust or a regional aggregator holds the verification contract for several thousand hectares, draws a stratified random subsample each cycle, carries the model, and distributes the proceeds by an allocation rule its members can read on one page. A farmer with two hectares is not excluded from the carbon market because she cannot fund an audit. She is a member of something that can, and her cost is a per-hectare levy of a few dollars a year rather than a five-figure invoice she could never justify.

Every published environmental claim carries its error bar in the same typeface as its value, because the market has learned to price them together and a claim without one does not trade. Verification bodies compete on the width of the confidence interval they can certify for a given spend, which is a real engineering competition and produces real improvement, in the way that competition over an unstated quantity never does.

Sampling happens on the interval the physics allows — five years for soil, annually for the proxies that screen which strata to visit — and nobody is paid to take a measurement that cannot resolve the thing it is measuring. The money that used to be spent on annual soil sampling with no statistical power is spent instead on more points, once, which is the same money buying an answer.

And the honest gap is visible. Where a claim genuinely cannot be verified at an affordable price, the system says so and pays for the practice instead, at a price that reflects the uncertainty. Nobody has to choose between lying and doing nothing, because a third option is written into the standard.

DESIGN

The structure that gets there

Three structures move kappa. They are complements, not alternatives, and each one moves a different term.

One: aggregation. It attacks C_fixed. Pool holdings into one registered activity; carry the sampling design, the model, the verification and the registry relationship at pool level; distribute by allocation rule. The precedents are specific and checkable: Verra's grouped-project provisions, Plan Vivo's community programmes since the 1990s, and the Kenya Agricultural Carbon Project, which put smallholders averaging well under a hectare each into a single verified activity.

What it moves: at $20/t, κ falls from 163.826 on a lone two-hectare farm to 0.578 at a thousand pooled hectares and 0.366 at five thousand. What it does not move: the floor of 0.312, because each new stratum brings its own forty-five cores. Aggregation alone never reaches a 25 percent tolerance at twenty dollars a tonne. It is necessary and it is not sufficient, and any scheme that claims otherwise has not printed its per-hectare term.

Two: tiered assurance. It attacks c_ha. Run a calibrated process model — this is the IPCC's Tier 3 in the national inventory tradition, and it is what Verra's VM0042 and the Climate Action Reserve's Soil Enrichment Protocol both permit — and verify it on a stratified subsample rather than on every stratum. Assurance level is matched to materiality, exactly as financial audit has done for decades through small-company exemption thresholds, and exactly as ISO 14064-3 does in distinguishing limited from reasonable assurance.

What it moves: sampling one stratum in five drops c_ha from $24.75 to $4.95. Adding a $6,000.00 model build, A* at $20/t and κ_max 0.25 becomes 1,818.2 ha, and at $50/t 606.1 ha. At five thousand pooled hectares κ is 0.131. This is the structure that makes the thing possible at all, and it is the one most often left out because it requires somebody to state a tolerance in public.

Three: outcome proxies. It attacks both terms and pays in precision. Screen with remote sensing at a marginal cost near $1.50 a hectare, and price the result honestly by widening the uncertainty deduction from 10 to 30 percent, so saleable abatement falls from 3.960 to 3.080 t CO₂e/ha. At $20/t, A* is 755.4 ha and κ at five thousand hectares is 0.058.

The discipline that makes this legitimate rather than convenient is the one Badgley and colleagues demonstrated in California's forest offsets: a proxy with an unstated or self-serving error term produces systematic over-crediting, and the over-crediting is invisible in every individual transaction. A proxy is admissible only where its error is measured against direct observation on a calibration subsample and the deduction is set from that measurement. That is why the three structures are one design and not three: the proxy is what makes it cheap, the tiered subsample is what keeps the proxy honest, and the pool is what pays for the subsample.

Put together at five thousand hectares and $20/t: cost $24,187.50, value $308,000.00, κ = 0.079.

DESTINY

How it holds when nobody is pushing

It holds when the markers outlive the scheme. A steel peg with a stamped disc survives a change of owner, a change of standard, a change of registry and a change of government, and every one of those changes is otherwise a reset to an unpaired design at $70,170.00 a claim. Convey the markers with the title. That one clause is worth more than any amount of methodological improvement.

It holds when the pool has a reason to exist beyond the claim. A cooperative that already buys inputs, dries grain, insures members or markets their produce can carry a verification function at near-zero marginal governance cost. A pool constituted solely to sell carbon dissolves the year the price falls — and prices do fall, and at $5/t the kappa floor for direct measurement is 1.250, which is to say the whole activity is under water.

And here is where it fails, stated plainly.

It fails when the pool's uncertainty deduction is applied uniformly, because the best-performing member then subsidises the worst and leaves, and the pool unravels from the top. Ostrom's design principles are precise about this: appropriation and provision rules must be congruent with local conditions, or the commons erodes from its most capable participants outward. A pool needs a within-pool allocation rule that differentiates, and building one costs real governance effort that nobody budgets for.

It fails when somebody resamples early. A three-year reading on this shape cannot resolve the accrual, and it will come back near zero or negative — and a near-zero reading, once published, is very hard to un-publish.

It fails when the proxy's calibration subsample is quietly reduced in a cost review, which is the single most likely thing to happen in year four of any programme, and which converts a defensible instrument into an unmeasured one without changing a single published number.

And it fails when the pool becomes the only counterparty the farmer has. Once verification is carried at pool level, no individual member's claim is verifiable even in principle. That is a real transfer of power to the aggregator and it should be priced, governed and disclosed as one, not treated as an administrative detail.

DELIGHT

What it feels like

There is a specific satisfaction in walking a field with a map and finding the peg exactly where the map said it would be. It is a small thing and it does not feel like finance. It feels like the moment a long argument becomes unnecessary, because the ground itself is now a witness that neither side can coach.

And there is a better one, which comes about five years later. You pull the second set of cores from the same forty-five holes, you run the paired difference, and the interval does not straddle zero. It is not a dramatic number. It is 1.6 or 1.8 tonnes a hectare and it took five years to arrive. But it is yours in a way that no modelled estimate ever is, because you know exactly where every one of those forty-five numbers came from and so does everybody else.

People who have done this describe the same thing: the pleasure is not in being right. It is in having built something that could have proven them wrong and did not. That is a rarer feeling than it should be, and it is available to anybody willing to spend a hundred and thirty-five dollars on steel before they spend anything on laboratories.

OPERATIONALIZE THIS

At the level of finance

The instrument: a measurement pool — a verification cost-sharing facility with a stratified draw and a per-hectare levy.

Not a carbon project. A verification utility that any carbon, water, biodiversity or supply-chain claim over the same land can buy service from. It is closer in structure to a captive insurer or a mutual audit fund than to a project developer, and that distinction is what keeps it alive when one credit market turns.

The mechanics.

The entity. A cooperative or special-purpose vehicle owned by the landholders, holding the sampling contract, the marker register, the model and the registry relationship. Members contribute hectares and a levy; they do not contribute claims.
The marker register. Every enrolled hectare carries permanent georeferenced markers at $3.00 each, installed once, recorded in a register held by the entity, and conveyed with title. This is the pool's principal asset and it should be on its balance sheet as one.
The draw. Each cycle, a stratified random subsample of strata is sampled directly at 45 paired points per stratum; the remainder is estimated by the calibrated model and screened by remote sensing. The draw is random, it is published in advance as a fraction and never as a list, and the fraction is a covenanted minimum.
The deduction. Uncertainty deduction set from the calibration residuals, not from a table. Measured strata carry 10 percent; modelled strata carry 30 percent. Members can see which they are in and why.
The levy. Per hectare per year, covering the pool's whole programme cost.
The term. Five years to the first verified reading, because 4.98 years is what the physics requires. Anything shorter is sold, not measured.

The balance-sheet treatment. The marker register and the model calibration are capitalised as intangible assets of the pool and amortised over the crediting period, because both are one-off expenditures that produce a stream of future verification capability — this is the same argument as any capitalised development cost and your auditor has it every year. The levy is deferred income recognised across the cycle. Member holdings are unaffected: the pool holds the measurement asset, the member holds the land.

The counterparty. An existing cooperative, processor, catchment trust or input supplier first, never a vehicle constituted for the purpose. Governance already exists there, the members already meet, and the marginal cost of adding a verification function to a body that already handles money is a fraction of standing one up.

The number that decides it. One inequality, on the front page:

                       C_fixed
      H*  =  ------------------------------
              κ_max · V_ha  −  c_ha

With C_fixed = $16,500.00 (analysis, verification and one model build) and c_ha = $2.74 (proxy screen plus the calibration draw), at a 25 percent tolerance:

  $   5/t   V_ha $  15.40   H* = 14,831.5 ha  =  7,416 holdings of 2 ha
  $  10/t   V_ha $  30.80   H* =  3,324.9 ha  =  1,662 holdings
  $  20/t   V_ha $  61.60   H* =  1,303.1 ha  =    652 holdings
  $  50/t   V_ha $ 154.00   H* =    461.4 ha  =    231 holdings
  $ 100/t   V_ha $ 308.00   H* =    222.2 ha  =    111 holdings

If you cannot assemble H\* hectares at today's price, you do not have a measurement problem. You have a membership problem, and that is a far more tractable one — it is solved by a cooperative's existing member list, not by a laboratory.

At $20/t across a 5,000-hectare pool the whole programme costs $30,187.50, or $6.04 per hectare per cycle, $1.21 per hectare per year — 9.8 percent of the claim. That is a levy a farm secretary can approve without a board paper.

The first ninety days.

Day	Action	Artifact
1–15	Count enrolled hectares; compute H\* at three prices	One page: H\* and the gap
16–30	Stratify the pool; fix the stratum size and the draw fraction	Sampling design memo
31–45	Install and register markers on the first drawn strata	The marker register
46–60	Agree the baseline, the method and the deduction rule with the verifier	The signed baseline
61–75	Baseline sampling on drawn strata; first proxy screen	Baseline dataset
76–90	Set the levy; publish the allocation rule to members	The levy and the rule

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

Where do we already return to exactly the same place to take a reading, and who set that up? What did it let us see that we could not otherwise have seen?
Which of our numbers is published with its error bar, and what made that possible where it was not possible elsewhere?
When have we shared the cost of proving something with somebody else, and what did the sharing make affordable that would not have been?

Dream — what becomes possible

If every parcel we touch carried permanent markers conveyed with the title, what would we be able to claim in ten years that we cannot claim now?
Imagine our verification cost at one tenth of what it is. What would we start measuring that we currently only assert?
If a two-hectare holding could make a credible claim through us, how many would join, and what else would they want from the same membership?

Design — what we build

What is the smallest pool we could convene from people who already meet for another reason, and who convenes it?
What tolerance would we be willing to publish — what value of kappa — and what would we have to be sure of before we published it?
What is the one clause in our agreements that would make a marker register survive a change of ownership, and who drafts it?

Destiny — how it holds

What would keep this pool worth belonging to in a year when the carbon price halves?
How would our best-performing member know they were not subsidising our worst, and what would they need to see to stay?
If somebody proposed trimming the calibration subsample to save money in year four, who in this room would notice, and what would they say?

WORKS CITED

Badgley, G., Freeman, J., Hamman, J. J., Haya, B., Trugman, A. T., Anderegg, W. R. L. and Cullenward, D. (2022). "Systematic over-crediting in California's forest carbon offsets program." Global Change Biology, 28(4), 1433–1445.

Biggs, J., Ewald, N., Valentini, A., Gaboriaud, C., Dejean, T., Griffiths, R. A., Foster, J., Wilkinson, J. W., Arnell, A., Brotherton, P., Williams, P. and Dunn, F. (2015). "Using eDNA to develop a national citizen science-based monitoring programme for the great crested newt (Triturus cristatus)." Biological Conservation, 183, 19–28.

Bradford, M. A., Carey, C. J., Atwood, L., Bossio, D., Fenichel, E. P., Gennet, S., Fargione, J., Fisher, J. R. B., Fuller, E., Kane, D. A., Lehmann, J., Oldfield, E. E., Ordway, E. M., Rudek, J., Sanderman, J. and Wood, S. A. (2019). "Soil carbon science for policy and practice." Nature Sustainability, 2, 1070–1072.

Brus, D. J. and de Gruijter, J. J. (1997). "Random sampling or geostatistical modelling? Choosing between design-based and model-based sampling strategies for soil." Geoderma, 80(1–2), 1–44.

Climate Action Reserve (2020). Soil Enrichment Protocol, version 1.0.

Conant, R. T., Smith, G. R. and Paustian, K. (2003). "Spatial variability of soil carbon in forested and cultivated sites." Journal of Environmental Quality, 32(1), 278–286.

Cooperrider, D. L. and Whitney, D. (2005). Appreciative Inquiry: A Positive Revolution in Change. Berrett-Koehler.

de Gruijter, J. J., Brus, D. J., Bierkens, M. F. P. and Knotters, M. (2006). Sampling for Natural Resource Monitoring. Springer.

de Gruijter, J. J., McBratney, A. B., Minasny, B., Wheeler, I., Malone, B. P. and Stockmann, U. (2016). "Farm-scale soil carbon auditing." Geoderma, 265, 120–130.

Deiner, K., Bik, H. M., Mächler, E., Seymour, M., Lacoursière-Roussel, A., Altermatt, F., Creer, S., Bista, I., Lodge, D. M., de Vere, N., Pfrender, M. E. and Bernatchez, L. (2017). "Environmental DNA metabarcoding: Transforming how we survey animal and plant communities." Molecular Ecology, 26(21), 5872–5895.

Dubayah, R., Blair, J. B., Goetz, S., Fatoyinbo, L., Hansen, M., Healey, S., Hofton, M., Hurtt, G., Kellner, J., Luthcke, S., Armston, J., Tang, H., Duncanson, L., Hancock, S., Jantz, P., Marselis, S., Patterson, P. L., Qi, W. and Silva, C. (2020). "The Global Ecosystem Dynamics Investigation: High-resolution laser ranging of the Earth's forests and topography." Science of Remote Sensing, 1, 100002.

Environment Act 2021 (United Kingdom), Schedule 14 — biodiversity gain as a condition of planning permission.

Ficetola, G. F., Pansu, J., Bonin, A., Coissac, E., Giguet-Covex, C., De Barba, M., Gielly, L., Lopes, C. M., Boyer, F., Pompanon, F., Rayé, G. and Taberlet, P. (2015). "Replication levels, false presences and the estimation of the presence/absence from eDNA metabarcoding data." Molecular Ecology Resources, 15(3), 543–556.

Intergovernmental Panel on Climate Change (2006). 2006 IPCC Guidelines for National Greenhouse Gas Inventories, Volume 4: Agriculture, Forestry and Other Land Use. IGES.

International Organization for Standardization (2019). ISO 14064-2: Greenhouse gases — Part 2 and ISO 14064-3: Greenhouse gases — Part 3: Specification with guidance for the verification and validation of greenhouse gas statements.

Johnston, A. E. and Poulton, P. R. (2018). "The importance of long-term experiments in agriculture: their management to ensure continued crop production and soil fertility; the Rothamsted experience." European Journal of Soil Science, 69(1), 113–125.

Kravchenko, A. N. and Robertson, G. P. (2011). "Whole-profile soil carbon stocks: the danger of assuming too much from analyses of too little." Soil Science Society of America Journal, 75(1), 235–240.

Lowder, S. K., Sánchez, M. V. and Bertini, R. (2021). "Which farms feed the world and has farmland become more concentrated?" World Development, 142, 105455.

MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Royle, J. A. and Langtimm, C. A. (2002). "Estimating site occupancy rates when detection probabilities are less than one." Ecology, 83(8), 2248–2255.

Norris, C. E., Bean, G. M., Cappellazzi, S. B., Cope, M., Greub, K. L. H., Liptzin, D., Rieke, E. L., Tracy, P. W., Morgan, C. L. S. and Honeycutt, C. W. (2020). "Introducing the North American project to evaluate soil health measurements." Agronomy Journal, 112(4), 3195–3215.

Oldfield, E. E., Eagle, A. J., Rubin, R. L., Rudek, J., Sanderman, J. and Gordon, D. R. (2022). "Crediting agricultural soil carbon sequestration." Science, 375(6586), 1222–1225.

Orgiazzi, A., Ballabio, C., Panagos, P., Jones, A. and Fernández-Ugalde, O. (2018). "LUCAS Soil, the largest expandable soil dataset for Europe: a review." European Journal of Soil Science, 69(1), 140–153.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Paustian, K., Collier, S., Baldock, J., Burgess, R., Creque, J., DeLonge, M., Dungait, J., Ellert, B., Frank, S., Goddard, T., Govaerts, B., Grundy, M., Henning, M., Izaurralde, R. C., Madaras, M., McConkey, B., Porzig, E., Rice, C., Searle, R., Seavy, N., Skalsky, R., Mulhern, W. and Jahn, M. (2019). "Quantifying carbon for agricultural soil management: from the current status toward a global soil information system." Carbon Management, 10(6), 567–587.

Plan Vivo Foundation. The Plan Vivo Standard for Community Payments for Ecosystem Services Programmes. Successive editions.

Smart, A. S., Weeks, A. R., van Rooyen, A. R., Moore, A., McCarthy, M. A. and Tingley, R. (2016). "Assessing the cost-efficiency of environmental DNA sampling." Methods in Ecology and Evolution, 7(11), 1291–1298.

Smith, P., Soussana, J.-F., Angers, D., Schipper, L., Chenu, C., Rasse, D. P., Batjes, N. H., van Egmond, F., McNeill, S., Kuhnert, M., Arias-Navarro, C., Olesen, J. E., Chirinda, N., Fornara, D., Wollenberg, E., Álvaro-Fuentes, J., Sanz-Cobena, A. and Klumpp, K. (2020). "How to measure, report and verify soil organic carbon to realize the potential of soil carbon sequestration for atmospheric greenhouse gas removal." Global Change Biology, 26(1), 219–241.

Verra (2023). VM0042 Methodology for Improved Agricultural Land Management, version 2.0. And VCS Standard, grouped project requirements.

Viscarra Rossel, R. A., Behrens, T., Ben-Dor, E., Brown, D. J., Demattê, J. A. M., Shepherd, K. D., Shi, Z., Stenberg, B., Stevens, A., Adamchuk, V., Aïchi, H., Barthès, B. G., Bartholomeus, H. M., Bayer, A. D., Bernoux, M., Böttcher, K., Brodský, L., Du, C. W., Chappell, A. and others (2016). "A global spectral library to characterize the world's soil." Earth-Science Reviews, 155, 198–230.

World Bank. State and Trends of Carbon Pricing. Annual. And BioCarbon Fund project documentation for the Kenya Agricultural Carbon Project.

Within this edition. Chapter II.10, Measurement and What It Does, on the measure as an intervention and on what targeting does to a proxy. Chapter IV.03, Agriculture as the Model Case, on the agronomy of soil carbon itself. Every figure above is computed in lib/verify/IV_10.py and can be re-run with python3 lib/verify.py IV.10; inputs are printed with their units and their sources, and model assumptions are labelled as such.