La Bourse / Volume IV / Nº IV.10 / Ten concept briefs

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

TEN CONCEPT BRIEFS · Chapter IV.10 — Measuring Regeneration

One page each. A reader who reads only these ten pages has the chapter.

BRIEF 1 — Kappa, the Verification Ratio

The idea. Every claim about land has two numbers attached: what it is worth, and what it costs to prove. The ratio between them decides whether the claim can exist.

  κ  =  total verification cost  ÷  value of the claim verified

κ below about 0.10 is a commodity claim. Between 0.10 and 0.25 it is workable with discipline. Above 1.00 the claim costs more to prove than it is worth, and no amount of conviction changes that.

Worked example. A five-year soil carbon claim on a farm accruing 0.30 t C/ha/yr is worth 3.960 t CO₂e/ha after deductions — $79.20 per hectare at $20 a tonne. Verifying it honestly costs $25,950.00. On two hectares: κ = 25,950 ÷ 158.40 = 163.826. On 5,000 pooled hectares: κ = 144,750 ÷ 396,000 = 0.366.

Why it matters. κ tells you instantly which conversation you are in. A high κ is not an argument against the practice — the carbon is still accruing. It is an argument against that party making that claim alone, which is a different and much more fixable problem.

You already know this because you have decided not to claim a small expense on tax, not because it was not a real expense, but because the receipts, the form and the risk of a query cost more than the relief was worth.

BRIEF 2 — Effect Size Against Noise

The idea. You cannot detect a change smaller than your own measurement scatter unless you take enough samples, and how many is arithmetic rather than judgement.

A temperate cropland soil holds around 40 t C/ha in the top thirty centimetres. Within-field variation runs about 20 percent, so a standard deviation of 8.00 t C/ha. Five years of improved management adds 1.50 t C/ha — 3.75 percent of the stock.

Worked example. The effect is 1.50 and the scatter is 8.00. The signal is one fifth of the noise. That single comparison is the whole cost problem in soil verification, and it does not go away with a better laboratory, because the scatter is in the field, not in the instrument.

Why it matters. It explains why soil carbon is expensive and why forest biomass is not: a growing tree changes a lot relative to the variation around it, and a recovering soil changes very little. Cost follows the ratio, not the importance.

You already know this because you have tried to tell whether a diet was working by weighing yourself daily, and found that the day-to-day swing was larger than the week-to-week trend.

BRIEF 3 — Paired Sampling, and Why ρ Is Purchasable

The idea. If you measure the same points twice, the variance you must overcome is not the field's variance but the variance of the difference:

  σ_d² = 2σ²(1 − ρ)

where ρ is the correlation between the two visits. In soil, ρ is high — most of the variance in a field is spatial and travels with the point.

Worked example. At σ = 8.00 t C/ha and Δ = 1.50 t C/ha:

  independent samples       n = 2kσ²/Δ²  =  447 cores per date
  paired at ρ = 0.70        n = 134 cores
  paired at ρ = 0.90        n =  45 cores
  paired at ρ = 0.95        n =  23 cores

At ρ = 0.90 the design needs 9.93 times fewer cores.

Why it matters. ρ is not given by nature. It is bought — with a stake, a stamped number and a coordinate. It is the only term in the whole cost stack that can be moved by a decision costing three dollars.

You already know this because a before-and-after photograph taken from the same spot on the same tripod tells you something that two photographs from different angles never can.

BRIEF 4 — Minimum Detectable Difference and the Sampling Interval

The idea. Fix the number of samples and the design tells you the smallest change it can see. Divide that by the rate of change and you get the soonest you may usefully come back.

  MDD = (z₀.₉₇₅ + z₀.₈₀) × √(σ_d²/n)          T = MDD ÷ rate

Worked example. With 45 paired points at ρ = 0.90, σ_d² = 12.80:

  MDD = 2.801585 × √(12.80/45)  =  1.4942 t C/ha
  T   = 1.4942 ÷ 0.30           =  4.98 years  →  5 years

At one, two and three years the accrual is 0.30, 0.60 and 0.90 t C/ha — all inside the noise band.

Why it matters. Annual soil carbon sampling on this shape cannot resolve the accrual it is paying for. It is sold anyway, and it is the most common waste in the whole field. The same budget spent on more points, less often, buys an answer instead of a series of shrugs.

You already know this because you do not weigh a child every morning to see whether they are growing. You mark the doorframe once a year.

BRIEF 5 — The Cost Stack, and Where It Actually Sits

The idea. Verification cost has a fixed part and a per-hectare part, and they behave completely differently under scale.

  per sample:  field 18.00 + combustion 25.00 + bulk density 12.00  =  $55.00
  per event:   mobilisation and relocation                          =  $2,000.00
  per claim:   design 3,000 + verification 12,000 + registry 2,000  =  $17,000.00

Worked example. A paired event costs 2,000 + 45 × 55 = $4,475.00. Two events plus the fixed programme gives $25,950.00 for the whole five-year claim. The unpaired equivalent is $70,170.00.

Why it matters. Only the fixed part is shared by pooling. The per-hectare part — $24.75 at one stratum per 200 ha — travels with every new hectare, and it is what sets the floor that aggregation cannot reach below.

You already know this because you know the difference between the fee your accountant charges to exist and the fee they charge per transaction, and which one gets cheaper when you merge two companies.

BRIEF 6 — The Kappa Floor

The idea. Aggregation spreads fixed cost. It does not touch marginal cost. So κ does not fall toward zero — it falls toward a floor set by the per-hectare term divided by the per-hectare value.

  κ_floor  =  c_ha ÷ V_ha

Worked example. With c_ha = $24.75 and full direct sampling:

  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

At twenty dollars a tonne, direct measurement never reaches a 25 percent tolerance at any scale whatsoever, because $24.75 already exceeds the whole tolerated budget of $19.80 per hectare.

Why it matters. It is the single most important check on any pooling scheme. If the promoter cannot tell you their per-hectare term, they have not found the floor and neither have you.

You already know this because a bulk discount on a thing that costs real money to make still never gets you below the cost of making it.

BRIEF 7 — Minimum Credible Plot Size

The idea. Given a tolerance, a price and a cost structure, there is a smallest area that can be honestly verified, and it is one line of algebra.

  A*  =  C_fixed  ÷  (κ_max · V_ha  −  c_ha)

If the denominator is zero or negative, no area clears — the claim is unverifiable at that price and that tolerance however much land you assemble.

Worked example. Direct measurement, C_fixed = $21,000.00, c_ha = $24.75, κ_max = 0.25:

  at $ 20/t   NO AREA CLEARS   (24.75 > 0.25 × 79.20 = 19.80)
  at $ 50/t   A* =   848.5 ha
  at $100/t   A* =   282.8 ha

With tiered assurance sampling one stratum in five, c_ha falls to $4.95 and at $20/t A\* = 1,818.2 ha.

Why it matters. It converts an argument about ambition into a number of hectares, and a number of hectares is a membership list.

You already know this because every business you have run had a break-even volume, and below it no amount of enthusiasm made the unit economics work.

BRIEF 8 — Proxies and Their Real Error Bars

The idea. A proxy is admissible when its error is measured, and inadmissible when its error is asserted. Price the proxy by widening the deduction, not by hoping.

Worked example. Visible–near-infrared spectroscopy predicts soil carbon with a relative RMSE near 25 percent — 10.00 t C/ha on a 40 t C/ha stock, which is 6.67 times the 1.50 t C/ha it is asked to resolve. GEDI's mission requirement is a 1 km cell mean within 20 Mg/ha or 20 percent; detecting a 3.00 Mg/ha biomass gain at ρ = 0.70 needs 210 footprints, and at 3.60 ha of land served per shot that is 756.0 ha of minimum credible area.

Priced honestly — 30 percent uncertainty deduction instead of 10 — saleable abatement falls from 3.960 to 3.080 t CO₂e/ha, and at $20/t, A* = 755.4 ha.

Why it matters. Two routes sharing no assumptions land in the same few hundred hectares. The binding constraint is the ratio of error bar to effect and a cheaper sensor does not move it.

You already know this because a bathroom scale accurate to half a kilo is useless for weighing a letter, however cheap it is.

BRIEF 9 — eDNA, Detection Probability and Replication

The idea. A molecular survey does not return presence or absence. It returns a detection, with a probability attached, and the number of replicates you need follows from that probability.

  1 − (1 − p)ⁿ ≥ 0.95     →     n ≥ ln(0.05) ÷ ln(1 − p)

Worked example. At p = 0.30, nine replicates. At p = 0.50, five. At p = 0.70, three. Detecting a shift in occupancy from 0.30 to 0.45 needs

  n = 7.848880 × (0.30×0.70 + 0.45×0.55) ÷ 0.15²  =  159.59  →  160 sites

At five replicates and $55.00 each, that is $44,000.00 per event and $88,000.00 for a two-date claim — 3.39 times the soil claim on the same land. For κ = 0.25 the claim must carry $352,000.00, or $465.61 per hectare over 756 ha.

Why it matters. Biodiversity is the dearest thing to verify and the thinnest thing to sell. Outside a statutory duty like England's biodiversity net gain requirement, a standalone biodiversity claim cannot fund its own proof.

You already know this because a negative test result on a test you know misses a third of cases does not mean you are clear, and you would take a second one.

BRIEF 10 — The Measurement Pool

The idea. Do not build a carbon project. Build a verification utility that any claim over the same land can buy service from, owned by the landholders, with a marker register as its principal asset.

The three structures it combines. Aggregation spreads C_fixed. Tiered assurance — a calibrated model verified on a stratified subsample, the IPCC's Tier 3 logic and ISO 14064-3's limited-versus-reasonable distinction — cuts c_ha. Outcome proxies cut both and pay in a wider deduction.

Worked example. At $20/t across 5,000 pooled hectares:

  aggregation alone        cost 144,750   value 396,000   κ = 0.366
  + tiered assurance       cost  51,750   value 396,000   κ = 0.131
  outcome proxies          cost  18,000   value 308,000   κ = 0.058
  all three                cost  24,188   value 308,000   κ = 0.079

The pool's programme costs $30,187.50, or $1.21 per hectare per year — 9.8 percent of the claim. The deciding number is H\*, the hectares that must be pooled: 1,303.1 ha at $20/t, which is 652 holdings of two hectares.

Why it matters. The precedents are real and checkable: Verra's grouped projects, Plan Vivo's community programmes, and the Kenya Agricultural Carbon Project, which verified tens of thousands of sub-hectare smallholders as one activity.

You already know this because you have belonged to something — a cooperative, a mutual, a trade body — that bought an audit, an insurance policy or a laboratory you could never have bought alone.