A sampling plan is the document that decides how many credits a project can defend. Most are written backwards, meaning a plot count is chosen, and then a justification is assembled around it. Written forwards, the logic runs from precision target to variance to plot count, and every number has a reason.
Start From the Precision You Need
Before any fieldwork, decide what confidence interval the project is aiming for. This is a commercial decision as much as a statistical one: standards deduct conservatively in proportion to reported uncertainty, so the precision target sets how many tonnes survive to issuance.
Working backwards from that target is what makes every subsequent decision defensible.
Building the plan, in order
Define the frame
Establish the exact area being sampled and what is excluded, with boundaries you can defend geospatially.
Stratify
Divide the area by the variables that actually drive biomass variance, forest type, stand age, soil, disturbance history. Not administrative boundaries.
Estimate variance
Use pilot plots or prior inventory to measure within-stratum variance. This is the number everything else depends on.
Set precision
State the target confidence interval and the confidence level, and record why that level was chosen.
Compute plot counts
Derive plots per stratum from variance, target precision and stratum area.
Allocate
Distribute plots across strata by area AND variability, not area alone.
Locate
Place plots by a probabilistic rule, recorded in advance. Accessibility must not drive placement.
Document
Write down every decision and its basis while it is fresh. This becomes the verification evidence.
Stratification Is Where the Money Is
Everything after stratification is arithmetic. Stratification itself is judgement, and it is the highest-leverage decision in the plan.
Good strata are internally uniform in carbon density. That pushes variance into the between-stratum component, which the area weighting handles for free, and leaves a small within-stratum variance driving plot counts.
| Stratification basis | Reduces variance | Verifier view |
|---|---|---|
| Forest type / ecosystem | Strongly | Expected |
| Stand age or successional stage | Strongly | Expected in ARR |
| Soil and hydrology | Strongly where soil counts | Expected on peat and mangrove |
| Disturbance history | Moderately | Supportive |
| Ownership or compartment boundary | Rarely | Treated as unjustified |
The Two Mistakes That Recur
Allocating by area alone. A large uniform stratum gets many plots it does not need; a small chaotic one gets too few, and its variance dominates the total. Allocation should weight both size and variability.
Letting access choose locations. Plots placed where the crew could reach are not a probability sample. The bias is unmeasurable and unfixable after the fact, and verifiers treat it conservatively because they have no basis to do otherwise.
What the Plan Has to Contain
A verifier reading the plan should be able to reconstruct the design without asking a question: the frame and its exclusions, the strata and the reasoning behind them, the variance estimates and their source, the precision target, the plot count derivation, the allocation rule, the location rule, and the field protocol including what is measured in which nested layer.
Everything in that list is cheap to record at design time and expensive to reconstruct at verification.
Image credit
Hero image: Canopy in protected rainforest, Eungella National Park, Queensland by Kimberly Melissa Booth, CC BY-SA 4.0, via Wikimedia Commons.
Turn uncertainty into a plan
Sampling design and growth assumptions decide how much measured carbon survives conservative deduction. TREEO Carbon Simulator models sequestration scenarios and biomass growth before the field season.
