Credits are issued for periods. Trees grow continuously and are measured occasionally. Somewhere between the field campaign and the monitoring report, a handful of measurements becomes a year-by-year table, and that conversion is where a surprising amount of reported carbon originates.
What You Actually Have
A project with monitoring events in years 3, 6 and 10 has three observations. It reports carbon for ten years.
The seven other values come from a model: interpolation between observations, and a growth function fitted to whatever data exists. That is legitimate and standard. It stops being legitimate when the modelled values are presented with the same authority as the measured ones.
| Value type | Source | Uncertainty |
|---|---|---|
| Measured year | Field inventory | Sampling + model error |
| Interpolated year | Between two measurements | Adds interpolation error |
| Extrapolated year | Beyond the last measurement | Largest — and unbounded |
Why Extrapolation Is the Dangerous One
Interpolating between two measured points is constrained on both sides. The curve has to pass through both, so error is bounded.
Extrapolating past the final measurement has no such constraint. A growth function fitted to years 3 through 10 and projected to year 20 is asserting a shape nobody has observed, and the shape matters enormously, because tree growth decelerates and the deceleration point varies with species and site.
A linear extrapolation from the juvenile phase will overstate later stock substantially, because that is exactly the phase where growth is fastest.
The Curve Shape Is a Claim
Most growth functions used in forestry, which are sigmoid forms of various kinds, share a characteristic shape, comprising a slow start, a rapid middle, and a tapering approach to an asymptote.
Choosing the function and fitting its parameters is a modelling decision with real consequences for the reported numbers. The requirements that make it defensible are unglamorous:
Fit to local data where it exists. A curve fitted to plantation data from another region is an assumption wearing the clothes of a measurement.
State the asymptote. If the function implies a maximum stock, say what it is and whether it is plausible for the site. Implausible asymptotes are the fastest way to spot an overfitted curve.
Do not fit more parameters than the data supports. Three observations cannot support a four-parameter function. The fit will be perfect and meaningless.
Reconciling Model and Measurement
The discipline that keeps this honest is simple: when a new monitoring event arrives, compare the measured value against what the previous curve predicted.
If they agree, the model is behaving. If the measurement comes in materially below the projection, the curve was optimistic and everything projected beyond it needs revisiting, including, potentially, previously issued periods.
Projects that perform and document this comparison at every monitoring event catch drift early. Projects that simply refit the curve to include the new point and carry on have discarded the one diagnostic that would have told them the model was wrong.
What Belongs in the Report
The measured values, clearly marked as measured. The modelled values, clearly marked as modelled. The function and its parameters. The comparison against the previous projection.
An assessor reading that can evaluate the growth claim. One reading an undifferentiated table of annual figures has to ask which ones came from a field visit and that question, asked late, is rarely comfortable.
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.
