Two things are true simultaneously and are usually discussed separately. Satellite-derived biomass estimates are models that require ground data to calibrate and to have their accuracy assessed. And in most projects, that ground data is produced by the party whose revenue depends on the result. The gap between those two facts is the layer that is missing.
Why Remote Sensing Does Not Close It
Earth observation has transformed change detection. It has not removed the need for field measurement, for a reason that is structural rather than technological.
A satellite measures reflectance or backscatter. Biomass is inferred from that signal through a model fitted to ground plots. The model is only valid within the range of conditions it was calibrated on, and its accuracy can only be assessed against reference data it was not trained on.
| Task | Satellite alone | Requires ground data |
|---|---|---|
| Detect that canopy changed | Yes | No |
| Map where and when | Yes | No |
| Say how much biomass was there | No | Yes — calibration |
| State the accuracy of that figure | No | Yes — independent assessment |
The last two rows are where carbon credits are actually generated. More satellite data does not substitute for them; it increases the amount of inference resting on the same calibration plots.
The Independence Problem
Set the technology aside. The structural issue is that the measurement is usually produced inside the organisation that benefits from it being large.
This does not require anyone to act in bad faith. Measurement involves dozens of small discretionary choices, such as which plots to select, which allometric equation to apply, how to treat an outlier, or whether a stem is in or out. Each choice is individually defensible. However, when made by someone who knows which direction helps, these choices do not distribute randomly.
That is the same compounding problem that made project-level REDD+ baselines untrustworthy, applied to the measurement side instead of the baseline side. The response there was to remove the discretion from the interested party. The measurement side has not had that reckoning yet.
What an Independent Layer Would Do
Not replace the verifier, because the verifier checks whether reported numbers trace to evidence, on a sample, after the fact.
The missing function sits earlier: producing the evidence itself, to a documented protocol, by a party with no stake in the magnitude. Plot establishment and measurement, tree-level records with provenance, calibration data for whatever remote-sensing model is used, and an accuracy assessment on reference data held separate from training data.
The verifier's job becomes far easier when that exists, because the chain they are asked to trace was built to be traced.
Why the Market Is Moving This Way
Buyers have exhausted the easy filters. Methodology approval no longer differentiates because most serious projects use approved methodologies. Standard names no longer differentiate. What remains is the quality of the underlying numbers, and buyers are beginning to ask about uncertainty, sampling design and who produced the data.
A project that can answer "an independent party, to this protocol, with this assessed accuracy" is in a materially different position from one that answers "our team".
The Honest Limitation
An independent measurement layer costs money and does not increase the tonnage. It frequently reduces it, because rigorous uncertainty quantification triggers conservative deductions that casual measurement avoids by never computing the uncertainty at all.
The argument for it is not that it produces more credits. It is that it produces credits that survive scrutiny, in a market that has repeatedly discovered the alternative, and where the cost of a credit class losing buyer confidence has been far larger than the cost of measuring it properly.
Move beyond estimates
Verifiers test sampling design, uncertainty and whether a number traces back to the field. TREEO dMRV captures that evidence in real time, in one auditable chain.
