Scientific rigour. Commercial precision.
Atmospheric measurement, a tamper-evident ledger and a reconciliation loop that keeps the two honest. None of it is novel science — the novelty is running it continuously, at a price a project can carry.
We measure the air above the claim, not a coefficient in a table.
Eddy-covariance towers sample the exchange between land and atmosphere ten times a second. Regional inversion fills the space between towers using satellite column measurements, so coverage is continuous rather than a set of points with assumptions in between.
The output is a flux series with an explicit uncertainty band. Where the band is wide we say so, which is the part a modelled estimate structurally cannot do.
- Sampling rate
- 10 Hz raw, aggregated hourly
- Spatial method
- Tower flux + satellite column inversion
- Reported with
- 68% and 95% uncertainty bands
- Latency
- Readings available within the hour
Every number keeps the chain that produced it.
Each reading is committed with its instrument, its calibration state and the processing version applied to it. A figure in a report can be walked back to the raw sample that produced it, by someone who does not work here.
Restatements are appends, never edits. If a calibration is found to have drifted, the correction and the original both stay on the record — which is the only version of “auditable” that means anything.
- Record
- Reading, instrument, calibration, processing version
- Corrections
- Appended; the original is never overwritten
- Verification
- Third parties can re-derive a published figure
- Export
- Raw series and provenance, on request, in full
Measured against modelled, published either way.
The reconciliation runs whether or not the answer flatters the project. Below is a worked example from a peatland site: the model expected one figure, the measurement found another, and the difference is what the buyer was actually paying for.
| Period | Modelled (tCO₂e) | Measured (tCO₂e) | Variance | 95% band |
|---|---|---|---|---|
| Q1 | 4,200 | 4,910 | +16.9% | ±310 |
| Q2 | 5,100 | 6,040 | +18.4% | ±355 |
| Q3 | 5,400 | 6,120 | +13.3% | ±390 |
| Q4 | 3,800 | 4,270 | +12.4% | ±280 |
| Year | 18,500 | 21,340 | +15.4% | ±670 |
Illustrative figures from a demo build. Not a published result.
No spreadsheets.
No guesswork.
Instrument the site
Towers, power and backhaul. Two weeks on a site with road access, longer where a helicopter is involved.
Establish the baseline
A season of readings before any claim is made, so the counterfactual is measured rather than assumed.
Run continuously
Hourly aggregates, monthly reconciliation, and an alert when a sensor drifts outside its calibration envelope.
Hand over the record
Figures, bands and provenance, exported whole. If you leave, the series comes with you.
Bring us the figure your auditor keeps circling.
We will tell you whether it can be measured, what that would cost, and how long it takes.
Let's talk