The Data Problem: When the Numbers Are Estimates
The entire framework — pricing physical risk, committed emissions, resource adequacy, carbon incidence — rests on numbers that are far softer than their decimal points suggest. Much of what the market treats as measured fact is estimated, self-reported, or modelled. That gap is the last strain, and the most investable.
They drew the map in a fair and steady hand, / and marched by it, and trusted every line; / but no one walked the land to see it true, / and the marsh was where the road was meant to shine.
The Numbers Are Softer Than They Look
Every piece in this framework runs on data — emissions figures, reserve estimates, resource volumes, climate projections, asset performance. And almost all of it is reported to two or three significant figures, which invites the reader to treat it as measured. It is mostly not measured. It is estimated, modelled, or self-declared, and the gap between the reported number and the real one is the quiet strain underneath everything else.
This matters because the whole discipline of the series is pricing what markets misprice, and markets misprice what they cannot measure. A carbon price rides on an emissions number. A stranding estimate rides on a committed-emissions number. A physical-risk premium rides on a climate-model output. If those underlying numbers carry a large, systematic error — not random noise that averages out, but a consistent bias in one direction — then everything built on top of them inherits the error, and the market is confidently pricing a figure that is simply wrong.
Data uncertainty is not a footnote to the transition; it is a hidden, investable inefficiency. Where reported and real diverge systematically, there is mispricing — and therefore both a risk (greenwashing, regulatory catch-up, stranded-asset surprise) and an opportunity: the edge accrues to whoever can measure what everyone else estimates.
The Methane Case
Methane is the cleanest illustration because the gap is now measurable. It is also the case that matters most: methane has roughly eighty times the warming power of carbon dioxide over twenty years, and cutting it is the single most cost-effective near-term climate lever there is. Yet the fossil-fuel sector emits an estimated 124 million tonnes of it a year — oil 45, coal 43, gas 36 — and the number is rising even though the abatement is cheap and proven. The reason nothing moves is partly that, until recently, no one could see the true scale.
The mechanism of the error is instructive. Countries and companies estimate methane bottom-up: count the facilities, multiply by a standard leak rate. Satellites and aircraft measure it top-down, and they keep finding that a handful of super-emitter leaks — a stuck valve, an unlit flare, a blowout — dominate the total and are almost entirely absent from the factor-based inventories. The same average also hides enormous variation between producers, which is itself a data point: a single global figure is nearly meaningless for pricing any specific barrel.
We Built the Instrument, and It Died in Orbit
The measurement gap is closing, but the story of how tells you how hard and how fragile the work is. In March 2024 the Environmental Defense Fund launched MethaneSAT, an $88 million satellite — one of the most advanced ever flown — built specifically to see the methane the inventories miss, sensitive to changes of three parts per billion and able to catch both super-emitters and the diffuse sources that had been invisible from space. It worked. Then, on 20 June 2025, fifteen months into a five-year mission, it lost power and went silent, and was declared unrecoverable.
The loss is not the end of the point — it is the point. The capability was proven, the algorithms and data live on, and a growing constellation of other instruments continues the work. But the episode captures the strain exactly: seeing the truth clearly enough to price it is expensive, technically fragile, and only partly done. Until it is finished, the reported number and the real number will keep diverging, and the divergence is where the mispricing lives.
The Problem Is General
Methane is the sharpest case, not the only one. The same structure — a reported figure treated as fact, resting on estimation that carries a systematic bias — runs through most of the data the transition is financed and regulated on.
| Data domain | What gets reported | Why it is uncertain |
|---|---|---|
| Methane emissions | Factor-based national and corporate inventories. | Measurement finds ~80% more; super-emitter leaks are missed entirely. |
| Carbon offsets | “One tonne avoided or removed.” | Baselines and additionality are unverifiable; chronic over-crediting is documented. |
| Reserves & resources | Oil, mineral and groundwater estimates. | Self-reported, sometimes political, model-derived — with wide error bars. |
| Physical-climate risk | Asset-level flood, heat and fire scores. | Model spread is large; a point score hides a wide distribution. |
| Corporate Scope 3 | Self-reported value-chain emissions. | Estimated, inconsistent and largely unaudited. |
In each row the reported number is precise-looking and the real number is a distribution. The danger is treating the first as the second — buying an asset-level climate risk score as though it were a measurement, or a carbon credit as though a tonne had been verified. The discipline is to ask, every time a clean figure appears, whether anyone actually walked the land to check it.
The Positioning Read: Measure What Others Estimate
If the gap between reported and real is a systematic inefficiency, then it is tradeable in both directions — own the tools that close it, and reprice the assets whose reported numbers the closing will expose.
Measurement & verification
Satellites, sensors, aerial surveys and digital MRV are the picks and shovels of a data-scarce transition. As disclosure and regulation tighten, the ability to verify emissions and performance becomes a priced service, not a cost centre.
Assets where reported < actual
Producers whose real methane or emissions intensity exceeds what they report face regulatory catch-up — EU methane rules, waste-emissions charges, import standards. The gap is a latent liability the market has not yet marked.
Physical-risk scores
Do not buy asset-level climate risk as a point estimate. The disagreement between models is information; the spread, not the midpoint, is what should size the position and the premium.
Books built on self-reported data
Offset portfolios and Scope 3 targets underwritten on unverified numbers are exposed to a measurement revolution that can revalue them overnight. Discount claims that no independent instrument has checked.
The through-line of the whole Strain thread has been that the build runs into physical limits the market prices badly. This is the limit underneath the others: the limit on what we actually know. You cannot price heat you have not measured, water you have not gauged, or emissions you have not seen. The transition is being financed on a map drawn in a fair and steady hand — and the single most valuable act in the space may simply be to walk the land and check that the road is where the map says it is.
This note sits under the whole framework. It sharpens The Climate Clock (model uncertainty in the warming path), underlies Forestry & Offsets (whose credibility crisis is a measurement crisis), and qualifies Committed Emissions and Carbon Pricing — both of which price a number this note argues is uncertain. It completes the Strain thread: eleven ways the build meets a limit, ending with the limit on what can be known.
The framework runs on numbers, and the numbers are estimates dressed as measurements. Methane is the proof: energy-sector emissions run about 80% above what governments report, a systematic bias in the single most important near-term climate lever — and the satellite built to see it clearly died in orbit fifteen months in. Treat every clean figure as a distribution, own the tools that close the gap between reported and real, and reprice the assets that gap will expose. With this, the Strain thread closes: the last and deepest limit the build runs into is the limit on what we actually know.
Count not the harvest by the promise sown, / nor trust the tally that was never weighed; / for what is written is not what is grown, / and the ledger lies until the field is surveyed.
Leave a Reply