Measured, not assumed

A surprising number of the figures we act on were never measured. They were assumed: pulled from a standard factor, a model, or a round number that got copied forward until it looked like fact. Usually nobody checks, because checking is harder than trusting, and the assumed number is good enough to keep the meeting moving.

Ground Truth is about the distance between that assumed number and the one you get when you actually measure. It tends to run wider than anyone admits, and it hides where being wrong costs the most: the emissions we claim to cut, the risks we sign off on, the numbers a whole system quietly leans on. Three examples of the same shape, each with the source to check it.

Nitrous oxide from wastewater and farms

Nitrous oxide (N₂O) warms the climate roughly 270 times as much as CO₂ per tonne, and most of what a farm or a treatment plant emits is never measured directly. It is inferred from an emission factor: a fixed fraction of the nitrogen input, assumed to escape as N₂O. For agricultural soils the IPCC default is 1% of applied nitrogen, yet the official uncertainty on that single number runs from 0.1% to 1.8%, an eighteen-fold spread on the figure whole national inventories are built from. Real emissions arrive in short “hot moments” after rain or fertiliser that a flat annual factor cannot see. Wastewater tells the same story: a 2024 study in Nature Sustainability that expanded the measurement database twelvefold found the uniform factor systematically misestimates real plant emissions, which swing with process, scale, and season.

Methane from oil and gas

Oil and gas methane leakage was long reported bottom-up: count the components, multiply each by an assumed leak rate, add it up. Then researchers measured the supply chain directly. A 2018 assessment in Science put 2015 US oil and gas methane near 13 million tonnes a year, about 60% above the EPA inventory, and traced the gap to abnormal operating conditions the component-by-component method was never built to catch: a small population of “super-emitters” doing most of the leaking. The assumed number and the measured number were describing different worlds through the same pipes.

Carbon credits

A carbon credit is a claim: one tonne of emissions avoided or removed. Often that tonne is modelled rather than measured, a baseline of what supposedly would have happened, set against what a project reports. Baselines meet measurement badly. A 2023 study in Science of voluntary rainforest offset projects (REDD+, which pay to avoid deforestation) found most did not cut deforestation as claimed, and the credits sharply overstated the real benefit. The exact share is contested (press analyses put the worthless fraction above 90%, the registry disputes the method), but the direction is not: the market was trading assumed tonnes as though they were measured ones.

The method, held constant

That is the whole thing here. Take a number a system depends on. Find the claim behind it, and the measurement that should back it. Put the two side by side, quantify the gap, and keep the uncertainty visible instead of rounding it away. The sources sit in the text, so you can run the same check. Sometimes the gap is small and the claim holds. Sometimes it is not. Either way the point is steady: a number you can check beats a number you have to trust.