The carbon numbers are being re-measured

The canopy is lying

Three preprints posted to EarthArXiv this week do the same job in three different places: they re-measure carbon, and in each case the old measurement turns out to be shakier than people assumed. One looks at forests, one listens to compost, and one reads the chemical traces left in ocean mud. None of them is peer-reviewed yet. Taken together, they make a simple point for anyone who buys, sells or reports carbon: the numbers underneath the market are being rebuilt, and some of the proxies everyone relied on do not survive the check. m:…8265 m:…4242 m:…1584

Start with the forests. In a preprint, researchers apply deep learning to 3D point-cloud data to produce high-resolution time series of forest biomass and carbon stocks — a direct read on what is actually there, rather than an estimate from above. The direct read disagrees with the shortcut. The preprint reports that canopy height is an unreliable proxy for forest carbon stock changes in managed forests, and that biomass loss often occurs in forests with steady or increasing canopy height. A forest can lose carbon while its canopy looks fine. m:…8265

The implication for anyone who buys forest carbon credits is blunt: if canopy height cannot be trusted as a stand-in for stored carbon, then a credit or a report built on that proxy needs a second look. The preprint does not claim all forest carbon accounting is wrong. It claims the shortcut fails in managed forests — and that a forest can lose biomass while its canopy holds steady. m:…8265

Listening to rot

The second preprint listens instead of looking. It tests whether sound can tell the phases of organic decomposition apart — the stages material passes through as it breaks down — and reports that nonlinear acoustic complexity measures do the job better than the conventional ecoacoustic indices. The numbers in the preprint: a median predictive skill of 0.48 for the nonlinear measures, against 0.12 for the conventional ones. And one frequency band stands out: 128-256Hz, which yields the highest predictive skill, a median of 0.70, for identifying decomposition states. m:…4242

That is a proof of concept, and the preprint says so in its title — a small, controlled test, not a field-ready monitor. But the direction is clear enough to matter. Decomposition is the process that decides how much carbon stays in the soil and how much returns to the air, and an acoustic read on which phase it is in would be a new kind of sensor for it. What the preprint offers is evidence that the signal is there, in a specific band, at a specific strength — and that the standard indices miss most of it. m:…4242

Reading the mud

The third preprint reads the record rather than the present. It works with glycerol tetraethers — chemical traces preserved in marine sediments — and reports that the ratio of isoGMGTs to crenarchaeol serves as a tracer of relative soil organic matter input in marine sediments. In plain terms, the chemistry of the mud keeps a record of how much soil carbon washed into the sea. The same preprint also reports that marine-derived brGMGTs, known as H1020a and H1020b, are present in oxygenated marine water columns — a detail that matters because it pins down where those traces come from. m:…1584

What to do with this

None of these three preprints is a finished tool. The forest study is a re-measurement, not a new market. The acoustic study is a proof of concept. The sediment study traces carbon that moved in the past, not carbon being counted today. But the pattern across them is the story: carbon accounting is moving from proxies and estimates toward direct measurement — point-cloud scans of trees, sound signatures of decay, chemical fingerprints in mud. For a professional who relies on carbon numbers, the question to ask of any figure now is not whether it is big or small, but what it was actually measured with. m:…8265 m:…4242 m:…1584