Three numbers nobody has

Public figures exist for the floats only: Argo at least $40 million a year, OneArgo about $100 million, BGC-Argo about $25 million. There is no public figure for GO-SHIP repeat hydrography, for the ship-of-opportunity pCO₂ lines, or for the moorings, and none for the system as a whole, which nations already support together. A literature search of 53 references and 34 programme sources found nothing more, and a novelty check of the peer-reviewed and grey literature confirmed it: the fragments that exist (a measurement-team budget here, a per-ship instrument cost there) have never been assembled, and no one has set the cost of each network next to the uncertainty it removes.

The global sink is known to about ±0.4 GtC a year on about 3 GtC a year, and its trend since 2001 is disputed between observation-based products and models. Ship-based data sets fell by about 35% between 2017 and 2021, and the spread among flux estimates rose by 65%. Against that background, local errors in the global products are more than ten times the signal an open-ocean carbon-removal project would produce, about 1 µatm.

Many documents describe what MRV needs in words. None states the accuracy and density in numbers, maps them onto what today's networks deliver, and ranks the technology advances by the gap each one closes. Since 9 October 2026, when the parties to the London Convention and Protocol adopted assessment frameworks for research on ocean fertilisation, alkalinity enhancement and biomass sinking, every research experiment is expected to show rigorous monitoring, which in practice means telling its effect apart from the background ocean.

Monitoring, reporting and verification can exceed half the cost of some removal techniques; a cost model for ocean iron fertilisation finds that verification alone raises the cost per tonne three to four times. Every project builds and pays for its own MRV, background ocean included. The idea of treating the background as a public good is now voiced by many groups, but nobody has priced a shared system against every project measuring for itself, or found the scale at which the shared system pays for itself. Earlier studies valued better information; none costed the infrastructure.