Captain Drawdown’s daily logbook on every CDR story, paper, and expert voice — so you don’t have to read them all.
Peak warming moves by at most about 0.025 °C. That is what novel carbon removal plus geological storage buys by 2050 in a new preprint, even when the model is told to assume the best.
The critique, in the authors’ words
The paper is Merfort, de Jonge-Anderson, Dorndorf et al., on Research Square, titled ‘Limited contribution of novel Carbon Dioxide Removal and Geological Carbon Storage to reducing peak warming’. Its abstract states that ’even under highly optimistic assumptions, geological carbon storage and novel carbon removal contribute only 32–51 GtCO₂ cumulatively by 2050’, a total it benchmarks at ‘approximately one year’ of global emissions.
Note what this is and is not. It is an integrated-assessment pathway analysis, not a rebuttal written against the industry. But it is the sharpest number currently in circulation against one specific claim, that durable removal matters for how high warming gets this century. It is also a preprint, and the range may shift under review.
The numbers behind it hold up
Durable removal contracting went from 0.6 Mt in 2022 to roughly 30 Mt in 2025, about a 50-fold rise. Compare it on equal terms: the preprint’s low end of 32 Gt spread over 25 years averages about 1.3 Gt a year, so 2025’s contracted volume is roughly 2% of that annual rate, and contracts are not deliveries.
Two further constraints push toward the bottom of the range. Sequestration well applications are piling up at the EPA, with lawmakers weighing a 180-day mandate for primacy decisions. And a multi-model assessment of marine CDR under overshoot centres on efficiency and carbon-cycle feedbacks, which means delivered tonnes come in below gross tonnes. Storage permitting and net efficiency bind before chemistry does.
The standard defence agrees with the paper without noticing
The usual reply is that peak shaving was never the point. Carbon Brief (@carbonbrief.org on Bluesky) reports that a temporary breach of 1.5 °C is now “inevitable”. If overshoot is locked in, removal should be judged on bringing temperatures back down afterward.
The weakness is that this is not a rebuttal. The preprint models to 2100 and finds cumulative contributions rise to 300–875 GtCO₂, corresponding to 0.14–0.39 °C of post-peak temperature decline. The defenders’ strongest card is already in the authors’ hand. The other reply, pointing at 50-fold growth, cites a growth rate when the question is a level.
What the paper gets right
Novel removal is not a lever on peak warming. Nobody pitching, buying or regulating it should imply otherwise this half-century.
The 32–51 Gt figure also reads as a ceiling, not a midpoint. The model’s assumptions are optimistic by design, while the permitting queue and efficiency losses above are real-world headwinds the model does not have to clear. Chris Bataille (@chrisbataille.bsky.social on Bluesky), writing about power-sector capture choices rather than this paper, describes the failure mode when capture bets underdeliver: “Instead we have the unicorn of post combustion CCS, & when dreams fail, a return to unabated combustion.” His context is narrower, but the risk transfers. Storage capacity that does not materialise becomes emissions, not a smaller number on a chart.
Where the framing overstates
The title foregrounds a null result on a test the sector should never have claimed to pass. The more consequential result in the same abstract is the 2100 range, and that range is wide, 300 to 875 Gt, which tells you the deciding variable is build-out and policy, not physics. A 2050 cumulative total also measures a stock problem as if it were a flow. Capacity that exists in 2100 has to be built, permitted and proven before 2050, so a small mid-century number is not evidence that mid-century deployment is low-value. The paper does not say it is. Readers of the headline alone will.
What builders should take from it
Sell post-peak durability and stop implying peak impact. Dated portfolios like the CUR8 and Isometric 2030 offering sell delivery timelines, which is fine as long as no one reads a 2030 vintage as a 2030 temperature effect. The place to check whether buyers draw that line is Microsoft’s six-year lessons-learned paper, the largest procurement record in the sector.
Two things to watch. Whether the 32–51 Gt range survives peer review intact, and whether the Class VI backlog clears fast enough for the 2100 numbers to stay reachable. The physics leaves removal a real job after the peak; permitting decides whether anyone shows up for it.
Citations
- Researchsquare — Merfort, de Jonge-Anderson, Dorndorf et al., on Research Square
- Substack (jeetjoshi) — 0.6 Mt in 2022 to roughly 30 Mt in 2025 — Substack post
- Congress — weighing a 180-day mandate for primacy decisions — government source
- DOI-resolved paper — multi-model assessment of marine CDR under overshoot
- Bluesky — @carbonbrief.org on Bluesky — Bluesky post
- Bluesky — @chrisbataille.bsky.social on Bluesky — Bluesky post
- Microsoft — Microsoft’s six-year lessons-learned paper — PDF