Three enhanced weathering papers landed on CDRXIV in close succession, and read together they give me the clearest snapshot I’ve had of where this pathway actually stands in late 2025. The Suhrhoff et al. systematic review — with a co-author list that reads like a who’s-who of the field — puts a median area-normalized CDR flux of 0.84 tCO₂ ha⁻¹ yr⁻¹ on the table and converges global technical potential at ~0.2–2 GtCO₂ yr⁻¹. Cobo and Guillén-Gosálbez then argue that if you take empirical (rather than theoretical) weathering rates seriously in a life-cycle assessment, inland deployments starting in 2030 face a median 11-year lag before net CDR turns positive. And Gholamahmadi et al. offer the first structured synthesis of how ERW changes the soil’s ability to hold and move water — with mixed results that complicate the “co-benefits” story.

What was already on the table

The status quo going into 2025 was a pathway with strong theoretical promise and thin field evidence. The recent Nature Reviews Earth & Environment piece framed scaling challenges; a Communications Sustainability paper put maximum deployment at 0.76–1.1 GtCO₂ yr⁻¹ by mid-to-late century. But field trials have been landing harder than models predicted — a large basalt trial covered by C&EN reported CO₂ removal at less than 10% of expected rates, and a Frontiers in Climate midwestern trial found evidence for CDR from steel slag but not basalt. Project Drawdown’s own explorer has flagged “unpredictable effectiveness” as a top risk. So the theory-vs-field gap was known; these papers quantify what it means.

What’s genuinely new

Three things stand out to me.

First, Suhrhoff and colleagues normalize the sprawling literature into comparable units and make the variance itself the message: reported fluxes span several orders of magnitude, driven not just by real context-dependence but by differing system boundaries and treatment of loss processes (secondary phase formation, cation exchange). That framing is useful because it reframes “why don’t field trials match models?” as partly a measurement-convention problem, not only a physics problem.

Second, Cobo and Guillén-Gosálbez’s 11-year median carbon payback is the first LCA I’ve seen that anchors its weathering kinetics to experimental rather than theoretical rates. Their probabilistic toxicity finding is the sharper surprise: basalt-derived zinc may pose greater long-term toxicological risk than nickel from dunite — inverting the common assumption that basalt is the safer feedstock. Their recommendation to prioritize non-agricultural deployment until field trials better characterize food-chain risks is worth sitting with.

Third, Gholamahmadi et al. give the “soil co-benefits” narrative its first real audit. Across 17 studies and 261 observations, field capacity (+9.4%) and plant available water (+10.1%) rose modestly, but saturated hydraulic conductivity fell 27.5% and porosity dropped 6.3%. Bulk density didn’t move. So ERW doesn’t uniformly enhance the “soil sponge” — it appears to trade some flow properties for some storage properties, context-dependent.

Who this touches

The empirical-rates finding matters most for companies selling near-term credits from inland silicate application on croplands: Lithos, InPlanet, Terradot, and Alt Carbon all operate in that space, with basalt prominent in several portfolios. The Cobo paper doesn’t name companies, and I won’t over-read it — but its policy implication (evaluate near-term net climate benefits before commercializing credits) intersects directly with how MRV protocols count upstream emissions against delayed drawdown. CREW Carbon and CarbonRun, which work in river/alkalinity-enhancement modes rather than cropland spreading, sit somewhat outside the specific inland-agricultural framing here.

What I’m watching

Whether the toxicity signal on zinc from basalt gets replicated in field measurements, not just LCA modeling. Whether MRV frameworks start requiring disclosure of expected payback lags at the project level. And whether the Suhrhoff review’s push toward standardized system boundaries actually gets adopted — because until the field agrees on what to count and where to stop counting, “0.84 tCO₂ ha⁻¹ yr⁻¹” will keep meaning different things to different buyers.

The papers

I read the full text of each. Context drawn from prior literature and public discussion; company links where the connection is clear.