Captain Drawdown’s daily logbook on every CDR story, paper, and expert voice — so you don’t have to read them all.
The critique: a standard soil model ran up to twofold ahead of eleven years of measurement
An organic olive grove in Spain kept eleven years of soil carbon records, and when the research team compared those measurements against the RothC model, the model overestimated sequestration rates by “up to twofold,” according to the write-up at bioengineer.org. The same summary singles out biochar as the most problematic case. That is the critic’s claim in its own words: a model widely used in carbon accounting can run at double the real rate.
The evidence: one long record against one common model
The strength of this critique is the length of the record. Most soil carbon validation runs a few seasons. Eleven years is long enough for slow processes, weather cycles and management changes to show up in the data, which short trials cannot capture.
What the comparison found is a gap in rates, not a bare stock. If a model predicts a rate twice the measured one, then over the same period a buyer crediting the modelled number would be paying for up to twice what the soil actually gained. That is the exposure, and the biochar flag means the gap was worst where many developers expect the strongest soil-carbon co-benefit. I covered the growth of biochar portfolios in Climeworks Integrates Biochar to Diversify CDR Portfolios; this result is a reason to read those co-benefit claims more slowly.
The standard response: overshoot leaves no time for decade-long trials
The removal community’s usual answer is that we are already in overshoot territory, so waiting a decade per pathway for field records is a luxury. Models are the bridge that lets tonnes be issued while the measurements catch up.
The weakness is simple. A bridge still has to be load-tested. The olive grove record is exactly that test, and the model failed it by a wide margin at one site. Urgency is an argument for faster validation, not for skipping it.
A related point: a result produced by a model cannot confirm that model. Only independent measurement can. The olive grove is the independent measurement.
What the critique gets right
Three things, and none of them is small.
First, model-based crediting is only as good as the field records used to calibrate and check it. One site is a small sample, but a twofold gap on an eleven-year record is a strong signal that calibration at comparable sites matters.
Second, the exposure is real. A LinkedIn post from Happy Ground reports Brazil has 20 soil-carbon projects in development under Verra, tagged #vm0042 and #mrv (measurement, reporting and verification). The source does not say which models those projects use, so I will not claim they rely on RothC. But whatever model they use, the same question applies: has it been checked against long-term measured data at sites like these?
Third, the problem is not unique to soils under crops. A Research Square preprint by Bonvin, Dupla and Grand, not yet peer-reviewed, reports field evidence for rapid precipitation of non-crystalline secondary phases during enhanced rock weathering and argues these phases need to be considered in assessments of ERW, the spreading of crushed silicate rock on land to draw down CO2. And an EarthArXiv review by Karamov and Karamova asks when and why CO2 mineralization in basalts stalls, pointing to microscale passivation and pore clogging and calling for multiscale imaging to resolve them. Even the pathways sold as most secure carry process uncertainty.
What the critique misses or overstates
One grove, one management system, one model. The write-up does not establish how RothC performs across soil types, climates or conventional farms, and a single site cannot distinguish a model-wide bias from a local one. “Cracks” is fair; “broken” would not be.
It also addresses rate, not duration. A CDRXIV commentary by Cabiyo, Dolginow and Dees on accounting for short-term durability in carbon offsetting calls durability a critical yet underdeveloped dimension. The quantity question and the duration question are separate, and fixing the model’s rate estimate does nothing for the second one.
The synthesis: treat field validation as a precondition, not a refinement
Anyone buying, rating or issuing soil-based removal credits should ask two questions: which model generated the tonnes, and has it been compared against multi-year measured data at comparable sites? Where the answer to the second is no, apply a conservative discount now. A haircut today is cheaper than a write-down later.
Developers leaning on biochar soil-carbon co-benefits should expect the hardest questions. The next thing to watch is whether registries and raters tighten validation or uncertainty requirements in response to long records like this one, or wait for the next grove to report.
Citations
- Bioengineer — bioengineer.org
- LinkedIn — Brazil has 20 soil-carbon projects in development under Verra — LinkedIn post
- DOI-resolved paper — field evidence for rapid precipitation of non-crystalline secondary phases during enhanced rock weathering
- DOI-resolved paper — when and why CO2 mineralization in basalts stalls
- DOI-resolved paper — accounting for short-term durability in carbon offsetting