Authors: Matthew Clarkson, Christina Larkin, Junyao Kang
Acknowledgements
We gratefully acknowledge the support and contributions of the many individuals that made this deployment a reality including the mining and farming partners, Antonio Azevedo, Mariane Chiapini, Jessica Ferrarezi, Marcella Daubermann, Veronica Furey, Niklas Kluger, David Manning, Murilo Nascimento, Elisabete Pedrosa, Igor Nogueira, Bruno Ramos, Felipe Reis, Mayra Maniero Rodrigues, Leticia Schwerz, Karla Nascimento Sena, Philipp Swoboda, Jeandro Vitorio
Reviewed by: Zeke Hausfather, Frauke Kracke
Introduction
Robust and transparent Measurement, Reporting, and Verification (MRV) is paramount for establishing the credibility of the nascent Enhanced Weathering (EW) industry. For EW, translating theoretical geochemical principles to a complex real world deployment can present both challenges, but more importantly opportunities for learnings and improvement. Here, we report on the methodological learnings from our first large-scale EW deployment.
In April 2025, InPlanet achieved another milestone, our first credits were delivered to the Frontier buyers coalition, stemming from a pre-purchase agreement awarded in Fall 2022. It was part of their second set of carbon removal purchases, and it was the second enhanced weathering (EW) purchase made by the group. The purchase, from the $1B+ advance market commitment, directly funded our first deployment, Project Beija-flor. Project Beija-flor is based in São Paulo state, Brazil, and was designed to remove CO₂ while regenerating degraded tropical pasture soils. The project was instrumental in refining the MRV framework that ultimately led to the issuance of the world’s first third-party verified EW credits by Isometric in late 2024, which were associated with a subsequent deployment (Project Serra da Mantiqueira). This blog post documents the learnings as well as the solutions implemented during the foundational Project Beija-flor. The piece is intended for those with some technical understanding of EW measurement in order to help inform others looking to implement EW, and builds upon our technical review paper published last year1.
Project overview
Project Beija-flor encompasses approximately 900 hectares across 13 farms, with basalt powder applied at a rate of 10 tonnes per hectare (t/ha). The deployment targeted pasture, a biome where rock powders can potentially restore degraded soils through nutrient addition and pH adjustment. Brazil contains an estimated 160 million hectares (Mha) of pasture2, highlighting the immense scaling potential for EW in this context. Participating farms ranged in size from 10 to 400 ha, enabling engagement with both smallholders and larger commercial operations.
The feedstock, a basalt powder certified as an agricultural remineraliser in Brazil, was sourced as a byproduct from a local quarry. Reactive minerals include Andesine (42%), Augite (25%), Albite (10%) and Orthoclase (7%), contributing to a gross CDR potential of 0.268tCO2/trock. The powder has a D50 of 111 µm.
To minimise transport-related emissions3, all deployment sites were situated within a 100 km radius of the source. The surface application of the feedstock using conventional agricultural spreaders occurred between August 2023 and November 2024. Significant logistical delays at several farms extended the spreading interval, necessitating a phased delivery of credits to Frontier.
Monitoring design
Our initial monitoring framework was centred on a solid-phase mass balance approach, combining broad-scale soil sampling (one sample per 10 ha) with high-intensity monitoring at two dedicated field monitoring stations (FMS). Our measurement and calculation approaches are aligned with the Isometric protocol, although the sample density was lower than the recommended 1/1ha. Learnings from early operations prompted a refinement of this protocol, increasing the standard sampling density to one sample per hectare to help reduce uncertainty. Although 1 sample per 10 ha, in the case of our farms, captured the variability across the field scale, a tailored approach is needed for different soil types and regions. In addition, when crediting at one standard deviation below the mean, lower uncertainty in the measurement results in a lower uncertainty discount, therefore increasing sampling density helps increase the certainty, and therefore number of credits issued.
We incorporated data and learnings from our first pasture FMS, based at Luiz de Queiroz College of Agriculture, University of São Paulo (ESALQ) as part of a joint research grant funded by the Grantham Foundation, in collaboration with Profs Antonio Azevedo (ESALQ) and David Manning (University of Newcastle), for which we gratefully acknowledge the support of.
The monitoring protocol integrated two primary measurement strategies:
Solid-Phase Measurement: This approach quantifies net CDR by first calculating the gross potential dissolution of the applied rock powder4–6 using shallow samples at 0–5cm. Deductions are then made for system losses, including cation sequestration on soil exchange sites (0–20 and 20–40 cm depth samples), strong acid losses (to 40cm depth) and uptake by vegetation.
Liquid-Phase Measurement: This method provides a direct validation6 by measuring the flux of bicarbonate (the aqueous product of weathering) in soil water (leachate) at 40 cm depth.
While solid-phase samples were collected across the entire project area (1/10ha), liquid-phase sampling was concentrated at the FMSs. Here, leachate was collected approximately bi-weekly, at 20 cm intervals down to 1 m depth using suction-cup lysimeters within a randomised controlled plot design. At these FMS sites we also monitored the soils down to 1 m depth, and carried out soil characterization with a 2m soil pit. This was supplemented by lower-resolution liquid sampling across a subset of farms. In addition to this, we maintained control areas which were monitored in the same manner, but rock powder was not applied and the farmers maintained business as usual.
Results
CDR calculations were made following the approach of the Isometric protocol, conservatively crediting below the mean (16th percentile) for the gross CDR potential and, and using loss values above the mean (84th percentile). Calculation results are summarised in Table 1. Operational emissions are amortized equally across the first two years of the project.
The deployments achieved a net CDR of 81t, after accounting for Near Field Zone (NFZ) and Far Field Zone (FFZ) deductions. This equates to reaching ~40% of the total gross CDR potential of the rock powder within the first year of weathering.
There was no significant change in biomass yield or cation concentration during the reporting period, and hence no biomass losses were accounted for. Soil carbonates do not form in this environment due to high rainfall and low pH soils, again excluding this loss term. Strong acid losses, due to non-carbonic acid dissolution of the rock powder, were minimal due to the lack of widespread fertilization of the farms and because the cations charged balanced dissolved inorganic carbon (DIC) in measured pore waters. Despite the relatively low soil pH (range 5.8-6.1), soil pore waters had a higher pH (average pH=6.7), and showed stable bicarbonate concentrations. Additionally, there was no displacement of liming activity and hence no counterfactual losses.
The only significant NFZ loss term applicable to the deductions related to soil exchangeable phases. The temporary retention of cations in exchange sites is a well established concept in both soil science and natural weathering studies7 and an observed feature of ERW experiments8–11. This cation store is a temporary sink in the NFZ, resulting in a temporary deduction under the Isometric protocol, which acts as a safeguard against overcrediting.
For this deployment, ~175t CDR equivalent was ‘lost’ due to a statistically significant increase in control-corrected exchangeable Ca concentrations at 20–40cm, relative to baseline conditions (p=0.009; Wilcoxon signed rank test). This represents a deduction when calculating net CDR for the reporting period (Table 1). We emphasise that these loss terms are likely to be site-specific as we did not observe any significant changes in cation exchange in the first year of Project Serra da Mantiqueira. We hypothesise that this difference is due to the larger and significant plant uptake of basalt derived cations in sugarcane, a large and fast growing crop, compared to pasture, where no significant changes were observed.
Hypothetically, a later release of the exchangeable cation store, during a subsequent reporting period, would result in a reduction in the loss term for the later CDR delivery, allowing the temporary deduction to be claimed back. With this approach, there is no need for ex-ante projections of lag times and future credit availability, rather they are dealt with empirically in an ex-post manner across multiple crediting events. Multi-year deployment data are required to test this hypothesis.
Major learnings
From this deployment, there were several learnings, which we break down into 3 sections covering: 1) control and treatment designation, 2) solid phase measurement and 3) validation with additional measurements (liquid phase).
Control and treatment comparability
Control (13% of total area) and deployment areas were designated based on soil mapping, but final placement was subject to practical constraints, including farmer agreement and participation. This resulted in a non-contiguous control design where some smaller farms lacked dedicated control plots. In future deployments we focus on larger farms that are better able to manage a control plot in closer proximity to the treatment plots.
Despite this, a comparative analysis demonstrated that the aggregated control plots broadly captured the range of baseline agronomic conditions present in the treatment areas, including soil texture, pH, cation exchange capacity (CEC) and base saturation (Table 2).
We note that whilst the control sites were situated on farms with lower sand content compared to the treatments, the mean CEC is more comparable (Table 2). A wider CEC range would be expected to result in more variable exchange processes, and temporary CDR losses, however we did not observe any trends to suggest systematic differences in CDR performance across the CEC range.
Solid phase measurements
Rock powder resolvability
CDR quantification was performed using the immobile trace element (ITE) method4,5, which reconstructs feedstock dissolution by tracking the change in concentration of mobile cations relative to an immobile element present in the applied rock. This method’s efficacy is conditional on the geochemical distinctiveness of the rock powder relative to the baseline soil.
For this delivery, a statistically significant increase in Titanium (Ti) was observed in treatment sites post-application (p=0.046, one tailed Wilcoxon signed-rank test) but not in control sites (p=0.371, one tailed Wilcoxon signed-rank test), confirming Ti as a resolvable tracer for the applied basalt. However, ITE resolvability was found to be problematic in clay-rich soils, which exhibited high baseline concentrations of Ti and other potential ITEs (Fig. 1). This high background concentration masked the signal from the applied rock powder. Consequently, two farms with particularly high clay content were excluded from the final CDR delivery, as weathering rates could not be confidently determined. Future deployments in such soils may necessitate a total cation inventory approach, foregoing ITE-based calculations. Nevertheless, without an immobile tracer, it is not possible to account for physical movement or loss of the rock powder affecting the measured loss of cations. Future research will cross-validate these two methods.
Figure 1. Cross plot of TiO2 versus SiO2 in the soil at 0-20cm and 20-40cm depth. Ti concentrations are negatively correlated with SiO2 (a proxy for sand content) indicating a decreasing ability to resolve ITEs as clay content increases.
This resolvability issue is exacerbated by sampling depth, as deeper sampling depths effectively dilutes the signal of the rock powder. The project’s initial 0-20 cm sampling protocol was later revised to 0-5 cm to concentrate the rock powder signal (with surface application and shallow integration). This adaptation proved essential for resolving the weathering signature, although it introduces a potential risk of sampling noise for this specific delivery, as the baseline samples at 0-20 cm were compared with a post deployment sample at 0-5 cm. Nevertheless, we checked depth variations in ITEs in these soil types and saw no difference between 0-5 and 0-20 cm samples. Moving forward, we include 0-5 cm samples at baseline to ensure that soil depths are comparable across all sampling intervals.
Correction for background cation addition
A key assumption in EW MRV is that control plots can be used to quantify the background loss of cations from natural weathering. However, in this project, as in the subsequent Project Serra da Mantiqueira, an increase in cations was observed at the control sites. This finding points to external cation inputs, which is likely in farmland due to fertilizer inputs, such as phosphate or lime, which contain cations. In the case of pasture, the hypothesised source is cation addition from intensive cattle grazing across the sites, as livestock effluent will add cations to the soil. This is a key learning, as previously project developers and other scientific stakeholders have assumed the control site would have a stable or decreasing cation concentration. However, in farmland under standard management, including fertilization, the cation concentrations in the control and deployment areas will likely increase due to additional inputs.
Provided both sites are treated in the same manner, this can be accounted for using the control sites. Therefore, to account for this external input of cations, a conservative correction was applied. The change in cations at control sites was used to adjust the deployment data, preventing the misattribution of these external inputs as feedstock-derived weathering. It is important to examine evidence for the applicability of the background correction, confirming that the control and treatment are indeed managed in the same manner. In this case, it was unrealistic to track cattle grazing directly beyond farmer land use reports, however evidence from liquids samplers in both the control and treatment areas confirm an exogenic source of cations. This correction results in an increase in net CDR but is important in order to account for background farm activity. To ensure a conservative estimate and mitigate the risk of over-crediting, the lower of the mean or median cation change from the control dataset was used for this correction.
Data heterogeneity and outlier management
Soil systems are inherently heterogeneous, leading to noisy datasets. Mass-balance calculations for the fraction of dissolved rock powder (Fd) can consequently yield non-physical values (Fig. 2; Fd < 0 or > 1). While these values are part of the true uncertainty, extreme outliers can disproportionately skew the mean dissolution estimate5.
Figure 2. Raw dataset (upper) showing calculated Fd for Mg and Ca, illustrating the potential for non physical values due to system noise. Bootstrapped results (lower panel) removes the extreme outliers and we conservatively credit based on the 16th percentile to avoid any risk of over crediting.
To address this, a standard 3-standard-deviation threshold was employed for outlier removal during the bootstrapping uncertainty analysis (n = 10,000 simulations). This common statistical method proved to be a conservative approach; for our dataset, 80% of the simulation replicates required the removal of one or zero data points, with a maximum of two points removed in any single simulation (Fig. 3). This demonstrates a robust method for managing statistical noise while preserving the integrity of the underlying data distribution.
Figure 3. Number of outliers removed during every bootstrapping selection using 3 standard deviations (n=10000).
Redundant measurements for validation
The Isometric protocol requires that solid-phase CDR estimates be validated with an additional, redundant, measurement phase, such as with liquid-phase measurements. The dissolved inorganic carbon (DIC) measured in soil water should align with the net CDR calculated from solid-phase measurements after accounting for all loss pathways (e.g., cation exchange, biomass uptake).
Consistent with findings from other studies8,10, our results showed that net CDR estimates derived from liquid-phase measurements were systematically lower than those from the solid phase, although the wide uncertainty range in the solid phase measurement meant that they overlapped within uncertainty.
It is crucial to note that the liquid-phase dataset was limited due to challenges in sample recovery from the high-permeability sandy soils. Potential reasons for this discrepancy fall into two categories: an underestimation by the liquid phase or an overestimation by the solid phase.
Challenges in Soil Water Sampling: Passive lysimeters are often inefficient in fast-draining soils, leading to low sample recovery. Furthermore, sampling often occurs days after rain events, potentially missing the “first flush” of water that carries the highest solute concentrations. Significant spatial variability in DIC concentrations was also observed between different lysimeter locations, further complicating direct upscaling.
Potential for Solid-Phase Overestimation: To proactively address the risk of overestimation from the solid-phase data, a highly conservative approach was adopted for the final credit calculation. This involved using the 16th percentile of the estimated gross dissolution and the 84th percentile of the CDR loss terms. This conservative bounding reduced the final net CDR figure by an additional 32%. The data confirmed that cation exchange was the dominant loss pathway (56% of CDR potential), with negligible losses to non-carbonic acids.
Ongoing research continues to investigate other potential unquantified sinks, such as the binding of cations to iron/aluminium oxides or organic matter12, which could lead to overestimation in current solid-phase models. Continued refinement of MRV methodologies, including the integration of novel sensor technologies, is essential to resolving these discrepancies and further reducing uncertainty in field-scale EW deployments.
At InPlanet, we are actively investigating these mechanisms further, whilst maintaining conservative estimates to mitigate the likelihood of overcrediting. We currently have two research projects underway specifically on these topics, jointly with the University of Antwerp (Prof. Sara Vicca and Dr Harun Niron), funded by Cascade Climate and with Everest Carbon, funded by the Milkywire Climate Transformation Fund.
Conclusion and future outlook
The successful delivery of credits from Project Beija-flor to Frontier provides a critical proof-of-concept for deploying enhanced weathering at an operational scale. However, as with all early-stage deployments, its primary value to the scientific and climate community lies in the transparent documentation of key methodological learnings. Learnings, such as limited ITE resolvability in certain soil types, confounding background cation inputs in control sites, and the persistent discrepancy between solid- and liquid-phase quantifications are key learnings for the EW sector. Navigating these complexities requires adaptive field protocols and a commitment to conservative accounting principles, which proved essential for achieving credible, third-party verification for a subsequent deployment.
Ultimately, these findings underscore that the trajectory toward large scale EW deployment is contingent not on ignoring challenges, but on confronting and addressing them directly. Focused research into novel MRV technologies and the refinement of our understanding of fundamental soil processes is imperative to reduce uncertainties and build a robust scientific foundation for enhanced weathering as a durable climate solution.
References
1. Clarkson, M. O. et al. A Review of Measurement for Quantification of Carbon Dioxide Removal by Enhanced Weathering in Soil. Front. Clim. 6, (2024).
2. MapBiomas Project – Collection 9 of the Annual Series of Land Use and Land Cover Maps of Brazil,. (2024).
3. Lefebvre, D. et al. Assessing the potential of soil carbonation and enhanced weathering through Life Cycle Assessment: A case study for Sao Paulo State, Brazil. J. Clean. Prod. 233, 468–481 (2019).
4. Reershemius, T. et al. Initial Validation of a Soil-Based Mass-Balance Approach for Empirical Monitoring of Enhanced Rock Weathering Rates. Environ. Sci. Technol. (2023) doi:10.1021/acs.est.3c03609.
5. Suhrhoff, T. J. et al. Updated framework and signal-to-noise analysis of soil mass balance approaches for quantifying enhanced weathering on managed lands. CDRXIV (2025).
6. Sutherland, K. et al. Enhanced Weathering in Agriculture v1.0 — Isometric. https://registry.isometric.com/protocol/enhanced-weathering-agriculture (2024).
7. Tipper, E. T. et al. Global silicate weathering flux overestimated because of sediment–water cation exchange. Proc. Natl. Acad. Sci. 118, e2016430118 (2021).
8. Kanzaki, Y. et al. Soil cation storage is a key control on the carbon removal dynamics of enhanced weathering. Environ. Res. Lett. 20, 074055 (2025).
9. Kelland, M. E. et al. Increased yield and CO2 sequestration potential with the C4 cereal Sorghum bicolor cultivated in basaltic rock dust-amended agricultural soil. Glob. Change Biol. 26, 3658–3676 (2020).
10. te Pas, E. E. E. M., Chang, E., Marklein, A. R., Comans, R. N. J. & Hagens, M. Accounting for retarded weathering products in comparing methods for quantifying carbon dioxide removal in a short-term enhanced weathering study. Front. Clim. 6, (2025).
11. Larkin, C. S. et al. Quantification of CO2 removal in a large-scale enhanced weathering field trial on an oil palm plantation in Sabah, Malaysia. Front. Clim. 4, (2022).
12. Niron, H., Vienne, A., Frings, P., Poetra, R. & Vicca, S. Exploring the synergy of enhanced weathering and Bacillus subtilis: A promising strategy for sustainable agriculture. Glob. Change Biol. 30, e17511 (2024).
