The field site is located on The Future Forest Company-managed Estate, Dumyat, near Stirling, Scotland (56.14930789 N; −3.89292692 W), and is referred to as Dumyat throughout this paper (Figure 1). The site is a grassland pasture, which is periodically used for grazing cattle and sheep from late summer to early autumn. The site is situated on a flat area of land at the bottom of a hillslope (field site gradient = 2°). The site is underlain with historic drainpipes running in the North–South direction. During the field trial, no fertilizer was applied to the field site.

The climate of the site is characterized as temperate oceanic, according to the Köppen-Geiger classification scheme (Beck et al., 2018). Directly before the experiment began, the 10-year average annual temperature was 10.1°C and the average daily precipitation-evapotranspiration (P-ET) was 1.6 mm. During the course of the experiment, the average daily P-ET was 2.8 mm and annual average temperature was 8.2°C. The seasonality in the observed P-ET over the duration of the experiment can be found in Figure 2.

Four experimental plots, each 48 × 100 m in size were set up in September 2022. Between the plots a buffer of 12 m was established, corresponding to the width of one pass with the agricultural lime spreader that was used for basalt deployment. Each plot was amended to a different basalt application rate (0, 23, 78 and 126 tons ha⁻¹, see Figure 1). Within each plot a central area was established, and three sets of soil porewater macro-rhizon samplers (pore size of the membrane = 0.12–0.18 μm) were installed to the North, East and West of this central point via soil pits at 5 and 10 cm depth. The soil pits in which the rhizons were placed were 3 m from a central point, with straight-line distances of 5.2 m between the pits. The rhizons were fenced off using wooden fencing to prevent equipment damage from livestock and wildlife.

The basalt application was surface-applied in two stages. The majority of the plot was spread using a commercial lime spreader on the 20th September 2022. The central fenced off area was spread using a hand sieve on the 27th September 2022 to ensure an even spreading distribution across the area where porewater extraction was conducted.

Baseline soil samples were collected on the 19th September 2022 (1 day before basalt spreading) using an Ejkelkamp ‘Edelman’ auger (⌀ = 4 cm) and a sample depth of 30 cm. Soil samples were taken in the larger plots where basalt was spread using a commercial spreader at the locations indicated in Figure 1. The soil samples were taken in a W-pattern across each plot and georeferenced using a Trimble dGPS. Larger soil clumps were removed with a cloth between sampling events if needed. Before each sample the auger flight was prepared by taking a disposable sample directly adjacent to the sample point, to avoid contamination between samples. Samples were collected using plastic bags. The sampling depth was measured with a ruler until 30 cm depth was reached. Once collected, all samples were dried at < 40°C, sieved to < 2 mm and stored in plastic containers until analysis.

Both baseline and monitoring soil samples were sent for analysis to the United Kingdom Accreditation Service (UKAS) tested and ISO accredited (ISO/IEC 17025:2017) lab NRM—part of Cawood Scientific Limited. Baseline soil samples were analyzed for soil texture, cation exchange capacity and major exchangeable cations, and soil inorganic and organic carbon. Soil pH was determined in UNDO’s in-house lab and bulk density measurements were determined at the UK Biochar Research Center (University of Edinburgh). The analysis methods are described in Supplementary material S4. Analysis of the baseline soil samples showed a very uniform soil texture across the entire field, classified as a silty clay loam according to the USDA soil taxonomy. Mean and standard deviation for soil organic carbon content, and exchangeable calcium, magnesium, potassium and sodium, as well as soil pH and bulk density in three different depths are reported in Table 1. Baseline soil samples were also analyzed for soil inorganic carbon, but all samples were below detection (<0.1 wt.%). Differences in baseline variables across the four plots were evaluated using an ANOVA. There were no differences between the four plots, except for exchangeable calcium and magnesium (p < 0.05). A Tukey’s HSD post-hoc test revealed that the significant differences were present between the control vs. 23 t ha⁻¹, control vs. 126 t ha⁻¹ plots for calcium and control vs. 23 t ha⁻¹ for magnesium. Model requirements of normality and homoscedasticity were tested using Shapiro–Wilk and White’s Lagrange multiplier test, respectively. All variables met model requirements except exchangeable magnesium which showed slight heteroscedasticity (p = 0.03), indicating that differences should be considered cautiously. The statistical comparison was carried out using Python (version 3.9) and the package Statsmodels (version 0.14.4; Seabold and Perktold, 2010).

For each of the treated fields, basalt from the Hillend quarry was applied. The quarry is situated in Airdrie, Ayrshire (55.88552 N; −3.88448 W) and is owned by Tillicoultry Quarries Ltd.

Eight kilograms of rock was taken from a quarry stockpile on the 10^th August 2022. The mineralogy of the sample (using quantitative X-ray diffraction, XRD), chemistry (using X-ray fluorescence, XRF), particle size distribution (using laser diffraction analysis), and specific surface area (SSA; using 5-point Brunauer–Emmett–Teller (BET) methodology N₂-adsorption analysis) were analyzed. Detailed methods for the mineralogical, chemical and SSA characterization referred to in this paper can be found in Skov et al., (2024) and Supplementary material S1 in this paper.

The samples contained 69.9 wt.% fast weathering minerals, when compared to mineral groupings from Lewis et al. (2021) (Table 2). The crushed basalt had a BET SSA of 0.917 m²g⁻¹, and the D₅₀ particle size was 1,006 μm. The particle size and whole-rock chemistry are described in the Supplementary Figure S1 and Supplementary Tables S1, S2, respectively.

The first porewater samples were collected from the rhizon samplers installed at the site on the 15^th October 2022 (23 days after the basalt was applied to the fenced-off areas). Porewater samples were collected c. every 2 weeks (between 10 and 18 days) when soil saturation levels were sufficient for porewater extraction. Porewater was extracted using a vacuum that was set and left for 48 h (for the first 170 days, until 7th March 2023) or 24 h (for the last 410 days, from 21st March 2023 to 18th April 2024) to allow water to filter into a 30 mL plastic syringe.

After collection, the porewater samples were poured into clean, 50 mL polypropylene conical flasks and c. 3 mL of the solution was pipetted into a 15 mL falcon centrifuge tube and stored in a refrigerator before submission for Inductively Coupled Plasma Mass Spectrometry (ICP-MS) and ion chromatography (IC) analysis at Nottingham University. The ICP-MS and IC analysis methods can be found in Supplementary material S2, S3, respectively. pH was measured using an OMNIS Advanced Titrator with Magnetic Stirrer.

Soil monitoring samples were collected from the same locations as the baseline soil samples, using dGPS to relocate the sampling points shown in Figure 1. The monitoring samples were analyzed for exchangeable cations, soil inorganic carbon, as well as soil pH, following the same procedure as for the baseline soil samples (Section 2.2; Supplementary material S4).

Baseline samples (Section 2.2) were taken the day before basalt application on the 19th September 2022. The following three monitoring samples were taken 4, 7 and 10 months after basalt application, whereas the last monitoring samples were taken in January 2025, one and half year after the previous monitoring sample and 28 months since the experiment began.

Additionally, on one singular sampling event, five composite samples each comprising 20 subsamples were taken in the control and 126 t ha⁻¹ plots. The samples were taken to a depth of 10 cm and subject to ammonium acetate extractions to calculate inferred pCDR from exchangeable cations only (Section 2.6.2) by Weathering Industries Ltd.

The cumulated bicarbonate flux between treatment plots relative to the control in both depths (5 and 10 cm) was evaluated statistically at the end of the porewater timeseries, after 576 days (c. 1.5 years). The overall effect of basalt amendment on the cumulated bicarbonate flux, estimated from porewater samples, was evaluated using an Ordinary Least Squares model (OLS) with treatment as a continuous explanatory variable. In order to evaluate the relative differences between the plots/treatments, the prediction interval around the linear models were inspected. The estimated inferred pCDR from exchangeable cations, ExPot and soil pH was evaluated using a GLM with a gaussian distribution family and treatment and sampling date as interacting, categorical variables. To account for inter-plot variability the mean ExPot and soil pH in the baseline samples, the baseline was subtracted at each monitoring sampling event. The model assumptions were evaluated using Shapiro–Wilk (for normality) and White’s Lagrange multiplier test (for homoscedasticity). Statistical analyses were carried out using Python (version 3.9) and the package Statsmodels (version 0.14.4, Seabold and Perktold, 2010).

The raw porewater concentration data for calcium, magnesium, sodium and potassium, as well as pH and electrical conductivity (EC) from both extraction depths are shown in Figure 3. Data for iron, barium, strontium, chloride, nitrate, sulfate and fluoride are plotted in Supplementary Figure S2. Porewater pH varies between 3.6 and 7.6 (Figures 3A,G), and EC between 2.3 to 304.9 μS m⁻¹ (Figures 3B,H), respectively. The cation concentrations were ranked in descending order according to their magnitude as follows: calcium > sodium > magnesium > potassium.

Largely, higher calcium, magnesium and sodium concentrations were identified in basalt amended treatments (e.g., 126 > 78 > 23 > 0 t ha⁻¹). This trend is also evident in the magnesium and sodium time series, relative to calcium (Figures 3C–E,I), indicating weathering. Calcium concentration is often more variable in soils (Dietzen and Rosing, 2023), which could explain the greater variance in the measured calcium concentrations in the pore water. EC data also broadly follows the same trend of values in the plots that received basalt amendment (Figures 3B,H), indicating a measurable contribution of ions introduced by basalt weathering to the ionic strength of pore water solutions. Potassium concentrations are consistently higher and more variable at any given time point in the 78 t ha⁻¹ treatment relative to all the other treatments (Figures 3F,L). Variation in the 78 t ha⁻¹ treatment was most likely the result of the natural and unexplained variability of one rhizon. Variability in K concentrations were not identified in baseline soil samples (Section 2.2).

The pH data are highly variable, with no discernible trends (Figures 3A,G), suggesting the involvement of pH buffering reactions involving cation exchange (Weil and Brady, 2016), organic matter (Stevenson, 1995), surface complexation (Dzombak and Morel, 1991) and microbial and root respiration (Paul, 2013).

The results of porewater concentration data converted into bicarbonate flux are presented in Figure 4. Data from the control plot indicate that natural bicarbonate fluxes vary between 0.438 ± 0.09 and 0.656 ± 0.321 tCO₂ ha⁻¹ in the 5 and 10 cm samples over the duration of the porewater time-series.

After 576 days of rock weathering, a strong overall dependency of basalt amendment on the cumulated bicarbonate flux was found at both 5 and 10 cm depth (p < 0.005 in the OLS). The two models showed that the treatment accounted for a very large part of the variability in the cumulated bicarbonate flux, with R-square values of 0.91 and 0.56, for the 5 and 10 cm depth, respectively. The statistical models met model assumptions.

The prediction intervals between the control and the two high applications (78 and 126 t ha⁻¹) did not overlap in the porewater time series from 5 cm depth (Table 3). Despite the significant overall effect of basalt treatment on the cumulated bicarbonate flux in 10 cm, the prediction intervals for the control group overlapped with the prediction intervals for the treatments, indicating no statistical difference due to the variability in the control data (Table 3). When comparing the control to the basalt amended treatments in treatments where prediction intervals did not overlap, the mean difference±std. varied between 0.33 ± 0.11 and 0.53 ± 0.13 tCO₂ ha⁻¹ (Table 3) as direct pCDR.

Major exchangeable cations (calcium, magnesium, sodium and potassium) and soil pH were determined on soil samples taken to a depth of 30 cm across the four plots (Figure 1) at five time points during this experiment. The sum of the major exchangeable cations is shown in Figure 5. The results indicate that there is high variability both within and between the plots. The variability appears to be the greatest in the spring and summer samples from 2023. Both the baseline and the samples from January 2025 were also analyzed for soil inorganic carbon, to track the precipitation of carbonates, but all results were below the detection limit (0.1 wt.% inorganic carbon). Observing no precipitated inorganic carbon indicates that under the conditions observed in this experiment, that direct pCDR occurred only through the generation of bicarbonate. This was expected, given the pH conditions were below pH 7 (Figures 3, 5).

Contrary to Figure 5, data from the single time point (July 2023) between control and 126 t ha⁻¹ when ExPot was calculated on composite samples from a 0–10 cm depth identified statistically significant increases in ExPot in the 126 t ha⁻¹ treatment relative to the control. Significantly higher concentrations of calcium and sodium in the high application plot (p < 0.05, ANOVA) and marginally higher magnesium, as well as marginally lower potassium (Supplementary Figure S3). Overall, this led to a statistically significant inferred pCDR of 1.909 tCO₂ ha⁻¹ (Supplementary Figure S3).

While the soil pH was significantly higher than the control in the highest application plot (126 t ha⁻¹) on the 23^rd January 2023 (after c. 4 months) and 18^th April 2023 (after c. 7 months), as well as for 23 and 78 t ha⁻¹ on the 11^th July 2023 (after c. 10 months, 126 t ha⁻¹ was only marginally significant, p = 0.065; Supplementary Table S2), no significantly higher ExPot is observed in any of the time points measured in this study between the control and any treatment (Supplementary Table S3). This suggests that the exchange pool represents a large pool of retained metal ions, by comparison with the metal concentrations in porewaters. In this scenario large relative changes in porewater concentrations would be accompanied by small relative changes in exchanger loadings.

A key limitation to the experimental design in this study is the absence of plot replication; there was only one plot per treatment. A range of experimental designs for EW studies have been published previously, including paired catchments (control and amended; e.g. Larkin et al., 2022; Taylor et al., 2021), strip designs (e.g., Holden et al., 2024), split field designs (e.g., McDermott et al., 2024) and randomized block designs (e.g., Deng et al., 2024; Stubbs et al., 2025). These experimental designs are typically more statistically robust than the experimental design used to generate data in this study and as such the differences identified in this study should be evaluated with caution.

The experimental design presented in this study was initially designed to investigate whether pH and EC sensors could be used to fill the temporal gaps between porewater sampling events. Hence, the rhizon samplers were installed in soil pits in close proximity with the soil sensors with the aim of identifying empirical relationships between soil sensor data and porewater sample results. A key advantage of sensor-based measurements is that they would generate continuous data, rather than at discrete time intervals, which could be used to extrapolate direct pCDR using precipitation surplus (P-ET) data, as presented here. This is discussed in detail in Section 4.2. Due to the cost of sensor installation and the requirement for a central data logger for each 10 sensor pairs of pH and EC sensors, it was not financially feasible to install sensors across replicated plots. Unfortunately, the acquisition of soil sensor data was terminated due to vandalism at the experimental site.

Suitable porewater baseline samples would require multi-year sampling to account for both seasonal and inter-year variability. However, obtaining initial porewater samples before the rock was applied would have been useful for ensuring the soil had re-equilibrated following rhizon installation and for confirming whether, for a short period before the experiment took place, that the plots were statistically similar to each other. There were significant differences in the baseline soil samples for exchangeable calcium and magnesium as these were the only two variables which were significantly different in the baseline samples. However, the mean±std. for exchangeable calcium and magnesium varied by only 7 and 10% between the baseline samples.

Direct pCDR from cumulative porewater was significantly explained by the application of basalt in both 5 and 10 cm depth (OLS r² = 0.912 and r² = 0.56 for 5 and 10 cm, respectively), at the end of the 1.5 year long porewater time series. Where the prediction intervals for treatments did not overlap with the control in the 5 cm sampling depth (78 and 126 t ha⁻¹), linearity is observed between the tons of rock applied and percentages of CDR (Table 3) with mass-normalized-pCDR of 0.0042 tCO₂ ha⁻¹ tRock⁻¹.

The results also indicate that direct pCDR observed in the porewater accounts for only a small amount of the maximum theoretical CDR calculated through the E_pot equation—between 1.1 and 1.7%. These low percentages may indicate low weathering rates. In this experiment, the feedstock was not amended into the soil profile, but was surface-applied on the grassland and therefore was not exposed to the higher pCO₂ observed in soil profiles (e.g., Nan et al., 2016). Moreover, the feedstock applied in this study was coarse grained with a D₅₀ of 1,006 μm (e.g., Supplementary Figure S1). The theoretical maximum CDR does not take rock dissolution kinetics into account. Rock dissolution rates are influenced by the amendment’s specific surface area (which in turn is influenced by particle size and mineralogy; Lewis et al., 2021) and mineralogy (e.g., structural location of the cations), while E_pot values only use whole rock chemistry. These results confirm that it is not appropriate to use E_pot values to predict CDR in agricultural EW studies over annual timescales. It is not possible to test at what time scale an E_pot value is appropriate for, however, the value could be useful for comparing feedstock to select appropriate rock sources (Clarkson et al., 2024).

Given that the study duration is 1.5 years, the statistically significant CDR observed, ranging between 0.33 and 0.53 tCO₂ ha⁻¹ over c.1.5 years of weathering, as well as the linearity observed, may evolve over longer timeframes. Initial weathering rates tend to be higher due to the dissolution of finer particles and higher SSA asperities (e.g., Power et al., 2025). However, the detection of weathering signals may experience a time lag, as cations can initially be absorbed onto exchange sites, with a greater fraction of the weathering products remaining in pore water after several years (Kanzaki et al., 2024). The results presented in this study, therefore, should be considered as site- and time-specific. Additionally, the application of 78 or 126 t ha⁻¹ of rock amendment, where significant CDR is observed, are not operationally applied to agricultural land.

There are some limitations to using a porewater time series alone for direct pCDR estimations, and the results presented in this paper most likely represent the minimum direct pCDR over the measurement period. While the determination of the soil porewater concentration of bicarbonate at a particular point in time is the most direct measure of the instantaneous pCDR reservoir in the topsoil (Clarkson et al., 2024), using these measurements for temporal extrapolation have limitations including: (1) seasonal variability, (2) limited sampling duration and frequency and (3) practical limitations. Seasonal variability in both the water flux through the soil (i.e., the transport of CDR products to the groundwater; Figure 2) and porewater concentrations (Figure 3) were observed in this study, with Δbicarbonate varying by c. 2 orders of magnitude between time points over the year (Figure 4). This implies that the timing of porewater sampling in relation to rain events is pivotal to avoid missing a signal. Time series analysis has shown strong correlation (r² = 0.57, p = 0.0014) between rainfall averaged over a 10-day period prior to sampling and porewater cation concentrations (Betz et al., 2025). As the porewater sampling takes place over c. 24 to 48 h approximately every 2 weeks (in periods of sufficient soil moisture to extract porewater), the sampling period represents <14% of the total time integrated.

Additionally, the first porewater samples were taken 25 days after the rock application, due to insufficient soil moisture for porewater collection. Between the time of rock application and the first measurement, the fields were subjected to a total water flux (P-ET) of 106 mm. Weathering rates may decrease over time, and the results could be skewed by missing this initial weathering signal (Power et al., 2025). Therefore direct pCDR may have been underestimated.

Even if the sampling timing is based on analysis of the relationship between rain events and porewater concentration (Betz et al., 2025), the logistical challenge in collecting soil porewater is primarily governed by the availability of sufficient water yield. A potential solution for the latter utilizes centrifugation of soil samples (e.g., Jones et al., 2025). Integrating different instrumentation, such as passive wick drainage flux meters (as implemented in Holden et al., 2024) or hydrologically isolated plots which both collect water continuously may avoid these issues. Soil porewater extraction is notoriously known to be labor-intense with no guarantee of porewater yield, as samplers may lose vacuum or be ineffective at low moisture levels, leading to insufficient volumes of sample obtained for chemical analysis. The macro rhizons used in this study to collect porewater are also considered consumable, with an expected lifespan from the manufacturers stated as 6 months. Dry summers and cold winters affected the longevity of the rhizons in this study, but replacement is carried out with minimal disturbance.

Despite these drawbacks in the experimental data, the method to calculate flux based on daily, rather than monthly or annual, precipitation surplus data (P-ET), is likely to provide a more accurate indication of direct pCDR due to seasonality (e.g., Figure 2). Furthermore, as stated above, correlation data between rainfall averaged over 10-day periods also suggests that monthly P-ET values may not be sufficient when integrating precipitation surplus and concentration data to obtain direct pCDR estimates. This method could be improved in future iterations by incorporating soil hydrology into flux calculations.

Overall, no statistically significant changes were observed in the ExPot (Supplementary Table S3), which is likely a result of the experimental set-up and sampling design. The concentration of exchangeable cations in the soil samples taken in the treated plots was not significantly higher than in the control in any of the monitoring sample events (Figures 5, 6; Supplementary Table S3). This is despite soil pH readings appearing different for the highest application rate on several occasions (Figure 5; Supplementary Table S2). Several potential reasons for the lack of differences in the inferred pCDR stored on the soil exchangers have been identified, including: (1) inherent soil variability, (2) spreading heterogeneity, (3) sampling frequency and (4) basalt application strategy. The variability in exchangeable cations observed in this study suggests that compositing of multiple samples into one may be useful in decreasing within treatment sample variability. Further, the size of the baseline ExPot reservoir is large compared to direct pCDR measured from pore water (c. 1 order of magnitude), and so natural variation within soil could readily exceed the signal from any accumulated cations due to the EW process. The soil samples were taken within the greater plot boundaries where basalt was applied using a commercial lime spreader. Due to the inherent variability in spreading rate when using commercial equipment, it is likely that differences between the plots may have been discernible had the sampling density been higher. Additionally, the accumulation and flushing of inferred pCDR on exchangeable cations is likely to be time sensitive (Dietzen and Rosing, 2023), so infrequent sampling may have missed the accumulation peak before flushing.

The 30 cm sampling depth may have diluted any top soil differences, as evidenced from the singular time point tested in July 2023, where five composite samples each comprising 20 subsamples were taken to a depth of 10 cm from the control and 126 t ha⁻¹ plots, respectively. These values identified an ExPot inferred pCDR of 1.909 tCO₂ ha⁻¹ (Supplementary Figure S3), without accounting for the deduction factors which are briefly discussed in Section 4.4.

The direct pCDR values presented so far assume that all the pCDR calculated in the porewater will remain in solution. However, outgassing from riverine (i.e., equilibration of soil water into riverine water; Zhang et al., 2022, 2025; Harrington et al., 2023) and oceanic outgassing (due to carbonate buffering; Renforth and Henderson, 2017) can reduce CDR efficacy by 22% (Table 4). Riverine losses and ocean losses are likely to be highly dependent on catchment/environmental conditions (Zhang et al., 2022; Harrington et al., 2023), however, a factor of 5% is applied in this study based on results from Zhang et al. (2025). Oceanic carbonate buffering will reduce the CDR efficacy by 13–17% (Renforth and Henderson, 2017), depending on ocean temperature, salinity and atmospheric CO₂ concentrations.

Strong acid weathering is already excluded when calculating alkalinity by charge balance via Equation 3, but to calculate the net CDR from porewater data, additional deductions need to be taken for plant uptake. Plant uptake could have occurred in this trial because grass species typically root down to 10 cm depth (Schenk and Jackson, 2002), and therefore the cations measured during this study (particularly K), could be taken up by plants. However, the site was intermittently grazed and no grass cuttings were taken during the duration of the trial to confirm this. Moreover, grazing did not occur directly above the porewater rhizon samplers, as the central parts of the plots were fenced off. Hence any uptake of cations in the grass would be returned to the soil at senescence and therefore grass yields and chemistry were not determined. In addition to appropriate deductions for losses and leakages, a full life cycle analysis of all emissions associated with activities during the project lifetime would also have to be conducted to calculate net CDR (e.g., Puro, 2024; Sutherland et al., 2024).

The estimated direct pCDR, based on porewater data in this study (0.257–0.413 tCO₂ ha⁻¹ after deductions, Table 4), sits within the range of published data (0.025–0.55 tCO₂ ha⁻¹; Table 5) from previous EW studies that used aqueous sampling to deduce CDR (Kukla et al., 2024). However, there are challenges when comparing different EW studies due to the variety of climatic, soil and rock conditions used in field trials data. Given that a linear relationship between application rate and direct pCDR was observed in this study, inter-study CDR comparisons are discussed after being normalized per ton of rock applied and per year (tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹ and referred to as mass-time-normalized). Caution should be taken when interpreting mass-time-normalized CDR (or pCDR) or extrapolating further than simple comparisons; linear relationships are not always observed in enhanced weathering studies (Rijnders et al., 2024) and only three data points are used to interpret this trend, and only two of these were significant. Additionally, within the published studies cited in Table 5, several were subject to repeat basalt applications. Where measurements are taken at depths below 1 m depth (e.g., well below the NFZ), CDR is referred to as CDR while above 30 cm depth follows the same nomenclature as the results provided in this study and is referred to as direct pCDR.

Both McDermott et al. (2024) and Deng et al. (2024) use P-ET to calculate direct pCDR/CDR fluxes, while Holden et al. (2024) use passive wick drainage flux lysimeters. High mass-time-normalized direct pCDR (up to 0.088 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹) of c.1–4 orders of magnitude higher than other studies listed in Table 5 were observed in McDermott et al. (2024). The fast dissolution kinetics of calcite, c. five orders of magnitude faster than the labradorite which comprise 43.8 wt.% of the material used in this study, likely explains why the mass-time-normalized direct pCDR was higher. Deng et al. (2024) observed a similar amount of mass-time-normalized CDR (0.0023 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹) to the direct pCDR study (0.0021 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹, including deductions). There are similarities in the soil used in this study compared with Deng et al. (2024) (pH = 6.6; soil texture = silty loam/silty clay loam) and rock amendment used (basalt, with a p80 of 1,050 μm). However, leachate samples were collected from 1 m depth—up to 20 times deeper than the porewater extraction depth presented in this study. In Holden et al. (2024), non-carbonic acid weathering was reported to drive over 98% of weathering and therefore, coupled with deep flux meter installation, low and statistically insignificant (p = 0.082) bicarbonate fluxes were observed.

Both Taylor et al. (2021) and Larkin et al. (2022) utilized river discharge to determine CDR. Firstly, 3 years of annual applications of 50 t ha⁻¹ of basaltic andesite, CDR determined from alkalinity in river discharge was unresolvable in two out of three paired catchments when comparing control to amended paired catchments (Larkin et al., 2022). However, the paired catchment which showed c.1 tCO₂ ha⁻¹, converts to mass-time-normalized CDR of 0.022 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹ which is of a similar order of magnitude as observed in Deng et al. (2024). After applying a lower rate application (3.44 t ha⁻¹) of wollastonite amendment to a catchment which was monitored for 15 years, modest mass-time-normalized CDR of between 0.0005–0.0025 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹ was observed.

River transport is quicker than groundwater transport (Dingman, 2002) so once the charge-balanced bicarbonate is transported into rivers, and after the initial outgassing and assuming that no more outgassing will occur along a river course, weathering products could reach the ocean quickly. However, it can take time for weathering products to be transported to rivers, and weathering products may remain in groundwater for millennia. Many of the CDR/pCDR data used in this study are taken at depths greater than 1 m which is considerably deeper than the near field zone as currently defined in methodologies (e.g., Sutherland et al., 2024; Puro, 2024; Mills et al., 2024). At deeper depths, outgassing, and the potential for plant uptake may not be such an issue. Even so, more research is needed to identify the time point under different hydrological conditions where dissolved and charge balanced bicarbonate is at low risk of outgassing, or outgassing during throughflow until the water table is constrained.

Aqueous phase measurements from studies compared in this section are typically one order of magnitude lower than ‘natural’ pCDR measurements derived from soil- and mass- balance-based approaches. Mass-time-normalized soil-based calculations report values of inferred pCDR of 0.013 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹ when four annual applications of 50 t ha⁻¹ basalt (e.g., 200 t ha⁻¹ at the time of measurement) were applied to maize-soy rotations in the Midwest United States (Beerling et al., 2024). Tropical ERW measurements reported by InPlanet during their carbon credit verification by Isometric obtained mass-time-normalized inferred pCDR values of 0.055 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹ when applying 10.6 t ha⁻¹ of basalt and measuring after 1 year (Isometric HQ Ltd, 2025).

The time-mass-normalized direct pCDR reported in this study (0.0021 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹, including deductions) is within the midrange of time-mass-normalized results reported in Table 5 (0.001–0.088 tCO₂ ha⁻¹ tRock⁻¹ yr.⁻¹). The study took place in temperate climates, using relatively slow weathering feedstocks (e.g., basalt). Other feedstocks, such as wollastonite, and regions with warm, wet summers, would likely lead to faster weathering rates, as per common consensus on dissolution kinetic laws (e.g., Palandri and Kharaka, 2004). However, slower weathering feedstocks could be significant over decadal (e.g., 20-year) timescales and this study suggests that while detecting signals in slow weathering materials can be challenging, particularly at lower application densities, it is achievable via high-frequency and/or time integrated measurements.