A mesocosm experiment with 30 mesocosms was constructed at the experimental site at the Drie Eiken Campus of the University of Antwerp (Belgium). This experiment was part of a larger mesocosm experiment that aimed to investigate heavy metal fate and plant biomass in silicate amended maize plants (Rijnders et al., 2025). The mesocosms (0.6 m height, radius = 0.25 m) received natural rainfall and received additional water through manual irrigation (Fig. S2). In May 2021, the lower 40 cm of each mesocosm was filled with a slightly acidic sandy loam soil (Table 1).

The upper 20 cm was filled with the same soil, either unamended in the control treatment (five mesocosms), or amended with basalt (Fig. 2). Five mesocosms received 50 t basalt per ha, while six others received different amounts of basalt, ranging between 10 and 200 t ha⁻¹ (Table 2). The basalt was mixed homogenously in the control soil using a concrete mixer. Basalt was obtained from DURABAS (https://www.rpbl.de, last access: 25 March 2026). Particle size distribution (PSD) was analyzed using a mastersizer 2000 with a Hydro 2000G sample dispersion unit after removing larger particles with a 2 mm sieve. The P80 was 310.78 µm (see Fig. S5). The specific surface area (SSA) was determined with a Quantachrome Autosorb iQ using the Braunauer–Emmet–Teller (BET) method. The measurement used nitrogen gas as adsorbate with multi-point (5 points) and isotherm (77 K) settings. Samples with the same treatment were pooled in equal quantities into one sample to reduce the cost and time for analysis. All samples were degassed at 300 °C with 200 min of soak time. High measurement quality was ensured by frequent reference measurements (Bundesanstalt für Materialforschung und -prüfung, Germany) in addition to three technical repetitions for each measurement. The BET-SSA of the basalt rock was 9.226±0.08 m² g⁻¹. X-ray diffraction (XRD) and x-ray fluorescence (XRF) analyses are provided in the Supplement.

Basalt was mixed into the top soil on 17 May 2021. To allow leachate collection, mesocosms had a 2 cm diameter hole at the bottom. On the inside, the bottom of the pot was covered with a root exclusion mesh to prevent soil export. Glass collectors (2.3 L volume) were placed under the mesocosm to collect the leachates. Leachate volumes were determined throughout the experiment and were collected for chemical analyses on seven occasions. On 3 June 2021, two sweet maize seedlings (variety Tom Thumb) were planted in each the mesocosms and all pots received fertilization with nitrogen (N), phosphorus (P) and potassium (K) (96–10–79) kg ha⁻¹ by adding Ca nitrate (Ca(NO₃)₂), triple super phosphate (TSP, 45 % P₂O₅) and potassium sulfate (K₂SO₄). The experimental duration was 101 d; plants were harvested on 26 August 2021.

Soil water content and temperature were recorded using Cambell Scientific sensors (CS616) that are 30 cm in length. Watering (using rain water collected from a tank) was executed manually and total water manual inputs were tracked. In addition, daily precipitation amounts (in mm) were obtained for Wilrijk (Belgium) using the open source tool (https://www.visualcrossing.com/, last access: 25 March 2026). In the Supplement, an overview of environmental conditions (rainfall, total water inputs, temperature and soil moisture) is given.

Weekly pore water sampling was performed with rhizons (Rhizon Flex, Rhizosphere Research Products B.V., Wageningen, the Netherlands) installed at 5 cm depth in each mesocosm. Leachate and porewater samples were filtered through a 0.45 µm PET filter. The major cations (Ca, Mg and K) were measured through ICP-OES (iCAP 6300 duo, Thermo Scientific). Na was not directly measured in soil water samples; however, its charge contribution is accounted for indirectly in the measured soil water alkalinity. Before analysis, ICP samples were conserved using 1.5 mL (HNO₃ 69 %) per 30 mL sample. TA was determined using a SAN++ continuous flow analyzer (Skalar – NLD). The pH was measured using a HI3220 pH/ORP meter. Dissolved organic carbon (DOC) and DIC were measured using a FormacsHT with LAS sampler (Skalar – NLD). DIC and DOC were measured on eight and 12 occasions in leachates and pore water respectively.

Two quality control (QC) standards were analyzed for individual elements (Ca, K, Mg, sodium (Na), silicon (Si) and Fe). The mean precision of the QC standards was 0.84 %, 1.12 %, 0.54 %, 2.79 %, 1.67 % and 1.30 % for the respective elements. The mean accuracy for the two QC standards was 1.87 %, 2.30 %, 0.17 %, 1.88 %, 1.39 % and 2.65 % for Ca, K, Mg, Na, Si and Fe respectively. For TA soil water samples, mean accuracy and precision for two different QC standards were 1.51 % and 1.72 % respectively. The DIC measurements with FormacsHT had an accuracy and precision 1.09 % and 0.23 % respectively. Accuracy and precision were determined based on 12 measurements of a QC for TA (standards: 150 and 350 mg CaCO₃ L⁻¹) and DIC (range between 10 and 100 mg L⁻¹) and based on eight measurements of two different QC concentrations for each individual element.

Top soil pH was measured on five dates. To determine top soil pH, 4 g of air dried topsoil sample was dissolved in 10 mL deionized water and shaken before pH measurement using a HI3220 pH/ORP meter (Hanna Instruments, Temse, Belgium). After harvesting, soils were sampled using cylindrical soil cores (100 cm³, 5 cm length × 20 cm²). Samples were taken across the depth of the mesocosm and three sampling depths were considered (0–20, 20–30, 30–50 cm). One sample was taken for each depth and mesocosm. The cores were dried at 70 °C for 48 h to determine water content (g H₂O/g soil) and bulk density. An additional soil sample was taken at each depth, dried at 70 °C for 48 h and used for chemical analyses after sieving over a 2 mm sieve, removing the majority of the root biomass. Estimated base cations in sieved-off roots were negligible relative to soil pools and thus excluded from our main cation budget.

As conceptualized by Tessier et al. (1979), base cations can reside in four different soil pools: the exchangeable pool (cations weakly bound to SOM or clays), the carbonate pool (cations bound in pedogenic carbonates), the reducible pool (cations bound to Al/Mn/Fe hydr(oxide)) and the oxidizable pool (cations strongly bound to SOM). SOM bound to cations, extracted with weak salt solutions in the exchangeable pool typically has a low turnover time (Poeplau et al., 2018) and is therefore thought to be more susceptible to microbial decomposition than oxidizable SOM.

We adapted the original Tessier scheme by replacing 1 M magnesium chloride with 1 M ammonium acetate (NH₄(CH₃COO)) for extraction of the exchangeable pool, in order to be able to measure all base cations in the exchangeable pool (Table 3). Likewise, Na-acetate was replaced with a mixture of acetic acid and water to be able to measure Na in the carbonate pool. We quantified SIC changes from the base cations in these acetic acid extracts (as in Larkin et al., 2022) (see also Eq. S4). Additionally, three other SIC measurement techniques were explored to compare the sensitivity of detecting SIC changes after amending with a range of basalt (see Sect. S3.7).

Prior to extractions, approximately 1 g of soil was air dried. We also conducted the extractions for the pure basalt that was initially added to the mesocosms to be able to correct for the cations that were initially already present as exchangeable, carbonate, reducible and oxidizable pool cations. After each extraction, samples were centrifuged for 2 min at 2000 rpm, supernatants were collected for analysis. The remaining soil pellet after centrifugation was washed with 10 mL of demineralized water before the following step. Relevant elements (K, Na, Mg, Ca, Al, Fe and Si) were measured in each pool using ICP-OES (iCAP 6300 duo, Thermo Scientific) for each pool. Si was only assessed in the reducible pool and in the oxidizable pool to investigate whether Si forms amorphous oxides or allophane-like compounds or binds with organic matter. Al carbonates were not quantified here as naturally these carbonates are not commonly formed (Takaya et al., 2019).

On 26 August 2021 (101 d after basalt amendment in soils), the aboveground biomass was harvested and dried for 48 h at 70 °C to determine dry weight. For the results on root biomass we refer to Rijnders et al. (2025). Plant material was ground with an ultra-centrifugal mill (Model ZM 200, Retsch GmbH, Haan, Germany). Base cations (Ca, Mg and K) were measured through ICP-OES (iCAP 6300 duo, Thermo Scientific) in aboveground biomass to calculate plant base cation stocks. Na was not measured in aboveground biomass, because the amount of weathered base cations accumulated in plants is relatively small compared to that in soils, we do not expect this omission to substantially affect our results. Base cations were measured separately in all aboveground biomass parts: stems, leaves, flowers and maize ears.

We use the delta (Δ) symbol to denote the difference relative to unamended control soil. Accordingly, we quantify ΔTA (the change in total alkalinity in the basalt-amended soil relative to the control) based on the difference in base cation concentrations between amended and unamended soils. As basalt only contains cations and no conservative anions, we assume that ΔTA can be quantified from the change in base cation charges (Eq. 1). 1ΔTA≈2⋅ΔCa+ΔMg+ΔNa+ΔK-ΔconservativeanionswithΔconservativeanions=0 We thus use a base cation accounting approach as also proposed by Bijma et al. (2026) and explicitly assume no changes in conservative anions after basalt amendment. Still, nitrogen cycling processes (nitrification and denitrification) may be sensitive to soil chemistry and may therefore differ between treatments. Further research is needed to verify whether conservative anions such as nitrates and phosphates are affected by rock amendment.

The Wr corresponds to the rate of rock dissolution. The Wr can be expressed per element or as moles of alkalinity equivalents (i.e. the sum of base cation charges (Eq. 1) per amount of rock surface area per unit of time (in mol m⁻² rock per s). In addition, we calculate a “potential inorganic CO₂ removal”. We use the same definition for potential inorganic CO₂ removal as in Steinwidder et al. (2025). A “potential inorganic CO₂ removal” can be defined as the maximum amount of inorganic CO₂ that could be removed if all experimentally determined, weathered, soil-retained base cations were to leach from the soil and be completely balanced by carbonate anions. Potential inorganic CO₂ removal was previously named “CDR potential” by Niron et al., 2024). The concept of CDR potential was first introduced by Renforth (2019) to describe the maximum inorganic CO₂ removal achievable if all base cations within a rock were to completely weather. More recently, Beerling et al. (2024) quantified base cation losses from topsoils using an immobile/mobile tracer approach (see Sect. 4.2), from which they also derived a measure of CDR potential. To maintain conceptual clarity, we avoid using the term CDR potential for purposes other than its original definition by Renforth (2019). When the term is employed, its meaning should always be explicitly stated.

To calculate the Wr (from all base cation increases relative to controls in plants, extracted soil fractions and soil water leachates), we sum changes in TA in the following pools: exported soil water (leachates), plants and soil pools. Weathering rates are uncertain due to variability in soil cation pools considered, potential underestimation from unextractable base cations in secondary minerals and methodological limitations in detecting all weathering products. We can express changes in the cation pool of each reservoir as the equivalent Wr required to supply the cations (Wr_leachate, Wr_plant and Wr_soil) (Eq. 2). Conventionally Wrs are expressed using a logarithmic scale as absolute values can vary strongly. 2log⁡WrΔmolTAm2rockpers=log⁡ΔmolTAsoil+ΔmolTAplant+ΔmolTAleachatem2rockpers We used a gradient of rock applications, where we calculated the slope of the molar change in base cation charges (expressed as an equivalent “alkalinity” change if these base cations were dissolved in water) with higher rock amendment (TA slope) (Fig. 3). We compared logarithmic and linear regression and selected the linear regression approach, as both approaches had comparable R2 and Akaike Information Criterion (AIC) values and linear slopes ease further data processing (Fig. S23 and Table S5). We opted for the linear regression approach to simplify subsequent calculations. To make our gradient approach more robust, we also calculated the log⁡ Wr for individual application rates in Fig. S13.

Then we converted units of the alkalinity slope for increasing rock application (TA slope) per unit of rock mass to moles of alkalinity per rock surface area and per time (Eq. 3). Equation (3) was used to calculate Δmol TA m⁻² per rock per s of base cations in leachates, plants and of every measured soil pool at every soil depth. 3ΔmolTAm2rockpers=Scavengedalkalinity(=TAslope)ΔmolTAgrockExperimentalduration[s]⋅SSAsilicatem2surfaceareagrock Δmol TA m⁻² per rock per s was thus quantified per pool, based on the change in base cations in the basalt treatment compared to the control treatment. For plants, we calculate TA slope through regression of harvested base cations with basalt application. Harvested base cations were calculated as the product of harvested aboveground biomass and their base cation content. Charge contributions of Na were not included; Na was not quantified at the time of plant biomass elemental analysis, which may lead to an underestimation of the alkalinity equivalent increase in the plant pool. However, given that base cation charges in the plant pool were about two orders of magnitude smaller than in the soil pool, we expect the effect of this omission to be limited. In addition, maize plants aim to actively increase their K/Na ratio which avoids salt stress, the K content of maize shoots is typically about 2 orders of magnitude larger than Na (Gao et al., 2016; Suarez and Grieve, 1988). For leachates, TA slope was calculated as the product of mean cumulative leachate volume and mean leachate TA concentration for each application rate and regressing them with the applied basalt as dependent variable.

Finally, the Wr attributable to the change in cation content of the soil pools (Wr_soil.) was calculated by summing the Wr_{soil_layer_k_pool_j} for each pool and depth (Eq. 4). Here, we sum changes in all pools of the topsoil (0–20 cm) and lower depths (20–30 and 30–50 cm) to obtain an aggregate value for Wr_soil. with Wr_{soil,layer k,fraction j} calculated as in Eq. (4) (with k = the number of depths and j = the number of considered soil pools). TA slope at every depth and soil pool was calculated as in Eq. (5). 4Wrsoil=∑k=13∑j=14Wrsoil,layerk,poolj5scavengedTA(TAslope)Δmolgrock=µmolTAgdrySoil(AmendedSoil)-µmolTAgdrySoil(controlSoil)Applicationrate(Amendedsoil)grockm-2groundarea⋅1000⋅BulkDensitykgdrySoilm3soil⋅thicknesssoillayer[m] Equivalents of soil retained base cation charge equivalents per gram of dry soil mixture can be calculated for each mesocosm by summing the charges from each base cation (Eq. 6). 6µmolTAgdrySoil=∑j=14µgCapooljgdrysoil40.078gCamolCa+µgMgpooljgdrysoil24.305gMgmolMg⋅2molTAmolcat+++µgNapooljgdrysoil22.990gNamolNa+µgKpooljgdrysoil39.098gKmolK⋅1molTAmolcat+ These individual base cations (e.g. Ca in pool j) are calculated from the difference of cations weathered during the weathering operation minus the cations initially present in that fraction in the applied feedstock (Power et al., 2025) (Eq. 7). For example, some cations can already exchange on the surface or edges of the applied minerals, so that these cannot be counted as weathered, they will however contribute to CDR once leached.

To calculate in-situ Wr, it is thus necessary to correct for the cations that had already been weathered from primary minerals at the time of silicate amendment. This correction is typically currently not being done in EW literature yet. As basalt is only added to the top soil and not deeper, this correction is only done for the 0–20 cm soil layer here. A limitation of the approach in Eq. (7) is that it assumes no physical transport of basalt to deeper soil layers, which may lead to underestimating weathered base cations in the 0–20 cm layer and overestimating them in the 20–30 cm layer. 7µgelementipooljgdrysoil=µgelementipooljgdrysoilPostweathering,soilmixture-µgelementipooljgdrysoiladdedwithfeedstockinitially The mass of a specific element (i) in each of the four (j) soil pools (in µg per element per g per soil) is calculated using Eq. (8). 8µgelementipooljgdrysoilPostweathering,soilmixture=concentrationelementiinpooljmgL⋅Volumeextractj[mL]massofsolidextracted[g] The initial addition of element i to pool j is calculated as in Eq. (9). 9µgelementipooljgdrysoiladdedwithfeedstockinitially=µgelementipooljgsilicate⋅Applicationrategsilicatem2Bulkdensitygdrysoilm3⋅depthofsoilamendment[m] According to the charge balance (Reaction R1–R3) during mineral dissolution, 1 mol HCO3- mol⁻¹ TA is generated (and thus 1 mol of CO₂ is sequestered). We define a factor η, that is equal to the ratio of HCO3- per mol of generated TA. According to charge balance, η=1. A more conservative approach is to assume that all this generated alkalinity will be exported to the ocean, after which chemical equilibrium degasses a portion of the alkalinity (η=0.7 mol CO₂ mol⁻¹ TA, assumed for oceans) (Renforth, 2012, 2019; Renforth and Henderson, 2017). According to Renforth (2019), the ocean alkalinization efficiency η ranged between 0.7 and 0.85. This η parameter is relatively uncertain given that model studies indicate that η can range between 0.65 and 0.8 mol CO₂ mol TA⁻¹ (see Sect. S6 in the Supplement of Kowalczyk et al., 2024). Alternatively, we can assume that all base cations will form solid carbonates in soils or rivers. In this case η=0.5 mol CO₂ mol⁻¹ TA (Reaction R4). In Table 4, we calculated potential inorganic CO₂ removals assuming conservative values of η=0.5 (carbonate precipitation scenario) and η=1(the highest possible η without any downstream DIC losses).

While calculating Wr, cations added with the rock feedstock that were already weathered were subtracted as in Eq. (7), yet these cations are not subtracted to calculate the potential inorganic CO₂ removal as base cations in weathered fractions of the rock feedstock can also leach to soil water whereby HCO3- is generated. Last, base cation changes in the plant pool were excluded from the potential inorganic CO₂ removal pool here, as a conservative approach we assume that base cations in plants will not reach the ocean. The latter assumption had a negligible impact on the potential inorganic CO₂ removal estimate (Table 4). 10potentialinorganicCO2removalkgCO2trock=ScavengedTAmolTAgrock⋅ηmolCO2molTA⋅44gCO2molCO2⋅1000withµgelementipooljgdrysoiladdedwithfeedstockinitially=0

To assess whether carbonate precipitation was theoretically possible during this experiment, we calculated SIc for dolomite and calcite. For Mg and Ca the SIc as the logarithm of the ion activity product and the solubility product constant of dolomite and calcite (SIc = log⁡IAP/K). Minerals have the potential to precipitate when a log⁡SIc > 0 is reached although substantial oversaturation of calcite (log⁡SIc > 1) without calcite formation is possible in rivers due to ion inhibition, e.g. by phosphate (Morse et al., 2007; Zhang et al., 2022). Likewise, they are in equilibrium at a log SIc = 0 and dissolve if log⁡SIc < 0. The R phreeqc package was used and the phreeqc.dat database was used. As an input, the experimental pore water (10 cm depth) composition of Mg and Ca was entered, as well as measured pH and TA. Daily SIc values were calculated by feeding unique combinations Mg, Ca and pH into the PHREEQC solution function for each day.

For SIC and elemental stocks in plant biomass, soil pools and soil water export, a linear regression with basalt application rate as a dependent variable was performed to test for a basalt effect. For measurements that were repeated in time (pore water and leachate DIC and DOC compositions), a linear mixed model was used with basalt and time as fixed factors and mesocosm as a random factor using the lme4 R package (version 1.1-33).

For measurements repeated in time, we assessed basalt × time interaction effects and discarded these if not significant. All analyses were executed in R version 4.3.2. As an additional sensitivity analysis for the determination of Wr using the slope of application rates approach described in the main text, we quantified Wr also for individual application rates in Fig. S13.

To propagate uncertainty between basalt and controls in Fig. 5, averages and standard errors for every replicated application rate (0 or 50 t ha⁻¹) were determined. The average from the 50 t ha⁻¹ was subtracted with the average from the control soil and se=(se_control2+se_basalt2). For non-replicated application rates (10, 30, 75, 100, 150 and 200 t ha⁻¹, se = 0) the measurement was subtracted from the control soil average and errors were also propagated with se=(se_control2+se_basalt2).

EW is typically considered as a durable CDR pathway that removes CO₂ from the atmosphere by producing DIC that is either transported to the ocean (Strefler et al., 2018) or precipitates as carbonates in the soil (Manning et al., 2013). Here, we observe a clear weathering signal (a TA and DIC increase) in top soil pore water (Fig. 4). These TA and DIC increases in the pore water of amended top soil are consistent with recent findings (Holzer et al., 2023; McDermott et al., 2024; Vienne et al., 2024). Increased DIC in basalt soils relative to controls may result from enhanced plant DIC exudation or from mineral weathering; our dataset does not allow these effects to be separated. DIC did however not leach from our soil columns within this experimental timeframe of 101 d.

Absence of substantial DIC leaching is in line with other short-term recent studies (Amann et al., 2020; Larkin et al., 2022; Niron et al., 2024; Vienne et al., 2024). For example, the log Wr of approximately -13 mol TA m² s⁻¹ quantified from DIC export after 1 year in a mesocosm trial with 220 t ha⁻¹ olivine-rich rock (Amann et al., 2020) was about three orders of magnitude lower than what would be expected from lab-scale weathering studies (roughly -10 mol TA m⁻² s⁻¹, Palandri and Kharaka, 2004). Vienne et al. (2024) amended soils with 100 t basalt per ha and quantified a CDR from exported TA that was in the same order of magnitude as in the work of Amann et al. (2020). Although the studies of Amann et al. (2020), Vienne et al. (2024) were relatively short (≤1 year) and used a relatively low water infiltration flux, also a longer (3 year duration) catchment-scale study in Malaysian oil palm plantations with high annual rainfall (>2000 mm yr⁻¹) detected no significant increase in TA leaching in the catchments (Larkin et al., 2022). There may thus be a substantial delay for DIC leaching.

A DIC leaching delay can have multiple causes (Fig. 1); a first possibility is pedogenic carbonate formation.

We observe that solid carbonates did not increase in our experiment, in contrast, SIC decreased in time. PHREEQC calculations for our experiment suggest that dolomite and calcite were undersaturated, so that carbonate dissolution was possible (Fig. S17). Saturation states are expected to be low in our experiment because control soil was undersaturated and dissolved base cations accumulated in other soil pools than the carbonate pool (Figs. 5 and 6). A decrease in SIC is in contrast with substantial SIC increases found after wollastonite rock amendment (Haque et al., 2019, 2020). The observed SIC increase in the latter field study may be partly attributed to residual carbonates from prior liming activities instead of new carbonate formation related to silicate weathering (Haque et al., 2020). Thus, not all measured SIC may reflect new carbonate formation in the study of Haque et al. (2020). For short-term basalt studies, using elemental C analysis, also no significant changes in SIC could be detected previously (Kelland et al., 2020; Vienne et al., 2022, 2024). In contrast, in the study of Larkin et al. (2022), a relatively small SIC increase was detected in amended soils, using carbonate pool extractions.

While TA was not exported or taken up by soil carbonates here and plant base cation losses were minor (Table 4) it was retained in top soil where the exchangeable and reducible pools reduced solute TA. We expect cations to be primarily associated with Fe- and Mn-(oxyhydr)oxides in the reducible pool and with organic matter in the oxidizable pool, as supported by literature (Tessier et al., 1979); However, the extraction chemicals of this sequential extraction scheme (hydroxylamine and H₂O₂) are known to have a limited specificity and may have also partially targeted other mineral phases (such as clays) (Ryan et al., 2008), which could explain the elevated Si observed in the topsoil pools (Fig. S15). In addition, the observed increase of aluminum in association with the reducible soil fraction indicate the formation of secondary minerals. While we cannot pinpoint the exact Mg-phases formed in our soils, our results do demonstrate substantial base cation retention in the soil and show that there can be more base cation losses to soils than to the exchangeable pool alone.

Our estimate of log⁡Wr, derived solely from significant increases in TA uptake at higher basalt amendment rates, was approximately -12 mol TA m⁻² s⁻¹. This estimate reflects changes in the exchangeable and carbonate soil pools, plant uptake, and leachate composition. Notably, this value aligns with previous studies that estimated log⁡Wr values between -12 and -11 based on base cation depletion from the exchangeable pool alone (Kelland et al., 2020; Reershemius et al., 2023; te Pas et al., 2023), as summarized in Vienne et al. (2024). Also in a batch leaching experiment with 1 mM CaCl₂ (designed to mimic soil solutions) the quantified log⁡Wr of basalt was found to be -11 (Van Der Bauwhede et al., 2024).

Estimates from Buckingham et al. (2022), based only on leachates, gave a much lower log⁡Wr of -15, partly due to low water infiltration rates. With a high infiltration flux (8000 mm yr⁻¹), Amann et al. (2022) estimated log⁡Wr between -12.5 and -13.5 from basalt leachates. This highlights the importance of including scavenged alkalinity to determine Wr in soils. When we also include non-significant regression slopes we derive a mean log⁡Wr estimate with substantial uncertainty (-11.11±2.70) mol TA m⁻² s⁻¹. From individual application rates, we quantify log⁡Wr ranging between -11 and -10 (Fig. S13); these values are comparable to those observed in soil-free, laboratory-scale basalt dissolution experiments conducted at circumneutral pH (Brantley et al., 2008; Gislason and Oelkers, 2003). They also approximate the dissolution rates of key mineralogical components in basalt – such as plagioclases (between -12 and -9 for Na and Ca endmembers respectively) and augite (-11.97) – under room temperature and neutral pH conditions (Gudbrandsson et al., 2011; Hermanska et al., 2022; Palandri and Kharaka, 2004).

Although this and other experiments show relatively consistent weathering rates from exchangeable base cations (comparable to those observed in lab-scale studies) we emphasize that, unlike laboratory conditions where base cations remain far from equilibrium in excess water, soils experience solid-phase cation scavenging, which promotes DIC degassing (Fig. 1). From the sum of significant TA slopes we calculate a relatively low potential inorganic CO₂ removal, equalling to approximately 0.36–0.71 kg CO₂ t⁻¹ basalt or 0.018–0.036 t CO₂ ha⁻¹ for a basalt application rate of 50 t ha⁻¹ (Table 4). Also the highest possible potential inorganic CO₂ removal realized within this experimental timeframe is modest; Including also non-significant TA increases and assuming η=1, the potential inorganic CO₂ removal is quantified to be 26.33±6.13 kg CO₂ t⁻¹ basalt (Table 4). We emphasize that a potential inorganic CO₂ removal is a maximum inorganic CO₂ removal that can be realized with the delivered amount of base cation weathering as strong acids associated with fertilizers (such as nitric acid and sulphuric acid), or organic acids and not carbonic acid may have initially weathered silicate rock which does not lead to a CDR within the soil system (McDermott et al., 2024; Taylor et al., 2021). Moreover, life-cycle emissions associated with mining, grinding and transporting rock are typically of the same order of magnitude as our relatively low potential inorganic CO₂ removal (Lefebvre et al., 2019).

Furthermore, for climate change mitigation, not only the amount of potential inorganic CO₂ removal is important, but also the timescale at which this CDR is realized (Kanzaki et al., 2025). A mass balance of base cations indicates that exported TA was negligible compared to base cation charges that were retained in the soil over the timeframe of our experiment (101 d) (Table 4). As long as TA is retained in soil pools, inorganic CO₂ removal through DIC export is delayed as equivalent amounts of protons have then been released into the soil water to maintain charge balance (Fig. 1). Realization of this delayed inorganic CO₂ removal depends on liberation of base cations from these soil pools and their transport out of the soil, charge-balanced by HCO3-. This export may take decades or longer, depending on the circumstances (Kanzaki et al., 2025).

The realization of CDR may be further delayed through the formation of base cation bearing clay minerals. Clay formation has previously been suggested for EW application based on changes in soil water Ge / Si ratios and Si isotopes (Vienne et al., 2024) and also based on Li isotope measurements (Pogge von Strandmann et al., 2022). These measurements indicated basalt induced clay formation, but it remains unclear what type of clays were formed and hence what the effect on inorganic CO₂ removal may be. In the best case for the inorganic CO₂ removal time lag, the formed clays are 1:1 phyllosilicates such as kaolinite that do not sequester base cations. In this case, DIC leaching is only retarded by base cation exchange. Worst case for the inorganic CO₂ removal time lag, the formed secondary minerals bear substantial amounts of base cations (e.g. chlorite, chrysotile, smectites, montmorillonites). These clays exhibit a log⁡Wr between -12 and -12.5 at neutral pH (Palandri and Kharaka, 2004), so that dissolution within decadal timescales is unlikely (Bullock et al., 2022).

Although unfavourable for inorganic CO₂ removal, if base cation bearing secondary clay minerals would form, they can increase SOC (Georgiou et al., 2022; Heckman et al., 2022; Steinwidder et al., 2025). Georgiou et al. (2022) refer to base-cation bearing clays (e.g. smectitic or illitic clays) as “high-activity minerals” due to their higher SOM stabilization capacity compared to secondary minerals that do not contain base cations (i.e., “low-activity minerals”, with a lower CEC such as kaolinite). Both high- and low-activity minerals can adsorb DOC and form mineral-associated organic matter-C (MAOM-C), which is believed to have a relatively high permanence (decades-centuries) in soils (Lavallee et al., 2020). Besides mineral surface however, plant inputs can also limit SOC accrual. In the latter case, SOC stocks can only increase if belowground plant C inputs increase, which could follow from increases in exchangeable bases or pH (Haque et al., 2019; Shamshuddin et al., 2011). Nonetheless, increases in decomposition can also stimulate SOC losses if rock dust increases soil pH (Klemme et al., 2022).

Different base cation monitoring strategies are possible. A first option is to quantify TA in soil water (Clarkson et al., 2024). A disadvantage is however that soil water samples have to be sampled across the soil depth. Alternatively, TA could be only monitored in top soil, yet then uncertain TA leaching models must be used (Kanzaki et al., 2025). To calibrate TA leaching models, soil measurements in depth profiles could be used.

A first soil measurement approach is a total cation accounting approach, which quantifies the loss of base cations from top soils (te Pas et al., 2025). However, this approach only focuses on the top soil and fails to account for physical cation transport from top soils due to erosion or vertical feedstock transport via infiltration or bioturbation. Alternatively, in a mobile/immobile tracer element approach (often named “TiCat” by the EW community), cation losses from amended top soils are quantified along with immobile tracers, which can account for cation losses through bioturbation or erosion (Reershemius et al., 2023). Nonetheless the disadvantage of TiCat is that it does not track potential TA scavenging (e.g. by organic matter or clays) at larger depth. Our potential inorganic CO₂ removal estimate will thus differ from a potential inorganic CO₂ removal estimate quantified using a TiCat approach.

Alternatively, entire depth profiles could be analyzed to spatially calibrate TA leaching models. Analysis of the exchangeable soil pool is already a well-established method in EW research (Beerling et al., 2024; Reershemius et al., 2023; Reynaert et al., 2023; Vienne et al., 2024). Adding also the carbonate, reducible and oxidizable soil pools to the analysis could make base cation mass balancing more complete. These protocols could calibrate predictive TA leaching models spatially. In addition there is an opportunity to quantify SOC and MAOM-C changes in the same samples, which have recently gained traction in EW research due to their role in stabilizing SOM (Buss et al., 2024; Sokol et al., 2024; Xu et al., 2024). Integration of these measurements can provide more accurate estimates of the climate impact of EW, but should take into account the difference in permanence of inorganic and organic carbon stocks.

However, this monitoring approach involves complexities such as feedstock correction, leaching solution strength and soil heterogeneity. Although correcting for pre-weathered elements was crucial in this study, it assumes perfect mixing based on a silicate-to-soil ratio. This correction was particularly significant for carbonate and reducible soil pools, where for some base cations, over half of the cation increase with basalt amendment originated from feedstock addition and not from weathering (Fig. S18). An alternative approach could involve creating time series from sequential extraction data and quantifying base cation changes based on the change in time between multiple measurements taken after rock amendment.

As discussed in the previous section, another key challenge is that the fate of base cations may remain uncertain if strongly bound crystalline organo-minerals (see Lopez-Sangil and Rovira, 2013) form that are unextractable by the Tessier scheme. Such processes may have contributed to the observed decrease in oxidizable elements at larger depth, although this could also be an artefact of the applied extraction procedure. Pogge von Strandmann et al. (2022) proposed substituting the H₂O₂ leaching step of the Tessier scheme with a dilute HCl leach, which is thought to extract clays as well. Alternatively, post-extraction analysis of residual solids using techniques such as XRD or QEMSCAN may be necessary to rigorously assess changes in rock mineralogy (Mason et al., 2022). Although deep soil core sampling and extensive mineralogical analysis are resource-intensive and not feasible for large-scale application, this monitoring strategy could be valuable during the initial adoption of EW in targeted “measure-all” experiments, as reliable TA leaching models require extensive calibration.