The field trial is situated on an eight-hectare agricultural field on the campus of Carleton College in Northfield, Minnesota (Figure 1). Since the 1930s, the field has been in near-constant cultivation, with corn and soybeans rotating as the primary crops for the past several decades. In the mid 2010s the field transitioned from conventional tillage to strip-tillage. Crop rotation during the field trial included soybeans in 2022 and 2024 and corn in 2023. Aerial images of the site show that there has been land-management variation across the site since at least 2003, with the northern portion of the field remaining out of cultivation in a subset of years (see Supplementary Figure S1). In addition, dredged sediments from an artificial lake on campus were spread across the northern portion of the field during the early 2000s (estimated extent of dredging covers the full extent of Block B and all of Block A except the easternmost slag plot). These variations in land use correspond to a gradient in soil properties across the field area (discussed below in Section 3.1), making it ideal to explore the effects of varying soil types on enhanced weathering.

Soils in the region are dominated by loam developed atop Quaternary glacial sediments or Ordovician sedimentary units (limestones, dolostones, and sandstones). In this field, soils are loam and clay loam developed atop Quaternary loess and fine loamy till mapped primarily as the Hayden loam and Blooming silt loam in the field area (Soil Survey Staff United States Department of Agriculture, n.d.). The Hayden loam is a mesic, superactive, Glossic Hapludalf and The Blooming silt loam is a fine-loamy, mixed, superactive mesic Mollic Hapludalf (Soil Survey Staff United States Department of Agriculture, n.d.).

Climate in the area is temperate, with a mean annual temperature (MAT) of 7.9 °C and mean annual precipitation (MAP) of 902 mm yr.⁻¹ (Supplementary Table S1; NOAA). Growing season climate in the area (May-Sept) has average temperature of 18.5 °C and rainfall of 574 mm (1990–2020 normals for Rice County, NOAA). Conditions during the 3 years of this experiment were generally warmer than normal in 2023 and 2024 and drier than normal in 2022 and 2023 (see Supplementary Table S1).

We used a randomized block design to distribute treatment and control plots (~0.5 acres each) across the study area to minimize impacts of soil variation on our interpretation of treatment effects (Figure 1). A set of baseline soil samples were collected in June 2021 to constrain soil properties across the field area prior to spreading of any feedstock. Baseline soil was collected at four depth intervals: 0–10 cm, 10–20 cm, 20–40 cm, and 40–60 cm. At each depth interval, a total of 12 individual soil cores were collected at semi-random locations across the plot area (~0.5 acres) and homogenized into one sample per plot (see Supplementary Figure S2). All homogenized baseline samples were air dried and sieved to remove any particles > 2 mm prior to analysis.

Each baseline soil sample was analyzed for soil pH, total soil organic matter (SOM) and SOM pools (MAOM and POM; see methods for this below in section 2.3), carbonate minerals, cation exchange capacity (CEC), and 27 element microwave digestion ICP-OES. Soil pH was measured using a Mettler Toledo Seven-Multi pH meter and InLab Routine Pro electrode on a 1:1 soil to water mixture of 10 g dried soil and 10 mL deionized water with a 15-min equilibration period. Weight percentage organic matter (SOM) and carbonate minerals were determined through loss-on-ignition (LOI) at temperatures of 360 °C and 1,000 °C respectively, with a 2-h heating period at each temperature and an intermediate heating at 105 °C to remove any residual water. For CEC and the 27 element ICP-OES, baseline samples were sent to the University of Minnesota Research Analytical Laboratory (RAL) in St. Paul, Minnesota, USA. Complete analytical methods for those analyses are available on the RAL website¹ but are briefly summarized here. CEC was calculated by summation of the exchangeable cations Ca, Mg, Na, K, and Al through an ammonium acetate extraction and subsequent evaluation of leachate using an ICP-OES. The 27-element analysis involved a microwave digestion in concentrated nitric acid following EPA procedure 3,051 and measurement using an ICP-OES. The data for the 27 element ICP-OES analysis is not shown in a figure but is available in the Supplementary data.

Feedstock used for treatments in this study was basalt and steel slag. The basalt is a commercially available quarry sand product derived from an aggregate quarry in the Chengwatana Volcanics of the Keweenawan Supergroup, a sequence of lavas, tuffs, and interflow sedimentary rocks associated with the Neoproterozic (~1.1 Ga) failed rifting of the midcontinent (Wirth et al., 1997). The material is dominantly coarse to medium grained sand (~63% passing 0.5 mm sieve) with 20% fine sand (< 0.25 mm) or finer, placing this on the coarse side of feedstocks used for EW trials. Geochemically, the basalt is 15% Ca + Mg and the mineralogy is dominated by plagioclase, diopside/augite, chlorite, and epidote. The presence of epidote and chlorite is related to low-grade metamorphism associated with hydrothermal fluids (Wirth et al., 1997). The steel slag used in this study is a Ca-rich byproduct of steel production. The material is ~50% CaO + MgO with mineralogy dominated by Ca and Mg rich silicate, (hydr)oxides, and amorphous phases. Grain size is finer in comparison to the basalt, with most of the steel slag gradation being medium sand (~65% < 0.6 mm) and ~25% being very fine sand to silt (~27.3% < 0.15 mm). Both the basalt and steel slag were byproducts of industrial processes. Chemical, mineralogical, and grain size data for the feedstocks is available in the Supplementary file.

Spreading of feedstock materials was completed in November 2021. Four treatment plots were amended with basalt at a target rate of 10 tonnes acre⁻¹ and four plots with slag at a target rate of 2 tonnes acre⁻¹ using commercial ag lime spreading equipment. To validate application rates, six pie pans were placed in each plot at random before spreading (for pan test results and average spreading rates see Supplementary Table S2). The lower application rate for steel slag reflects the much higher liming capacity of steel slag relative to basalt, primarily due to the abundance of hydroxide mineral phases in the steel slag. Care should be taken to not over apply steel slag purely for CDR, as application rates should be tailored to properly adjust soil pH depending on local conditions.

Soil samples were collected each year of the three-year field trial. Various sampling strategies were adopted in later years to constrain effects of the study treatment and deal with background soil variability. Below, sampling strategies are described for each sampling year: 2022, 2023, and 2024. While all soil data is included in the Supplementary files, we only discuss soil data in detail from the end of the experiment—sampling year 3 (2024). Below, we briefly describe both sampling strategies and conclusions from evaluating soil in the intermediate years of this study.

In October of 2022 (year 1), soil was collected at 0–10 cm and 10–20 cm depth increments for all treatment plots using the same sampling strategy as the baseline soils. These soils were evaluated for soil pH, organic matter, CEC, and the 27 element ICP-OES, all measured at the RAL facility at the University of Minnesota using methods described above.

In 2023 (year 2), soil was collected during October 2023 at a higher resolution in the southern blocks (C and D) of the field area to determine if higher resolution sampling would reveal a clearer picture of treatment effects. Within an experimental plot, 15 sampling points representing a circular area with a 3 m radius (~28.3 m² area) were established (see Supplementary Figure S2). At each sampling point, fifteen 0–10 cm soil cores were collected and homogenized into a single sampling point sample. Soil pH and exchangeable base cations (Ca and Mg) for year 2 soils significantly increased in both properties in the slag treatment relative to both the control and basalt treatments (see Supplementary Figure S3). The basalt treatment was not statistically different from control for either property (Supplementary Figure S3). Given the improved resolution using higher density sampling, this approach was adopted for evaluating soil properties in year 3 of this project.

In year 3 (2024), all plots were sampled from 0 to 10 cm at 15 sampling points per plot (following scheme outlined above and in Supplementary Figure S2). Year 3 soils were evaluated for soil organic matter (SOM; weight percentage; based on loss-on-ignition at 360 °C), soil pH (1:1 slurry), a series of extractable elements using the Mehlich 3 extraction including P, K, B, S, Fe, Mn, Cu, Zn, and exchangeable base cations Ca, Mg, K, and Na. Cation exchange capacity (CEC) and base saturation is reported for each sample. Analytical work on all soils from year 3 of this study was conducted by Waypoint Analytical, LLC (Champaign, Illinois, USA).

For baseline soils (2021), year 1 soils (2022), and year 3 soils (2024) we also evaluated effects on different SOM pools: mineral-associated organic matter (MAOM) and particulate organic matter (POM). We separated MAOM and POM pools using a standard physical fractionation method (Sokol et al., 2017; Bradford et al., 2008). First, 5-g of air-dried soil was weighed into a 50-mL falcon tube, and 30-mL of 5% sodium hexametaphosphate (NaHMP) was added to disperse aggregates. Samples were then vortexed and placed on a shaker table for 16 h at 200 rpm. After shaking, the soil solution was passed through a 53-μm sieve. The < 53-μm fraction (clay + fine silt fraction; defined as MAOM) and the > 53-μm fraction (sand, coarse silt, and large organic matter fragments; defined as POM) were oven-dried to constant mass at 105 °C. Dried samples were finely ground to a powder-like consistency with a mortar and pestle and analyzed on an elemental analyzer (Lawrence Livermore National Laboratory, Livermore, CA).

We determined the trace and major element composition of the mineral-bound phases to constrain cation loss from the soil surface (e.g., a soil mass balance or TiCAT methodology Reershemius et al., 2023). This approach uses a mass-balanced based mixing model that constrains mineral weathering rates in enhanced weathering feedstocks. As minerals in applied feedstock are weathered, there is a preferential release of mobile base cations relative to immobile elements like titanium. It can be feasible to disentangle the loss of mobile base cations from a soil-feedstock mixture that has weathered relative to initial baseline measurements (Reershemius et al., 2023). All soil and feedstock samples evaluated using this approach were analyzed for isotope dilution inductively coupled mass spectrometry (ID-ICP-MS) at Yale University following methods in Reershemius et al. (2023).

In each growing season of the study, lysimeters were used to collect shallow (0–20 cm) soil porewater. In 2022, porewater monitoring occurred in each of the 12 experimental plots. Porewater in the northern experimental plots with high initial soil pH (> 7; all of Block B and Block A basalt and control plots) had considerably elevated porewater alkalinity compared to soils with more acidic baseline soil pH, likely due to soil carbonates dissolving (see Supplementary Figure S4). Due to difficulties in acquiring sufficient water for analysis, all additional porewater monitoring was conducted on experimental plots with acidic initial soil pH, which includes all the southern plots in blocks C and D along with the slag-treatment plot in the easternmost portion of Block A (see Figure 1 for plot and block layouts). In addition, a set of lysimeters was placed in the buffer strip area between block C and D to serve as additional porewater control given the lack of replicates for control in acidic soil areas.

Porewater was collected weekly to bi-weekly during a 10-week monitoring period between mid-June and mid-August 2022, 2023, and 2024. In each plot, a minimum of 3 suction cup lysimeters were installed at depths of ~15 cm using a soil slurry. A plastic apron was installed around the surface of the lysimeter soil connection to avoid any direct flow paths of water adjacent to the lysimeters. Following installation, the first round of water collected was discarded to avoid any contamination caused by the installation soil slurry. Dry periods during each collection season created gaps in data due to lack of available water. Collected water was combined for all lysimeters within a single experimental plot and refrigerated immediately after collection. Prior to analysis, all water was filtered using a 0.45 μm filter. For elemental analysis, water samples were spiked to 1% nitric acid for preservation shortly after collection.

Total alkalinity and pH were measured on all samples with sufficient sample volume using a Hach AT1000 automated titration system with hydrochloric acid. Units for total alkalinity were initially reported as mg CaCO₃ L⁻¹. Reported units here are in milliequivalents (meq) L⁻¹. We make the conservative assumption that for samples with pH < 8.3, the total alkalinity is equivalent to the bicarbonate alkalinity (Holzer et al., 2023; Neal, 2001). In 2023, to mitigate limited sample volumes, we conducted alkalinity titrations on diluted samples. Based on regular measurements of alkalinity in reverse osmosis (RO) water used in dilutions, we calculated the samples total alkalinity assuming linear mixing. Because RO waters used in dilutions had relatively low pH (~6), there likely was a minor consumption of alkalinity in samples that were diluted—however this does not appear to be a large effect. For samples that were diluted, we often were able to measure the same sample 2–3 times and use an average value for reporting. When not diluting it was often not possible to obtain duplicate or triplicate measurements of the same sample because of minimum sample volumes (~50 mL) when using pure samples. In comparing alkalinity data, we evaluate measurements throughout the season rather than overanalyzing any one collection interval in part due to these complications.

Major anions, cations, and metals were evaluated throughout the three seasons of monitoring. In all three summers, all water samples were evaluated for a suite of 27 elements using an ICP-OES at the RAL at the University of Minnesota. Here, we report data from this analysis for Ca and Mg, but all elemental data is included within our Supplementary dataset. Notably, across all treatments and monitoring years, there are rarely detectable metals (Cr, Pb, Ni, etc) in soil porewaters (see Supplementary datafile). In 2022, nitrate was measured using colorimetric methods at the RAL at the University of Minnesota. In 2023 and 2024, a set of major anions, including nitrate, sulfate, phosphate, and chloride were measured using an Aquion Ion Chromatography system from ThermoFischer Scientific.

In 2024, water samples were evaluated for total carbon (TC), total organic carbon (TOC), and total inorganic carbon (TIC) at the RAL at the University of Minnesota as an independent check on dissolved carbon separate from alkalinity. Samples were first measured for TC in an Elementar TOC Select Combustion Analyzer. Following an initial TC measurement, an aliquot of the same sample was acidified to remove inorganic carbon and remeasured to quantify TOC. TIC was quantified by subtraction, where TIC = TC-TOC.

In each of the first two growing seasons (2022 and 2023), soil gas flux was measured using a combination of instruments, chamber types, and soil collar distributions. In 2022, we measured CO₂ gas flux from soils across the study area using an EGM-5 from PP Systems (Amesbury, Massachusetts, USA). Measurements were conducted weekly to bi-weekly at multiple locations throughout each plot. For each measurement, a survey chamber was sealed gently to either the bare soil or to a pre-installed soil collar (PVC pipe) and CO₂ concentration was monitored in the closed chamber for 90 seconds. The flux of CO₂ was then calculated using the change in temperature, air temperature, pressure, and volume of the chamber using standard methods (PP Systems, 2018). In 2023, we measured CO₂ gas flux from soil using both the EGM-5 system and a Gasmet GT5000 Terra multigas FTIR gas analyzer (Gasmet Technologies Inc.; Vantaa, Finland). Functionally, both instruments measure change in CO₂ concentration in a closed chamber over time. In 2023, all measurements were made on pre-installed soil collars at standard locations across each plot.

Baseline soil properties across the different blocks were variable (see Supplementary Figures S5, S6). Blocks A and B in the northern portion of the field trial had higher soil pH, organic matter, carbonate mineral content, and CEC compared with the southern Blocks C and D (Supplementary Figure S5). One exception to this pattern was the slag treatment plot on the eastern side of Block A (Figure 1) which had soil properties that more closely aligned with Blocks C and D (Supplementary Figure S5).

Comparison between soil from baseline and intermediate sampling years (2022 and 2023) revealed considerable variability (see Supplementary Figures S6, S7 for example with soil pH). Here, we focus on surface soils (0–10 cm) collected at higher sampling density in year 3 (2024, see Supplementary Figure S2). In the more acidic soil in blocks C and D, both soil pH and Ca-saturation were significantly higher for slag treated soils compared against control (Figure 2; p-values < 0.05 for Tukey HSD and ANOVA tests). For soils with more neutral initial pH (blocks A and B), we observed no consistent difference in any soil properties in enhanced weathering plots vs. control plots (Figure 2). For acidic soils in blocks C and D, slag treatments showed shifts in soil pH and Ca-saturation relative to baseline (boxplots compared with grey squares in Figure 2); statistical tests made difficult due to variable sampling strategies.

There was no impact of basalt or slag on total SOM (p > 0.05 for Tukey HSD and ANOVA tests; Figure 2). There was also no effect of basalt or slag treatment on the change in MAOM-C (mg C g soil⁻¹; p > 0.05 for Tukey HSD and ANOVA tests) or POM-C (mg C g soil⁻1; p > 0.05 for Tukey HSD and ANOVA tests) pools relative to baseline for both year 1 (2022) and year 3 (2024) soils (Figure 3).

Variation in soil porewater occurred throughout each growing season, in part due to changes in weather (rainfall and temperature), plant activity, and fertilizer applications. In Supplementary Figures S8–S10, major porewater chemical properties including pH, alkalinity, nitrate, and base cations (Ca and Mg) are shown with time for each growing season.

When taken in aggregate across each growing season, porewater pH and alkalinity were significantly elevated in acidic soils treated with steel slag relative to control (Figure 4; all differences are significant at the p < 0.05 level using a Tukey HSD and ANOVA test). There was an increase in Ca and Mg concentrations in slag treated relative to control porewaters in 2022 and 2023, but the difference was not significant in either year (p = 0.63 and p = 0.14 in 2022 and 2023, respectively; see Figure 4). We observed no difference between porewater chemistry in the acidic soils treated with basalt during the 2022 and 2023 growing season, and lysimeters were repurposed in 2024 to focus only on continued monitoring of steel slag treated soil compared to control soil. There were consistent, measurable concentrations of strong acid anions such as nitrate (NO₃⁻) throughout the trial, but there were no significant differences between non-alkaline anion concentrations observed in any of the monitoring years. However, as with calcium and magnesium concentrations, nitrate concentrations in slag treatments in 2023 trended higher than control plots (p = 0.2; see Figure 4).

In the cases where porewater was measured for heavy metals including Cr, Pb, and Ni, these metals were often below the analytical detection limits (data for Cr, Ni, and Pb are available in the Supplementary file). Ni is the only metal with enough values for statistics, and there was no significant difference with treatment (p-values > 0.05 a Tukey HSD and ANOVA test).

There was no significant difference in the CO₂ gas flux between basalt-amended or slag-amended soils and control soils when measured using survey chambers with or without soil collars during the 2022 and 2023 monitoring season (p > 0.05 for Tukey HSD and ANOVA test; Supplementary Figure S11).

CDR occurs in enhanced weathering through mineral dissolution buffering acidity in soil. In soil porewaters, pH and base cation concentrations should increase as mineral weathering occurs (see Equation 2 above). If carbonic acid is directly buffered, bicarbonate (HCO₃⁻) formation will drive an immediate increase in the alkalinity of the system. In soils amended with steel slag, we observed a significant increase in porewater pH and alkalinity in blocks with acidic soil (Blocks C and D) across 3 years of monitoring (Figure 4; p < 0.05 for Tukey HSD and ANOVA). While there was an apparent increase in base cation concentration that accompanied the changes in pH and alkalinity, this effect was only significant in 2024 (Figure 4). In addition to soil porewater, soil pH and Ca-saturation for steel slag treatments in Blocks C and D increased significantly in Year 3 (2024) relative to control (Figure 2), providing further support for sustained weathering and CDR over the three-year monitoring period. However, we did not observe significant effects of steel slag in soils with neutral pH (Blocks A and B) nor basalt treatments regardless of initial soil pH (Figures 2, 4).

In the acidic soils (Blocks C and D), there was consistent evidence for increased pH and increased alkalinity export in steel slag-amended plots, even at a relatively low application rate (2 tonnes acre ⁻¹) compared to the much higher application rate of crushed basalt (10 tonnes per acre⁻¹). This highlights the importance of considering the reactivity of materials in feedstock selection. Steel slag is composed of calcium and magnesium oxides and hydroxides, Ca-silicates, and Ca-rich amorphous phases that are highly weatherable and alkaline, making it attractive from the perspective of pH management and CDR (Das et al., 2020; Das et al., 2019; O’Connor et al., 2021). However, steel slag availability in the United States is limited and a large majority of steel slag produced is already utilized within construction markets. Legacy steel slag piles could represent considerable volumes of material but may have more complicated production histories and subsequent potential for heavy metal contamination.

While many studies have focused on basalt as an attractive feedstock for enhanced weathering, the results presented here fail to demonstrate any impact from the applied basalt. This is likely due to the physical and chemical characteristics of the particular basalt used in this study. The grain size of the quarry sand product was likely too coarse-grained to weather measurably in the field over a 3-year period. In addition, the metamorphic minerals epidote and chlorite present in our basalt suggest low-grade metamorphism that may lessen the materials reactivity. In addition, our application rate for basalt (10 tonnes per acre) was relatively low compared to other studies using basalt, and basalt was only applied a single application in the first year, compared to other enhanced weathering deployments with annual applications at higher rates (Kantola et al., 2023; Larkin et al., 2022). The lower total application volume could have limited our ability to detect a weathering signal. Higher application rates come with agronomic and economic trade-offs, especially considering typical application rates for ag lime are just 0.5–2 tonnes per acre for pH management. Increased trucking costs and soil compaction for spreading large volumes of material, particularly for annual applications, are important considerations. This is an obvious documentation that some fine grained commercially available basalt products will not be suitable for EW deployments. Future work to evaluate new basalt products at finer grain sizes in similar settings should help to determine if weathering signals can be detected at lower application rates with basalt.

There are two plausible explanations for why simple measurements failed to detect impacts from enhanced weathering amendments in soils with neutral pH. First, silicate weathering is pH-sensitive and tends to slow in neutral soils, where lower proton availability reduces mineral dissolution and the release of weathering products. Second, more neutral soils can contain soil carbonates (Supplementary Figure S5) which appear to dominate the soil porewater chemistry (Supplementary Figure S4) and have faster weathering kinetics than silicates. This is consistent with studies monitoring chemical weathering in catchments based on stream water chemistry, where carbonate weathering often predominates the weathering signal. Given the more rapid weatherability of carbonates, they require less rock material to be produced, transported, and spread to achieve a similar buffering effect for soil acidity compared to most silicate feedstocks for enhanced weathering. Prior work that evaluated the carbon removal effects of widespread ag lime usage (Hamilton et al., 2007; Oh and Raymond, 2006), and more recent work on carbonate-rich returned concrete as a feedstock for enhanced weathering (McDermott et al., 2024), both support relatively rapid rates of carbonate weathering that can lead to CDR. Given the requirements of voluntary carbon markets for comparison of project activities against a counterfactual (what would have happened absent the project), quantifying net CDR should consider comparing silicates against carbonates in areas where ag lime use is common (e.g., the Midwest U.S.).

Increased Ca-saturation and soil pH provides direct evidence that steel slag is actively weathering and releasing base cations into the soil environment. Given that Ca-saturation increased (Figure 2), at least a portion of the released calcium is adhering to exchange sites, rather than flushing through the soil column. The timescale and rate at which this calcium will migrate through the soil column is not clear, but will be important for full delivery of CDR (Kanzaki et al., 2025). To further quantify rates of cation release and weathering, we applied the Ti-CAT methodology (Reershemius et al., 2023). However, the results were inconclusive using both the high- and low-resolution soil sampling methodologies described above (see Section 2.3 and Supplementary Figure S2). This could be due in part to our relatively low application rates, low Ti abundance in steel slag, or lack of high-resolution sampling for baseline soil prior to application for a direct comparison to weathered mixtures. This serves as an example of the importance of high-resolution sampling for soil-based mass balance approaches.

Inorganic CDR that results from mineral weathering may be synergistic with processes that encourage the storage of organic carbon in soils, similar to the relationship between rock weathering and soil organic carbon accrual over long time scales (Slessarev et al., 2021). However, it remains unclear how different feedstocks will interact with organic carbon cycling in different soil types and climates, and on different timescales. Here we evaluate the impacts of enhanced weathering on soil organic carbon cycling by evaluating changes in total SOM, MAOM and POM pools, and monitoring soil CO₂ gas flux from the surface soil.

Both basalt and steel slag did not have any positive or negative effects on total SOM, or MAOM and POM pools (Figures 2, 3). There was an apparent positive increase in MAOM-C in steel slag-amended treatments in neutral soil (Blocks A and B), but not in acidic soil in southern plots (Blocks C and D; Figure 2). In contrast to most other results where steel slag weathering in more acidic soils in Blocks C and D produce significant results, perhaps steel slag in more neutral soils could have a positive effect on MAOM formation.

Interestingly, while steel slag applied to acidic soils (Blocks C and D) led to significant changes to porewater chemistry that indicates weathering and CDR, the increase in MAOM-C in more neutral soils (Blocks A and B; see Figure 3) treated with steel slag could highlight differing responses of inorganic and organic C pools to enhanced weathering feedstocks based on initial soil pH. In the more neutral soils, slag-derived Ca may have promoted MAOM formation through mechanisms like cation bridging, aggregate stabilization or by enhancing microbial transformation of organic inputs (Kang et al., 2024; Rowley et al., 2018; Shabtai et al., 2023). Even if indicators of weathering in soil porewaters were masked by high background carbonate weathering, Ca released from slag could still interact with native SOM. Additionally, higher initial SOM could support MAOM accrual through organo-organic associations (Kang et al., 2024; King, 2025). Future work should aim to determine the effects on SOM pools with higher resolution sampling and better replicated trials that vary in soil pH.

Soil CO₂ flux could be influenced by enhanced weathering directly or indirectly. The direct effect of increasing consumption of soil CO₂ through mineral weathering could lead to a decrease in the amount of CO₂ released from the soil surface. In addition, enhanced weathering-induced increases to soil pH may alter soil microbial activity and influence rates of gas flux in soils (Malik et al., 2018; Vermeiren et al., 2025; Fernández-Calviño and Bååth, 2010). However, there were no significant differences in CO₂ flux for any treatments in this study. The lack of any significant difference between basalt and steel slag treatments and control plots suggests there are not any significant losses of carbon through increase soil CO₂ flux resulting from enhanced weathering. Although these results are inconclusive, they do not support the hypothesis that mineral addition leads to soil organic carbon loss.

As for any direct impact, the background variability in soil CO₂ flux from biological respiration likely masks any signal from inorganic CDR via enhanced weathering. Diurnal cycles can show variation of ~3–5 μmol CO₂ m² s⁻¹, or even more. Spatially, there is also considerable variation even within a given area or experimental plot. Even a low-end rate of CO₂ flux in soils from this study of just ~2 μmol CO₂ m² s⁻¹ implies an annual CO₂ flux of ~7.5 tCO₂ acre⁻¹ yr.⁻¹ being respired out of soils over a frost-free 9-month period. For enhanced weathering amendments to have a chance at being detectable via gas flux measurements, annual rates of carbon removal would need to be perhaps 10–25% of this rate, implying 0.75–1.9 tCO₂ acre⁻¹ yr.⁻¹, or higher depending on resolution of measurements. Automated chambers with appropriate spatial resolution and replication could perhaps help to reduce the signal-to-noise issues we encountered here, although existing commercial chambers are expensive and unlikely to be applicable at scale without development of low-cost sensor alternatives. Notably, recent work used an automated array of CO₂ sensors embedded in the soil profile to directly detect decreases in CO₂ flux based on application of basalt in an enhanced weathering trial in Viriginia (Milliken et al., 2025), suggesting this approach could be developed into an effective monitoring strategy.

Soil pH management in agriculture is dominantly a response to the generation of strong acids from the degradation of fertilizers. As an example, ammonia (NH₄⁺), a common ingredient in nitrogen fertilizers, releases acidity during oxidation and forms nitric acid (HNO₃), which dissociates into acidity (H⁺) and nitrate ions (NO₃⁻) in most soil environments (Equation 3).

Like carbonic acid, strong acids weather silicate minerals and release base cations (Equation 4 below). Instead of achieving charge balance with bicarbonate, strong acid anions (e.g., nitrate, NO₃⁻; phosphate, PO₄²⁻; sulfate, SO₄²⁻) in porewaters charge balance cations. To illustrate this process, consider the stoichiometric reactions below (Equations 4–6):

Given that carbonic acid neutralization and bicarbonate generation is typically thought of as a key mechanism for delivering CDR in enhanced weathering, strong acid weathering introduces complexity for quantifying CDR in enhanced weathering deployments.

Pure carbonic acid weathering should yield a 1:1 equivalence ratio between base cations and alkalinity (bicarbonate; Equation 4), while strong acid weathering will act to increase this ratio by generating an abundance of base cations compared to alkalinity (McDermott et al., 2024; Hamilton et al., 2007). In each year of this study, base cations in soil porewaters exceeded measured alkalinity (see Figure 5). Nonetheless, in all cases, the base cation to alkalinity ratios exceeded 1:1 and in many cases exceeded 2:1 (expected for a mixed acid weathering regime that includes carbonic acid, Equation 6; Figure 5). The apparent strong acid effects were most pronounced in 2023 when the field was planted with corn, requiring the addition of N-based fertilizers. In 2022 and 2024, the field was planted with soybeans, which do not require additional N-fertilizer. Nitrate in soil porewaters often better aligns with base cation concentration (Figure 5), indicating the importance of N cycling in agricultural soil acid base balances.

Alkalinity loss (ΔHCO₃⁻) is defined as the reduction in alkalinity due to the charge balance of base cations with strong acid anions rather than bicarbonate (Perrin et al., 2008). We assume titrated alkalinity is equivalent to the concentration of HCO₃⁻ (in units of meq L⁻¹; referred to below as HCO₃⁻_measured). The theoretical concentration of HCO₃⁻ (HCO₃⁻_theoretical) is assumed to be equal to the concentration of Ca and Mg (in units of meq L⁻¹) and ΔHCO₃⁻ is calculated as in Equation 7:

Comparison of alkalinity loss to nitrate concentrations reveals a strong linear relationship that is close to a 1:1 line (see Figure 6), suggesting that N-fertilizers are dominantly responsible for alkalinity losses. The observation of this correlation in controls, as well as treated soils, highlights that acid addition in many soils likely leads to alkalinity loss as bicarbonate neutralizes acidity.

In acidic soils with low pH, the use of silicate minerals to buffer acidity and bring soils back towards neutral would be particularly attractive relative to carbonate rocks as there would be no risk of releasing fossil CO₂ from carbonates during neutralization.

The weathering of silicate minerals with strong acids does not lead to a direct removal of CO₂ and subsequent formation of bicarbonate. However, the buffering effect of a silicate mineral on strong acids like nitric acid should decrease CO₂ evasion due to acidification and help to retain bicarbonate alkalinity. For instance, consider the carbonate system present in soils (Equation 8):

The net effect of acidification due to nitrogen fertilizers would be to add H⁺, pushing the equilibrium towards CO₂ (Equation 8). From a purely inorganic carbon perspective then, acidified systems will degas CO₂. Buffering acidity (H⁺) from any source of acid (carbonic acid or, in this case, nitric acid) should limit CO₂ degassing and drive the carbonate system towards bicarbonate thereby encouraging the stability and preservation of alkalinity. The results here support the inclusion of robust accounting for strong acid effects and acid neutralization in frameworks for greenhouse gas accounting for enhanced weathering deployments.