Increased yield and CO2 sequestration potential with the C4 cereal Sorghum bicolor cultivated in basaltic rock dust‐amended agricultural soil

Land‐based enhanced rock weathering (ERW) is a biogeochemical carbon dioxide removal (CDR) strategy aiming to accelerate natural geological processes of carbon sequestration through application of crushed silicate rocks, such as basalt, to croplands and forested landscapes. However, the efficacy of the approach when undertaken with basalt, and its potential co‐benefits for agriculture, require experimental and field evaluation. Here we report that amending a UK clay‐loam agricultural soil with a high loading (10 kg/m2) of relatively coarse‐grained crushed basalt significantly increased the yield (21 ± 9.4%, SE) of the important C4 cereal Sorghum bicolor under controlled environmental conditions, without accumulation of potentially toxic trace elements in the seeds. Yield increases resulted from the basalt treatment after 120 days without P‐ and K‐fertilizer addition. Shoot silicon concentrations also increased significantly (26 ± 5.4%, SE), with potential benefits for crop resistance to biotic and abiotic stress. Elemental budgets indicate substantial release of base cations important for inorganic carbon removal and their accumulation mainly in the soil exchangeable pools. Geochemical reactive transport modelling, constrained by elemental budgets, indicated CO2 sequestration rates of 2–4 t CO2/ha, 1–5 years after a single application of basaltic rock dust, including via newly formed soil carbonate minerals whose long‐term fate requires assessment through field trials. This represents an approximately fourfold increase in carbon capture compared to control plant–soil systems without basalt. Our results build support for ERW deployment as a CDR technique compatible with spreading basalt powder on acidic loamy soils common across millions of hectares of western European and North American agriculture.

determine efficacy, co-benefits, environmental risks, scalability and costs of such CDR strategies (EASAC, 2018;NRC, 2015;UNEP, 2018). Enhanced rock weathering (ERW) is a land-based CDR strategy addressing multiple UN Sustainable Development Goals  which involves amending agricultural soils with crushed abundant silicate rocks (e.g. basalt) to accelerate carbon capture. Implemented on croplands it has potential to increase production, protect against pests and diseases and assist in restoring acidified nutrient-depleted soils (Beerling et al., 2018;Hartmann et al., 2013;Kantola, Masters, Beerling, Long, & DeLucia, 2017;Zhang, Kang, Wang, & Zhu, 2018). The applied silicate minerals undergo reactions with CO 2 in the rhizosphere, releasing base cations (e.g., Ca 2+ , Mg 2+ ) and alkalinity. Depending on soil chemistry, this can result in either the formation pedogenic carbonates or be delivered to the oceans via run-off; both routes store carbon with an estimated lifetime of tens of millennia (Hartmann et al., 2013;Renforth & Henderson, 2017). Minimizing adverse effects of ERW from mining, including associated impacts on biodiversity (Edwards et al., 2017) is contingent on exploiting existing silicate rock waste streams, for example, from quarries and legacy reserves, and artificial silicates (Beerling et al., 2018;Das, Kim, Hwang, Verma, & Kim, 2019;Manning, Renforth, Lopez-Capel, Robertson, & Ghazireh, 2013). Undertaken at large scale, ERW is a CDR strategy that may have potential to support food and soil security and help avert ocean acidification (Hartmann et al., 2013;Köhler, Hartmann, & Wolf-Gladrow, 2010;Taylor et al., 2016).
Numerical modelling investigations (Köhler et al., 2010;Taylor et al., 2016) have assessed the CDR potential of ERW largely in the context of tropical forests, but the focus is now turning towards croplands and soils on which deployment with existing land-spreading technology should be feasible (Beerling et al., 2018;Kantola et al., 2017;Strefler, Amann, Bauer, Kriegler, & Hartmann, 2018). Experimental and field trial programmes are thus underway examining CO 2 removal rates and feedbacks on crop performance and soil health. Such trials have reported rates of CO 2 removal following amendment of different agricultural soils across a range of application rates with the fast-weathering magnesium silicate, olivine Dietzen, Harrison, & Michelsen-Correa, 2018;Renforth, Pogge von Strandmann, & Henderson, 2015;ten Berge et al., 2012). However, CO 2 removal by weathering of olivine-dominant dunite is often accompanied by increases in potentially toxic nickel (Ni) and chromium (Cr) concentrations that may be problematic for agricultural applications. Basalt rock is a proposed alternative silicate, containing at least six plant-essential nutrients (K, P, Ca, Mg, Fe and Mn) and very low concentrations of Cr and Ni (Beerling et al., 2018;Hartmann et al., 2013;Kantola et al., 2017). Although silicon (Si) is regarded as a non-essential element for plants, cereals, the most important crops globally, accumulate Si in their straw and stover, with benefits for crop yield and resistance to abiotic and biotic stress (Debona, Rodrigues, & Datnoff, 2017;Guntzer, Keller, & Meunier, 2012). When crop residues are not returned to the land, this depletes the plant-available Si pools and can constrain yields (Haynes, 2017). Amending highly weathered soils of tropical agricultural regions with basalt increased yields of crops, and improved soil health (e.g. increased soil pH and cation exchange capacity [CEC]; Beerling et al., 2018;Edwards et al., 2017;Hartmann et al., 2013;Van Straaten, 2006).
Additionally, soils produced from basaltic quarry fines have inorganic carbon sequestration potential via in situ carbonate formation, especially when mixed with organic matter (Manning et al., 2013). However, the amendment of soils with crushed basalt on crop production outside the tropics, possibly substituting in part for expensive rock-derived P-and K-fertilizers (Amann & Hartmann, 2019;Beerling et al., 2018), and its CDR potential remain to be assessed.
Here we investigate ERW carbon removal and co-benefits with the C 4 crop Sorghum bicolor grown at bench-scale in mildly acidic clay-loam agricultural soil amended with basaltic rock dust.
Sorghum is one of the most widely cultivated cereal crops in the world and the fifth most important crop for food and animal feed, with the United States being the leading producer (Turhollow, Webb, & Downing, 2010). It also provides a representative functional type similar to the C 4 crop maize suitable for cultivation at the mesocosm scale under controlled environment conditions. Sorghum and maize are often grown in slightly acidic clay-loam soils that are important agricultural soils of western Europe and the United States. Depending on fertilizer treatment Sorghum can form symbiotic associations with arbuscular mycorrhizal (AM) fungi (Hindumathi & Reddy, 2011) that are important biotic agents of mineral dissolution . Our experimental approach, therefore, is a first step towards understanding the potential applicability of ERW with basalt in a temperate-zone agricultural context which will be important for understanding suitability for large-scale deployment of this CDR strategy (Beerling et al., 2018;Haque, Santos, Dutta, Thimmanagari, & Chiang, 2019;Zhang et al., 2018). Basaltic mineral dissolution and carbon sequestration potential were assessed with elemental budgets (plants-soil-leachate pools). Over longer multiyear time horizons, we assess dissolution trajectories and resulting CDR using a 1 D reactive transport soil profile geochemical model calibrated to our experimental results, measured basalt mineralogy and particle size distribution.

| Weathering reactor design
Experiments were conducted with replicated mesocosms housed within the University of Sheffield controlled environment facility.
Weathering reactors were constructed from two lengths of 152 mm internal diameter polyvinyl chloride pipe, coupled by a polyurethane-sealed joint, where a 600 mm long upper-section contained the soil and a 400 mm long base-section housed the leachate collection assembly ( Figure S1a). A 20 mm diameter nylon nozzle was set into the bottom of the soil-containing upper-section, to direct leachate into a 1 L amber Winchester bottle via a polytetrafluoroethylene funnel. A 50 mm deep drainage layer located at the base of the columns was used to prevent migration of soil particles into leachate nozzles, and comprised 5 mm diameter polyethylene beads placed on a high-density polyethylene screen (2 mm mesh size). Algal growth in the Winchester collection bottles, was minimized by enclosing them in detachable opaque plastic membranes. Column soil surfaces were dressed with a 50 mm deep polyethylene bead layer to reduce evaporation and minimize moss growth.

| Soil characterization and preparation
Mildly acidic soil (pH = 6.6; measurement protocol in Section 2. base saturation, and a low total organic carbon (TOC) content of 1.2% (wt%).
Prior to filling mesocosms, soil was manually comminuted by removing large stones and invertebrates, and then by passing through a rotary soil sieve (20 mm mesh size; model: CRS400, Clarke Power Products). A sub-and topsoil arrangement was engineered in the upper-section of the weathering reactor to simulate a tillage regime (Figure S1a), composed of 4.3 and 5.0 kg of soil, and compacted to an approximate bulk density of 1.05 and 1.2 kg/L respectively.
The effective soil column length was 500 mm. Well-characterized crushed basalt (see Section 2.6; total mass of 181.5 ± 0.5 g per column; equivalent to 10 kg/m 2 ) from a typical volcanic arc mountain in the Cascade Range, Oregon, United States, was mixed with soil in the upper 250 mm of the treated columns (n = 6) to simulate plough layer mixing depth. This application rate is towards the upper limit of the practical range but lower than that used in a recent modelling study (15 kg/m 2 ; Strefler et al., 2018). Into this layer, we mixed a commercial AM fungal inoculum (25 g per column) to promote establishment of this symbiosis (PlantWorks UK Ltd). Planted weathering reactors containing only soil and mycorrhizal inoculum (without the basalt amendment) were maintained as the control treatment (n = 6). We also buried basalt-filled nylon mesh (35 µm diameter) bags (50 × 50 mm) in the top 250 mm depth of the soil columns (2 per column) filled with 3 g of sieved and washed basalt grains (50-150 µm size fraction). The mesh was chosen to exclude roots but permit access to fungal hyphae from mycorrhiza. These bags were recovered at the end of the experiment to determine changes in surface chemistry and analysis of the microorganisms colonizing the rock grains.

| Plant materials
Dwarf hybrid S. bicolor (Pennsylvania 115; Oakbank Game and Conservation) was cultivated in the weathering reactors.
Approximately 300 seeds were pregerminated on filter paper (Whatman #1), moistened with distilled water, placed in zip-lock plastic bags and incubated at 25°C in darkness for 24 hr. About 100 seeds with roughly equal radicle length (ca. 5 mm) were selected for planting in seed-trays containing the experimental soil, transferred to the growth room, and maintained at 60%-70% relative humidity, with 18 hr/day at 25/17°C day/night temperatures.
After 24 hr, 30 seedlings emerged from the soil, of which the 12 largest plants were grown for a further 14 days, before being randomly assigned for transplantation to weathering reactors in the growth room maintained under the same controlled environment conditions. Plants were grown under a photosynthetically active radiation (400-700 nm) of 800 µmol photons m −2 s −1 , and for the first 60 days from emergence to boot stage, the day length was 18 hr but then shortened to 10 hr for the final 61 days until harvest. Plants were harvested at physiological maturity, segregated into root, shoot and seed parts, and dried in a forced-air oven at 60°C for 24 hr.

| Irrigation regime and effluent collection
Reactors were irrigated on a 5 day rotation basis using reverse osmosis water delivered via a bespoke drip-feed irrigation system at 12.5 ml/min ( Figure S2, total irrigation amount of 13.9 L over 120 days). Urea was dissolved in irrigation water (total 0.71 g prilled urea per reactor during 120 day experiment) and supplied periodically to each reactor with the total addition rate representative of high yielding croplands worldwide (180 kg N/ha; Potter, Ramankutty, Bennett, & Donner, 2010). No P-or K-fertilizer was added during the experiment. Irrigation water supply was adjusted to produce about 150 ml of leachate about 24 hr after each irrigation event ( Figure S2).
Mean infiltration rate during the experiment was 2.1 mm/day. A 5% subsample of collected leachate was filtered through a Minisart 0.22 μm cellulose nitrate filter (Sartorius UK Ltd), and refrigerated at 4°C prior to chemical analyses.

| Leachate, soil, plant and rock analysis
Leachate and soil pH were measured with a Jenway 3540 Bench Combined Conductivity/pH Meter (Jenway). Measurements were standardized using a 3 point calibration on the hydrogen ion molar activity scale (pH = −log{H + }) using standard solutions at pH 4.00, 7.00 and 14.00. For soil pH measurements, 15 g of air-dried soil was KELLAND Et AL.

| 3661
weighed into a 100 ml glass beaker and gradually mixed with 30 ml of deionized water (1:2 mass ratio of soil:water) using a glass stirrer, for 3 min. The solution was left to equilibrate with the atmosphere for 30 min before recording the measured pH of the slurry to 0.01 pH units. Dissolved cation concentrations in the leachate were quantified by Inductively Coupled Plasma Atomic Emission Spectrometry (ICP-AES; Spectro-Ciros-Vision, Spectro Analytical Instruments GmbH; detection limit for Mg/Ca/Sr = 0.001 mg/L; calibrated with multi-element standards from certified stock solutions). Soil exchangeable cations were isolated by leaching soil with 1 M ammonium acetate (NH 4 CH 3 CO 2 ; pH 7; Chapman, 1965), and subsequent elemental analysis of the extracts was conducted by ICP-AES. Soil and basalt total inorganic carbon (TIC) and TOC measurements were made using a CN Analyzer (Vario EL Cube; Elementar) on samples before and after treatment with 6 M HCl; soil: acid ratio of 90 mg to 0.5 ml (n = 6 homogenized samples per depth at four depths, 12.5, 25, 37.5 and 50 cm).
Sorghum root, shoot and seed masses were powdered separately DNA was extracted from basalt in the mesh bags using a DNeasy PowerSoil Kit (Qiagen Ltd). Amplification of partial 5.8S, ITS2 and partial 28S fungal sequences was performed by PCR with MyTaq™ Polymerase (Bioline) and primers gITS7 (Ihrmark et al., 2012) and ITS4 (White, Bruns, Lee, & Taylor, 1990), modified to include the Illumina overhang adapter sequences (Pétriacq et al., 2017). Triplicate PCRs were carried out on 1 μl of template DNA extract in the presence of the manufacturer's buffer (including dNTPs) and 0.4 μM of each primer in 20 μl reactions (PCR conditions: 94°C for 2 min; 35 cycles at 94°C for 30 s, 53°C for 30 s and 72°C for 30 s; and 72°C for 1 min).
Products of the three PCR reactions were then pooled per sample and purified using AMPure XP beads (Beckman Coulter Inc., USA).
Dual indexes were added using the Nextera XT Index Kit (Illumina Inc. UK) following the manufacturer's instructions. Amplicons were then purified again using AMPure XP beads, quantified and pooled.
Sequencing was performed using 250 bp sequencing on a MiSeq sequencer (Illumina Inc., at The Earlham Institute, Norwich, UK).
De-multiplexed FastQ files were filtered using USEARCH 9.1 (Edgar, 2010) with a maxEE value of 1. Only forward reads were analysed due to length variability in the ITS2 region producing potential biases for paired data. Primer sequences were removed and UCHIME was used to detect chimeras (Edgar, Haas, Clemente, Quince, & Knight, 2011). Comparisons were made to the QIIME/ UNITE reference ITS database (v7.2; Nilsson et al., 2019) and sequences were clustered at 97% using QIIME (Caporaso et al., 2010).
Sequences have been deposited in the European Nucleotide Archive under accession number PRJEB28082.

| Basaltic rock characterization
Pulverized basalt was sourced from the Cascade Mountain Range,

Oregon (Central Oregon Basalt Products LLC). The rock is middle
Miocene in age and belongs to the 'Prineville Chemical Type Unit' of the Columbia River Basalt (Smith & Hayman, 1987); with a chemical index of alteration of 38%, as calculated from the oxide data (Table S1; Nesbitt & Young, 1982), it is considered relatively unaltered. The mineralogy of the Oregon basalt (Table 1) was determined using X-ray diffraction (XRD) analysis at the British Geological Survey (BGS) Keyworth laboratories. Samples were spiked by 10% (wt%) with corundum (Al 2 O 3 ) to aid quantitative analysis and spray-dried from an ethanol suspension at 80°C to ensure random orientation of mineral phases. Measurements were conducted on a PANalytical X'Pert-pro diffractometer (Malvern Panalytical Ltd.) using CoKα radiation (λ = 1.78896 Å) from 4.5-85° 2θ run at 45 kV and 40 mA conditions. Crystalline mineral phases were identified using PANalytical HighScore Plus software coupled to the International Centre for Diffraction Data (ICDD) PDF-4+database. Crystalline mineral and amorphous material quantification was completed using the Rietveld refinement technique within the same HighScore Plus software package and using crystallographic information from the Inorganic Crystal Structure Database (Hellenbrandt, 2004) following the methodology in Kemp et al. (2016).
X-ray diffraction analysis indicated that the basalt contained a high percentage of amorphous material interpreted to be glass (~25%; Table 1). The elemental composition of the basaltic glass was determined with scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDX) to better constrain our reactive transport model (RTM) calculations ( Figure S9; Table S2). Samples were carbon coated to approximately 25 nm thickness in an Agar AGB7367A automatic SEM carbon evaporation coater. Analysis was performed using a FEI QUANTA 600 SEM, operating at an TA B L E 1 Basalt mineralogy determined by X-ray diffraction and kinetic rate parameters (Palandri & Kharaka, 2004) (Table S2). Basaltic glass dissolution rates modelled using the following equation (Gislason & Oelkers, 2003): where r +,geo is the geometric surface area-normalized steady-state basaltic glass dissolution rate, A A is a constant equivalent to 10 −5.6 (mol of Si) cm −2 s −1 , E A is a pH-independent activation energy of 25.5 kJ/mol, R is the gas constant, T is the temperature in K, and a i is the activity of the respective aqueous species.
e Weight derived from carbon analysis of unweathered basalt samples using a CN Analyser, before and after treatment with 6 M HCl. Calcium carbonate in basalt was modelled as an equilibrium phase.
accelerating voltage of 20 kV, spot size 5 and under high vacuum conditions (<1 × 10 −4 Torr) also at the BGS laboratories. Energy dispersive X-ray analysis was conducted using Oxford Instruments X-MAX large area (50 mm 2 ) silicon drift detector, running with Oxford Instruments INCA (v4) software. The EDX system was used to identify the point-elemental compositions from samples and is capable of detecting elements from atomic number 4 (B) to atomic number 92 (U) with a detection limit of the order of 0.2-0.5 weight % for most elements. Areas of the amorphous glassy phase were identified from backscattered electron images and relevant spectra and concentration determined from point analyses.
Bulk pulverized basalt was split into representative samples using a spinning riffler (Microscal), and the particle size distribution was established using a laser diffraction particle size distribution an-  Figure S12; Table S3).

| Reactive transport model
Column reactors were conceptualized as a 250 mm deep soil column constituting five 50 mm deep cells, and a 50 mm deep wellmixed fluid reservoir in contact with the atmosphere at the base of the column. The RTM was constructed with the PHREEQC platform (Parkhurst & Apello, 2013) using the T&H.dat geochemical reaction database (Appelo & Postma, 2005), which includes surface complexation constants for particulate organic matter based on the WHAM model (Tipping, 1998;Tipping & Hurley, 1992).  (Table 1). (1) Reactive surface area for each mineral was apportioned according to the mineral weight per cent of the basalt grains.
Chemical dissolution rate laws and coefficients for basaltic glass were sourced from Flaathen, Gislason, and Oelkers (2010) and those for other minerals from the US Geological Survey (Palandri & Kharaka, 2004). Solubility constants of the basaltic minerals were taken from the THERMODDEM database (Blanc et al., 2012) and for the basaltic glass from Aradóttir, Sonnenthal, and Jónsson (2012).
Changes in reacting particle sizes due to element mass transfer to solution during dissolution were simulated using the shrinking particle model (Rimstidt, 2014). Background weathering of soil minerals was represented by a porewater solution chemically identical to leachate from the basalt-free systems. Figures S13 and S14 provide schematic and systems diagrams, respectively, defining the process and interactions of the RTM.
Modelled geochemical processes included equilibrium of dissolved inorganic carbon species with atmospheric CO 2 , and a profile of partial pressures of CO 2 in the soil profile representing autotrophic and heterotrophic respiration (Nan, Yue, Li, Huang, & Shen, 2016). Apparent reactive surface area (RSA) and CEC were opti- Based on the smallest root-mean-square errors between model and observed leachate Al (Figures S15 and S16) and 18 leachate and extractable elements (Figures S17 and S18), our solid equilibrium phases are determined to be amorphous Al(OH) 3 and Fe(OH) 3 , along with calcite, amorphous silica and pyrophyllite (MnO 2 ). These phases react reversibly on the relatively short residence times of the soil fluids. We optimized apparent CEC because measured CEC will include the charge separately represented by the WHAM model, and because our single extraction method could have resulted in dissolution of existing phases, such as calcite. Our optimizations suggest less than 10 cmol c /kg soil of the total observed charge is unaccounted for in our model ( Figure S19).
Cation exchange on clay surfaces was modelled using the default exchange convention in the PHREEQC code. Exchange site composition was initially determined by equilibration with the measured ionic composition of soil pore waters in the basalt-free control experiments with optimized CEC. The total number of sorption sites attributable to organic matter was derived from the concentration of extractable organic matter, as determined experimentally.
Following the protocol described in Example 19 of the PHREEQC documentation (Parkhurst & Apello, 2013), these sites were apportioned to the 20 WHAM surface species using WHAM constant charges of −1.42 meq/g humic acid for four monoprotic phenolic sites, −2.84 meq/g humic acid for all diprotic or carboxylic sites and total charge −7.1 meq/g humic acid. CO 2 capture was calculated as the sum of HCO 3 − exported in leachate plus CO 2 incorporated into pedogenic carbonate minerals minus half of the HCO 3 − originating from weathering of carbonate mineral component of the basalt (Table 1).
In the model, growth of Sorghum influences the geochemistry of the system by removing Ca, Mg and Si from the soil with concomitant release of hydrogen ions to the soil. Using an average leachate chemistry from our control columns as our influent, implicitly accounts for both dissolution of existing soil minerals and 'background' plant growth unrelated to the treatment. Sorghum growth due to the treatment is therefore the difference between growth in the individual treated column and the average growth in control columns. Sorghum growth was modelled as a linear function which removed ions at a constant rate up to the maximum plant growth at 120 days, when harvesting of the plant was simulated by switching off the growth function for the rest of the annual cycle. In the 5 year modelling scenarios, a new growth cycle was modelled, as described above, for the first 120 days of each year, followed by a fallow period until the beginning of the next year.

| Sorghum yield response
After 120 days, yield of mature Sorghum, measured as seed dry mass Over five successive harvests in that study, sugarcane yields increased by up to 30% compared with plots receiving fertilizer and no basalt addition (de Villiers, 1961).
The average 21% increase in Sorghum yield we report

| Elemental budgets of Sorghum plants
Plant elemental budgets showed significant increases in whole plant Si (p < .05), Ca and Sr pools (p < .01; Figure 2a,b,f), with overall Mg, K and P budgets remaining unchanged, but seed K increasing significantly (p < .05; Figure 2c  Error bars represent SE (n = 6 columns per treatment). The asterisk denotes that the result is statistically significant at the 5% level (i.e. p < .05) Replenishment of depleted soluble Si pools by basalt weathering as seen in the present work represents a substantial co-benefit of ERW because repeated biomass harvesting of silica accumulating crops by intensive cultivation depletes plant-available Si in soils (Haynes, 2017). Crop production and harvesting in the United States, for example, removes an estimated 19 million tonnes of silica annually and negatively affects yields, especially of sugarcane and rice (Tubana, Babu, & Datnoff, 2016). Adoption of Si-fertilization practices, such as dressing soils with natural and artificial silicates, to maintain crop yields could help redress the problem in Europe and North America (Artyszak, 2018;Crooks & Prentice, 2017), and the tropics (Ma, 2004;Ma & Yamaji, 2006;Meena et al., 2014). In east Asia, this approach has been practiced for the past decade to improve crop production (Cuong, Ullah, Datta, & Hanh, 2017;Li et al., 2018).

| Leachate and soil biogeochemistry
We detected no significant changes in leachate Si, Mg or Ca following the basalt treatment after 120 days (Figure 3b-d), suggesting elements released by weathering were retained in the plant-soil system in this study. Lack of leachate chemistry response has also been observed following a very high olivine application (220 t/ha) to loamy soils (pH = 7.8) supporting the growth of wheat and barley . We also found no statistical

| Element release from basalt weathering
Comparison of the total measured amount of Ca, Mg and Si in the exchangeable soil cations, column leachate and plant tissues (biomass [ Figure 1a] × concentration [ Figure S20]) provides a minimum estimate of mass transfer of these elements from basalt weathering into the plant-soil system (Table 2). Additional mass transfer resulting in formation of secondary minerals is not accounted for in these sampling procedures but may be inferred from geochemical modelling. Measured major cationic elements released by basaltic grain weathering were generally retained in the soil exchange pool ( and is less affected by plant uptake (Figure 2) and secondary mineral precipitation than Ca. This is consistent with significant (p < .05) Mg depletion on the grain surfaces, as determined by XRF ( Figure S3). We have not used dissolved Si as a tracer for mineral dissolution due to potential retention as amorphous silicate coatings that have not been quantified, for example, on soil and basalt particles (Daval et al., 2011). Rates for Mg release normalized to the BET surface area of our milled basalt rock (10 −13 mol m −2 s −1 ; Table 2) are a similar order of magnitude to those observed for dissolving Mg-rich olivine mineral grains in planted mesocosm experiments (10 -13.1 -10 -13.7 mol m −2 s −1 ; Amann et al., 2018;10 -11.8 -10 -12.7 mol m −2 s −1 ; Renforth et al., 2015). The mineralogy of the basalt (Table 1) suggests that if Mg were being released primarily from the olivine exposed at the surfaces of the grains, the release rate normalized to olivine surface area would give a calculated rate approximately 100-fold higher. This assumes that olivine surface area as a fraction of measured surface area scales with wt% ol-  Table 2).
The BET method is postulated to overestimate RSA of dissolving minerals due, in part, to measuring internal pore surface area  where A, B, C, D and E are the measured moles of the element in the leachate, roots, shoots, seeds and soil extraction, and the subscripts Bas and Con relate to the basalt treatment and control.
The overall uncertainty, δQ, is calculated by summing the errors for each system component in quadrature, using the following equation (Kirchner, 2001), c Calculated by dividing the measured moles released from basalt by BET-determined rock surface area (SSA = 7.35 m 2 /g basalt).
that is inaccessible to reacting pore fluids (Brantley & Mellott, 2000) leading to underestimation of surface area-normalized weathering rates. We calculated the apparent RSA using our RTM, constrained by the elemental budgets ( Table 2). The RSA value was obtained by optimization, as described above, where the RTM input files were updated with new RSA values. Model goodness-of-fit was evaluated using the root-mean-square difference between Mg release by basalt dissolution and the experimentally determined mass transfer of Mg into the plant, leachate and soil exchange pools in samples collected after 120 days (Table 2). Based on this approach, apparent RSA in our RTM was 7.35 m 2 /g, essentially the same as the BET surface area. Our RTM does not capture repeated evapotranspiration-driven wetting and drying cycles experienced in the soil columns, the formation of preferential flow paths or possible precipitation of secondary minerals occluding dissolving surfaces, as seen in other mesocosm ERW experiments Zhang et al., 2018). With constant moisture and drainage, our RTM is comparable to batch reactor experiments where particle surfaces are fully wetted and exposed continually to the fluid. Nevertheless, our modelled leachate Si is slightly underestimated compared to observations, and our Si release rates (10 -13 mol m −2 s −1 , Table 2

| Strontium isotope evidence for basalt dissolution
Radiogenic Sr isotopes help determine the relative contribution of

| Mycorrhizal fungal grain interactions
Root-associating mycorrhizal fungi are implicated in field trials  and experiments (Quirk, Andrews, Leake, Banwart, & Beerling, 2014)   harvesting the plants revealed that up to 45% of the fungal sequences were Glomeromycotina, the phylum-forming AM ( Figure S4). This substantial contribution of AM fungi to total fungal DNA likely under-represents their hyphal biomass because not all AM fungal hyphal fragments contain nuclei and they produce fewer spores than other fungal groups identified in the bags and so have a lower DNA:biomass ratio.
Under scanning electron microscopy, weathered grains retrieved from the mesh bags colonized by AM fungi (i.e. postexperiment) had rough, disrupted, surface structures and attached hyphae ( Figure S5), whereas unweathered grains (i.e. pretreatment) were characterized by smooth planar surfaces, with sharper angular geometries, suggesting crystal cleavage surfaces ( Figure S6).

| Modelled basalt dissolution and CO 2 sequestration
Reactive transport model simulations, constrained by observations after 120 days of were run forward in time for 5 years to assess min-  (Figure 5a). Simulated basalt dissolution leads to the formation of secondary minerals including calcite (calcium carbonate) and the amorphous phases, Al(OH) 3 and Fe(OH) 3 over time, which act as sinks for the weathered Ca, Al and Fe ions respectively ( Figure 5b).
Cumulative total CO 2 sequestration by coarse-grained basalt weathering is simulated to reach ~3 t CO 2 /ha within 2 years (Figure 5c), and continues to rise towards a maximum of ~4 t CO 2 /ha after 5 years. Initially, pedogenic carbonate (calcite) formation dominates CO 2 sequestration as the fast-reacting minerals (olivine and diopside) undergo dissolution releasing rapidly divalent cations.
Diopside is exhausted after approximately 2 years, and soil-formed calcite begins to dissolve, exiting the systems in the effluent as bicarbonate and reprecipitated calcium carbonate, a storage reservoir that linearly rises over the 5 years of the simulations (Figure 5c).
Simulations with a 10-fold smaller particle size distribution result in a faster dissolution of olivine and diopside leading to more rapid CO 2 sequestration via pedogenic carbonate formation (Figure 5d), but with similar cumulative sequestration after 5 years (4.2 t CO 2 /ha for the coarse-grained basalt vs. 4.4 t CO 2 /ha for the 10-fold smaller particles). Amplification of carbon sequestration resulting from the basalt treatment, relative to the 'control columns' was 3.1-and 3.2fold after 5 years for the two sets of simulations (experiment and small particles respectively). Estimated CO 2 sequestration rates via basalt weathering in our planted mesocosms are comparable with those of prior mesocosm ERW studies in which soils were amended with milled silicates (Table 3). Application rates, types of silicate, irrigation regimes and duration of experiment contribute differences in estimated CO 2 sequestration rates, but lowest rates correspond to studies in which pedogenic carbonate formation has not been estimated (Table 3).
Both sets of RTM simulations support the application of larger particle sizes for Oregon basalt for carbon removal than are typically adopted in ERW numerical studies-or example, 10-20 µm diameter (Köhler et al., 2010;Moosdorf, Renforth, & Hartmann, 2014;Strefler et al., 2018;Taylor et al., 2016). This reduces the energy demands for milling and the associated carbon emissions penalty from the use of fossil fuels (Lefebvre et al., 2019;Moosdorf et al., 2014  wollastonite-amended agricultural soil supporting maize and beans (Haque et al., 2019), and a long-term (7 year) field study (Manning et al., 2013). However, the fate of pedogenic carbonate will depend on seasonal rainfall patterns and land management. Its fate requires assessment in field scale trials under different climatic conditions and farming practices (Zamanian, Pustovoytov, & Kuzyakov, 2016).
We recognize that CO 2 may have been lost from our experimental basalt-treated columns via soil organic matter mineralization, and CO 2 release by respiration in response to increased pH (Malik et al., 2018). However, the agricultural soil in the columns was low in organic matter (~1%), in common with most agricultural soils, and the control and treated soil organic carbon concentrations matched each other after 120 days (Figure 6c). Soil respiration measurements following low (10 t/ha) and high (50 t/ha) applications of finely powdered olivine to an acidic organic-rich podzol soil (pH = 4.9) support this view, with no significant increases in cumulative CO 2 fluxes relative to unamended controls (Dietzen et al., 2018).

| CON CLUS IONS
We have shown that amending a slightly acidic clay-loam soil with a high loading of coarse-grained basaltic rock dust substantially increases Sorghum yield without P and K fertilizer usage or adverse trace element uptake into the seed, under experimental conditions.
Elemental mass budgets indicate that products of basalt dissolution (alkalinity and cations) will not be immediately transported directly to the marine environment via surface waters because of uptake of elements into plant biomass and temporary sequestration onto soil exchangeable sites (e.g. clay and organic matter). Geochemical reactive transport modelling indicates that a single application of basalt might achieve carbon sequestration rates of 2-4 t CO 2 /ha over 1-5 years. The fates of the newly formed soil carbonates and weathered elements taken up into biomass will depend on land management practices. The long-term fate of these weathering products requires assessment through field trials. Currently, primarily crop grain biomass is removed from fields and the shredded fast-decomposing stover normally returned back to the soils, whereby cations taken up by this unharvested biomass become available to participate in carbon capture.
Our study with a clay-loam soil represents a first step towards supporting the applicability of ERW to European and North American agriculture in which cereals with high silica demand are the most important crops. However, further investigations are warranted at lower rates of basaltic rock dust application to optimize cost effectiveness. If yield improvements follow under field conditions, associated economic gains could offset costs of ERW | 3673 deployment. Furthermore, volcanic rock dust is a recognized fertilizer in organic agriculture (Van Straaten, 2006) which, by avoiding superphosphate fertilizers, supports diverse active mycorrhizal fungal networks (Verbruggen et al., 2019) to facilitate mineral dissolution and subsequent CDR (Burghelea et al., 2015(Burghelea et al., , 2018Quirk et al., 2012Quirk et al., , 2014. With organic agriculture land area increasing fourfold in 10 years to occupy 57.8 million hectares (in 2018) across 178 countries, our findings highlight new opportunities for upscaling ERW practices in this rapidly expanding sector (Organics International, 2018). Furthermore, it raises the prospect that conventionally managed agricultural land suffering depletion of available silicon pools as a result of intensive cropping, with cereals and straw and stover removal for biofuels and other uses (Haynes, 2017), may also be suitable for ERW.

ACK N OWLED G EM ENTS
We

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw datasets are included in the Supporting Information tables and the PHREEQ-C code used to interpret the geochemical data and model CO2 capture is listed in Appendix S1. Additional datasets are available on request from M.E.K.