Two sets of 42 bottles (1 L wide-mouth polypropylene) each were used to conduct the experiments at room temperature (24.6±1.4 °C). About 90 L of natural Indian Ocean seawater was collected from the surface ∼ 400 km southeast of Sri Lanka (2°59.995^′ N, 77°59.997^′ E) and stored in five 18 L borosilicate 3.3 glass bottles. The bottles were stored in the dark in a refrigerated container until delivery ∼10 weeks after collection. The entire amount of seawater was sterile filtered using a Whatman Polycap 36 AS encapsulated filter with a 0.2 µm pore size. The seawater was stored in a cleaned and MilliQ-rinsed 210 L PP rainwater tank (Garantia rainwater green round barrel). The seawater was bubbled for a week using MilliQ-saturated ambient air following Moras et al. (2023) to allow for the entire content of the tank to be equilibrated with atmospheric pCO₂. The experiments were conducted in 1 L PP bottles that were cleaned in a 3-step process. They were first soaked overnight in a 1 % solution of NaOH with dishwashing tablets. The next day, the bottles were rinsed with tap water three times and soaked overnight in a 0.5 % solution of HCl. Finally, all bottles were rinsed with tap water and MilliQ (18.2 MΩ) three times each, before being dried overnight in an oven at 60 °C. Each of the 84 cleaned bottles was filled with ∼900 g of sterile-filtered seawater and the weight was accurately recorded (balance accuracy ±0.01 g).

One set was placed on a table and kept static throughout the experiments, while the second set was placed on an Edmund Bühler 30 SM C shaking table, set to circular motion at 125 RPM. The choice of motion and RPM was made following a dye test where powdered phenolphthalein dye was added to seawater in a 1 L bottle. The dye changes from a colourless powder to a bright pinkish colour when dissolved in solutions of pH higher than 8. These experiments revealed that below 80 RPM, the vertical mixing was deemed insufficient, with most of the dye sinking and remaining at the bottom of the bottle. Above 130 RPM, the seawater contained in the bottles reached the rim of the bottle and was deemed too great a risk of spilling. Therefore, a slightly lower rotation rate of 125 RPM was selected, providing suitable particle resuspension yet avoiding water overflow. Similarly, the tests revealed that a circular motion was more appropriate than a linear motion. Linear motion produced insufficient particle resuspension and substantial disruption of the water surface, including splashing. Circular motion produced turbulent conditions with effective vertical mixing without visible splashing. Turbulence can be expressed as a dimensionless number which is also used in physical oceanography, known as the Reynolds number (Reynolds, 1883). To compare turbulence in our experimental setup to naturally occurring ocean turbulence, we derived the Reynolds number from the bottle diameter, tangential velocity, seawater density and dynamic viscosity, as per Griskey (2002). Our conditions yielded a Reynolds number approximately 1.7×104, above the threshold considered for turbulent systems (Millero and Poisson, 1981; Thorpe, 2007; Sharqawy et al., 2010). This setup was used for a total of six months to assess the long-term response of alkalinity addition to seawater.

To minimize evaporation, the headspace of the open bottles was continuously replaced with water-saturated air. For this purpose, air was bubbled through MilliQ contained in a sealed 1 L Schott bottle. Once the air was bubbled through the MilliQ, it was carried over into plastic airflow dividers similar to those used for aquariums (Fig. 1). One tube was positioned in the headspace of each bottle and secured to the bottle neck with binder clips (Fig. A1 in the appendix). The air did not directly bubble the seawater, but rather continuously renewed each bottle headspace with MilliQ-saturated ambient air. To compensate for evaporation, each bottle was regularly weighed, and accurate MilliQ additions were performed when a decrease in weight of more than 5 % was observed. To reduce the potential growth of autotrophic organisms, two germicidal UV lamps emitting at 254 nm were mounted on top of the setup and placed so that all bottles were exposed to the UV lights. These lamps were automatically switched on for 1 h twice a week in the middle of the night, allowing for a long enough exposure to limit microbial growth.

Each of the two 42-bottle experimental sets consisted of three control bottles filled only with sterile-filtered seawater, 18 TA-enhanced bottles with Ca(OH)₂ powder, 18 TA-enhanced bottles with Mg(OH)₂ powder, and three TA-enhanced bottles with a stock solution of NaOH. Every treatment was conducted in triplicate. Targeted alkalinity increases ranged from +100 to +600 µmolkg-1 in increments of 100 µmolkg-1 for both alkaline powders, and the NaOH-amended set consisted of a TA increase of +600 µmolkg-1. One bottle was lost from the 125 RPM NaOH treatment after day 6, and one was lost from the 0 RPM NaOH treatment after day 108. Subsequent values therefore represent the means of the two remaining replicates. The alkaline powders Ca(OH)₂ and Mg(OH)₂ and the NaOH pellets used to make the 2 M NaOH solution were all reagent-grade, obtained from Carl Roth (Germany). The Ca(OH)₂ and Mg(OH)₂ powders were analysed for grain size distribution with a Bettersizer S3 Plus, exhibiting similar D50 values, i.e., 31.98 and 29.79 µm, respectively, as well as comparable specific surface areas, i.e., 58.90 and 62.82 m² kg⁻¹. Both powders were dried overnight at 60 °C, and specific amounts for TA amendments were weighed on weighing paper using a Kern ABT 120-5DNM (±0.01 mg) balance. Ca(OH)₂ additions ranged from 3.67 to 22.26 mg L⁻¹, while Mg(OH)₂ additions ranged from 2.88 to 17.54 mg L⁻¹ to reach the desired ΔTA targets (Table A1 in the appendix). The NaOH solution was added using an Eppendorf Research Plus 1000 pipette. Overall, the weighing uncertainty led to an offset between the targeted and estimated TA of less than 1 %.

For each experimental set, Ca(OH)₂ and Mg(OH)₂ treatments were applied at six amendment levels (+100, +200, +300, +400, +500, and +600 µmolkg-1), while NaOH was applied only at +600 µmolkg-1. Each treatment and the unamended control were prepared in triplicate, resulting in 42 bottles per experimental set. The pH, temperature and conductivity of each sample were measured before manipulation and 1, 3 and 6 d after alkalinity addition. Thereafter, measurements took place twice a week for the first two months and once a week afterwards. pH and conductivity were measured using a multiparameter pocket measuring device MultiLine Multi 3630 IDS from WTW, with a SenTix 940 pH probe and a WTW TetraCon 925 conductivity probe. The pH probe was calibrated weekly using NIST-traceable buffers and measured values were converted to the total pH scale (pH_T) using measured salinity and temperature and inputting the values into the equation derived by Badocco et al. (2021). The conductivity probe was calibrated before the first measurement to determine the starting salinity of the seawater. Salinity was calculated as 35.346±0.077 on the practical salinity scale of 1978 (Lewis and Perkin, 1981).

TA and DIC samples were taken before manipulation to establish background concentrations. Then, sampling occurred on days 1, 3 and 6 after alkalinity addition, and monthly afterwards. Seawater was sampled using a peristaltic pump connected to a 0.2 µm filter from Sartorius. Borosilicate-glass 3.3 TOC vials (24 mL, Shimadzu) were filled slowly from the bottom and allowed to overflow by at least half of their volume (Dickson et al., 2007). The vials were then sealed without headspace using screw caps fitted with red silicone/white PTFE double-layer septa. TA samples were collected immediately after DIC samples, with ∼30 mL of water sampled in 100 mL PP bottles. The TA bottles were closed tightly to prevent any evaporation, and both TA and DIC samples were stored refrigerated in the dark until analysis (within one week of sampling). Up to 100 mL was removed at each sampling point, leaving less than 150 mL at the end of the experiment, as the sampled water was not replaced.

The remaining water was used for mineral microscopy and elemental composition analysis. For microscopy, about 20 mL of the remaining water was filtered through Whatman polycarbonate Nuclepore filters (0.2 µm) from bottles amended to a target ΔTA of +600 µmolkg-1. The filters were briefly rinsed for several seconds with a few millilitres of fresh MilliQ to remove residual salts before being dried at 60 °C overnight. Another set of filters was used for elemental analysis. ∼50 mL of water was accurately weighed and filtered on a Whatman polycarbonate Nuclepore filter (0.2 µm). The recorded weight was used to determine the exact amount of seawater filtered for the analysis. The filter was then placed into a 50 mL Falcon tube to dry overnight. The incubation bottles were rinsed three times with ∼15 mL of 3 % Suprapure nitric acid to remove precipitates from the bottle walls. The acid was then transferred into the Falcon tube with the filter to dissolve all collected particles before being made up to a final volume of 50 mL.

TA was determined by potentiometric titration using an 888 Titrando connected to an 814 USB Sample Processor (Metrohm) with a 0.02 M HCl solution as the titrant. The titrant ionic strength was adjusted to 0.72 mol kg⁻¹ by NaCl addition, corresponding to a salinity of 35 (Dickson et al., 2007). The raw TA values were corrected against certified reference materials (CRM #187) from Dickson (2010). The overall instrument uncertainty was estimated at 0.1 % via error propagation of all CRM measurements. DIC analyses were conducted on a Shimadzu TOC-L/CPH analyser calibrated using a CRM (#207). Samples were injected into the reactor filled with fresh 25 % H₃PO₄, and the evolved CO₂ was carried by CO₂-free air. The overall instrument uncertainty was also determined via error propagation of all CRM measurements and estimated at 3.5 %.

During the six-month experiment, atmospheric CO₂ concentration, temperature and humidity were recorded every 15 min with a HOBO MX1102A CO₂ logger to control for any spike in atmospheric CO₂ and/or temperature. The pCO₂ logger was automatically calibrated to 400 ppm (as per manufacturer recommendations) by placing the logger in open air until a stable reading was recorded. This calibration did not permit accurate pCO₂ measurement but allowed to monitor fluctuations in pCO₂ during the experiments. Overall, atmospheric pCO₂ (437.7±34.6 ppm) varied by less than 10 % throughout the experiment, and the mean temperature was recorded at 24.6±1.4 °C (later rounded to 25 °C for further processing). To compensate for such atmospheric pCO₂ measurement uncertainty, the estimated pCO₂ from the control batches was therefore used as the background concentration. Sample pCO₂ and ΩA were calculated in CO2SYS using measured TA, DIC, salinity and temperature, with the K1 and K2 dissociation constants from Schockman and Byrne (2021), KSO₄ dissociation constant from Dickson (1990), KF dissociation constant from Perez and Fraga (1987) and the borate-to-salinity ratio from Lee et al. (2010).

For microscopic analysis, the dried Whatman filters were mounted on an aluminium holder using double-sided carbon sticky tabs. Scanning electron microscopy (SEM) analyses were conducted on a Hitachi TM4000 II tabletop SEM equipped with a Bruker QUANTAX energy-dispersive X-ray spectroscopy system (EDS). The second set of filters stored in Suprapure nitric acid was used to determine the concentrations of solid Ca and Mg, to assess whether CaCO₃ precipitated on undissolved Mg(OH)₂ particles. Analyses were performed using flame atomic absorption spectroscopy (F-AAS) on a Thermo Fisher Solaar S Series.

Changes in DIC attributed to CO₂ ingassing were used to derive the gas-transfer velocities and equivalent open-ocean wind speeds in each of the experiments. The DIC fluxes were estimated from the net changes in DIC between the control and the amended bottles. The change in DIC per day was estimated and inserted into the CO₂ flux equation from Wanninkhof (2014): 1F=kK0(pCO2w-pCO2a) where F is the flux of CO₂ between the atmosphere and the seawater, K0 is the solubility of CO₂ in seawater (obtained from Weiss, 1974, at 24.6 °C and salinity of 35.346), and pCO_2w and pCO_2a are the partial pressures of the seawater and air in atm, respectively. This equation, combined with the water depth on the sampling day, was solved to calculate the gas velocity terms k0RPM and k125RPM in cm h⁻¹, for the 0 and 125 RPM experiments, respectively. Finally, these gas-transfer velocities can be correlated to open-ocean wind speed using the gas exchange wind speed relationship from Wanninkhof (2014): 2k=0.251U2Sc660-0.5 where U2 is the squared wind speed in m s⁻¹, and Sc is the Schmidt number at the measured temperature and salinity (Wanninkhof, 2014).

Throughout the experiment, laboratory conditions occasionally affected the performance of the DIC analyser. Therefore, measured DIC concentrations were routinely compared with DIC concentrations calculated independently from measured TA and pH_T measurements using CO2SYS. Because no changes in TA or pH_T accompanied randomly observed shifts in measured DIC, these deviations were attributed to analytical drift rather than changes in carbonate chemistry. For sampling campaigns showing systematic offset between measured and calculated DIC, measured DIC values were corrected using the mean ratio of all calculated to measured DIC values.

Changes in pCO₂ over time were analysed separately for the 0 and 125 RPM treatments using linear mixed-effects models implemented in MATLAB (MathWorks). The fixed effects included treatment, i.e., ΔTA targets, sampling time, and their interaction, while bottle identity was set as a random intercept to account for repeated measurements from the same experimental unit. Sampling time was treated as a categorical factor to avoid assuming a linear temporal response and to enable comparison among individual sampling dates. The control treatment and day 0 were specified as the reference levels for treatment and sampling time, respectively. Overall effects of treatment, sampling time, and their interaction were assessed using F tests with Satterthwaite's approximation for the denominator degrees of freedom. Upon significant interaction between treatment and sampling time, planned one-sided post hoc contrasts were performed to compare each treatment with the control at each sampling time, testing the hypothesis that treatment pCO₂ was lower than that of the corresponding control. To account for multiple comparisons between treatments and the control at each sampling time, the one-sided p values were adjusted using the Holm procedure. Treatment pCO₂ was considered significantly lower than the control when the estimated difference between treatment and control was negative and the corresponding one-sided adjusted p value was <0.05. Model assumptions were evaluated by visual inspection of residual-versus-fitted plots and normal probability plots of the residuals.

A TA generation efficiency metric was calculated by dividing the measured ΔTA after 24 h by the targeted ΔTA. This %TA_realised can be used to compare the dissolution efficiency of the minerals based on the treatment. Furthermore, a TA budget was derived from the carbonate chemistry measurements and mineral analyses at the end of the experiment. Measured TA loss (Meas. TA_loss) was calculated as the difference between the TA value at day 0 and the lowest measured TA value. Similarly, a theoretical maximum ΔTA (Max. ΔTA_theo) term was calculated using the mineral molar mass and the amount of material added in mg L⁻¹ (assuming complete dissolution). From both estimates, a maximum TA_loss term (Max. TA_loss) was calculated from the difference between the Max. ΔTA_theo and the lowest recorded ΔTA. An estimate of particulate Ca concentration was derived from Max. TA_loss with each mole of TA loss corresponding to 0.5 mol CaCO₃.

Alkalinity additions for each experiment were performed identically with the two solid feedstocks, with ΔTA targets varying from +100 to +600 µmolkg-1. In all experiments, the expected maximum ΔTA was not reached, but positive ΔTA values were observed at day 1 (Fig. 2). Stark differences emerged in realised TA based on the material used and the advection regime within the first 24 h. In the advected bottles, TA generation was higher, with %TA_realised values that were 2.5±1.0 and 4.3±2.1 times those in the static treatments for Ca(OH)₂ and Mg(OH)₂, respectively. For example, in the advected bottles with a ΔTA target of +100 µmolkg-1, TA increased by 55–58 µmolkg-1 with Ca(OH)₂ and Mg(OH)₂, respectively, corresponding to %TA_realised values of 55 %–58 %. However, TA generation in static bottles with a ΔTA target of +600 µmolkg-1 only yielded TA increases of 20–45 µmolkg-1 with Ca(OH)₂ and Mg(OH)₂, respectively, corresponding to %TA_realised values ranging from 3 % to 8 % (Figs. 2 and 3).

In all cases, the %TA_realised decreased with increasing ΔTA targets. Using Mg(OH)₂, the %TA_realised consistently decreased regardless of the advection speed, while using Ca(OH)₂, %TA_realised quickly decreased with advection and led to converging values for the 0 and 125 RPM regimes at ΔTA targets of 500 and 600 µmolkg-1. TA generation from Mg(OH)₂ was generally lower than that from Ca(OH)₂, with values varying between 3 %–18 % with Mg(OH)₂, and from 3 % to 30 % with Ca(OH)₂ relative to the expected TA increase. However, when exposed to advection, the TA generation of Mg(OH)₂ was mostly higher than when using Ca(OH)₂, and the decrease in efficiency was less pronounced than for Ca(OH)₂. Advection led to a %TA_realised varying from 10 % to 55 % using Ca(OH)₂, and from 29 % to 58 % using Mg(OH)₂.

In most cases, TA measured at 24 h was higher than measurements obtained during the remainder of the six-month experiment (Fig. 3). Two clear patterns emerged between the 0 and 125 RPM incubations. The absence of advection mostly led to a negative ΔTA over time. With Ca(OH)₂, clear TA losses followed the 500 and 600 µmolkg-1 additions, for which, after approximately two months, about 100±12 and 190±37 µmolkg-1 were lost, respectively. However, smaller additions led to slight changes in ΔTA but without a clear gain or loss over time. Using Mg(OH)₂ led to a clear decrease in alkalinity over time, with alkalinity loss increasing with the amount of material added. After a month, the two largest additions led to a loss in TA of 150–180 µmolkg-1. The two medium additions led to a loss in TA of about 50 µmolkg-1, while lower additions led to losses around 10–20 µmolkg-1.

However, once exposed to advection, ΔTA remained positive in nearly all cases, except for the 600 µmolkg-1 addition with Ca(OH)₂. When targeting ΔTA of 100, 200 and 300 µmolkg-1 with Ca(OH)₂, ΔTA increased by ∼50, ∼ 80, and ∼130 µmolkg-1 respectively, and remained relatively stable over time. The 400 and 500 µmolkg-1 additions did not generate more TA than the lower additions. Instead, their ΔTA values decreased over time, from ∼80 to 40 µmolkg-1 for the 400 µmolkg-1 target, and from ∼60 to about -5 µmolkg-1 for the 500 µmolkg-1. Finally, the 600 µmolkg-1 treatment led to the lowest ΔTA increase of about 50 µmolkg-1 before decreasing steadily over six months down to about -140 µmolkg-1. Conversely, the Mg(OH)₂ treatments led to increasing ΔTA with increasing additions, and positive values over the six-month experiment. The +100, +200 and +300 µmolkg-1 treatments saw an increase over time in ΔTA despite a rather stable phase over the first week. The +400 and +500 µmolkg-1 treatments did not lead to a secondary increase in alkalinity but remained stable over the six-month experiment, following a peak in ΔTA on day 1 and slight loss until day 3. Finally, the highest Mg(OH)₂ treatment saw the greatest decrease in ΔTA over time, from nearly 180 µmolkg-1 on day 1 to 99±7 µmolkg-1 after two months, after which it remained stable until the end of the experiment.

Following alkalinity additions, clear drops in the seawater pCO₂ were observed in all experiments (Fig. 4). In most cases, the decrease in pCO₂ was correlated with the target ΔTA, i.e., higher targeted ΔTA led to lower pCO₂. This was the case in most treatments except for the 0 RPM Ca(OH)₂ experiment, where the lowest pCO₂ was observed in the +400 µmolkg-1 experiment. In the 0 RPM experiments, the lowest pCO₂ recorded was 250–280 µatm in the Ca(OH)₂ and Mg(OH)₂ experiments. With both Ca(OH)₂ and Mg(OH)₂, pCO₂ slowly increased over one month but remained statistically significantly lower than the control pCO₂ values until day 3 on average in the Ca(OH)₂ treatments, and until day 6 on average in the Mg(OH)₂ and NaOH treatments (Fig. 7). Between days 6 and 35, the difference was no longer statistically significant (95 % confidence, Table A2). Over the following four months, pCO₂ slightly varied, but followed fluctuations in the control bottles. In the 125 RPM experiment, the sharp drop in pCO₂ after alkaline material addition was followed by a sharp increase over the next days. In the case of Ca(OH)₂, the pCO₂ values were no longer significantly lower than those of the control incubations after only 72 h. The Mg(OH)₂ incubations' pCO₂ also quickly increased, becoming no longer significantly lower than the control pCO₂ on day 3 for additions of 100 and 200 µmolkg-1, and on day 6 for additions of 300–600 µmolkg-1. In both cases, the pCO₂ remained relatively stable over the following six months, with variations following the same patterns as for the control incubations.

Throughout the experiment, DIC concentrations were measured directly and calculated from TA and pH_T measurements, as described in the “Material and methods” section. The ratio of calculated to measured DIC was determined for each sampling time point and is presented as individual boxplots, together with the combined dataset (Fig. 5). The boxplot analysis showed that the month 1 and 2 datasets had lower DIC ratios than the combined dataset, with mean ratios of 0.896 and 0.922, respectively, both below the lower quartile of the combined dataset (0.928). In contrast, the month 6 dataset exhibited higher DIC ratios, with a mean of 0.969, exceeding the upper quartile of the combined dataset (0.957). In accordance with the criterion described in the “Material and methods” section, DIC concentrations for the month 1, 2, and 6 datasets were corrected using the average ratio derived from all measurements. Measured DIC values in the month 1 and 2 datasets were higher by ∼101±5 µmolkg-1 and ∼47±3 µmolkg-1 respectively, while the month 6 values were lower by ∼67±4 µmolkg-1.

In almost all treatments, a decrease in DIC within the first 24 h was observed, followed by an increase in some cases. A decrease between days 1 and 3 was observed in the 0 RPM treatments and was particularly visible in the Ca(OH)₂ treatment with the highest target addition, i.e., ΔTA target of 600 µmolkg-1, and in all Mg(OH)₂ treatments with ΔTA targets higher than 200 µmolkg-1. Between days 1 and 3, rates of DIC change were -4.7±21.2 and 38.2±25.3 µmolkg-1d-1 for the 0 and 125 RPM experiments, respectively. Using these rates of DIC change, Eq. (1) was solved, taking into account that the water depth in the bottle was ∼10.5 cm at the time. For an average temperature of 24.6 °C and salinity of 35.4, K0 is approximated to 2.932×10-2 mol L⁻¹ atm⁻¹ as per Weiss (1974). The differences between pCO_2w and pCO_2a were derived from Fig. 6 and averaged to 221.6±71.2 and 256.5±96.8 µatm for the 0 and 125 RPM treatments, respectively. Solving Eq. (1) for a bottle with an internal diameter of 9.6 cm yielded a k0RPM and k125RPM values of 0.4±0.7 and 2.5±1.0 cm h⁻¹, respectively. Solving the second equation with a Schmidt number Sc of ∼533.2 at 24.6 °C and salinity 35.4 (Wanninkhof, 2014) yielded corresponding U wind speeds of 1.2±1.4 and 3.0±2.5 m s⁻¹.

NaOH additions were performed identically in static and advected bottles. However, a clear difference in ΔTA emerged between the two advection regimes. At 0 RPM, ΔTA only reached about 278 µmolkg-1 within 24 h and peaked at 337 µmolkg-1 after 3 d, corresponding to about 56 % of the expected TA (Fig. 7). The 125 RPM experiments quickly reached about 552 µmolkg-1 after 1 d, before reaching up to 598 µmolkg-1 after three months, corresponding to about 100 % of the expected TA increase (within measurement uncertainty). DIC followed a similar pattern in both cases, but the final ΔDIC was approximately half as large in the 0 RPM treatment as in the 125 RPM treatment. Also, while ΔDIC was positive throughout the experiment at 125 RPM, the 0 RPM experiment first showed a decline on day 1 of about 17 µmolkg-1 before consistently increasing until the end of the experiment. pCO₂ decreased after NaOH addition down to similar levels for both 0 and 125 RPM experiments, i.e., 209 and 167 µatm, respectively. The increase that followed was quicker in the 125 RPM experiments, where, after only 6 d, pCO₂ was no longer significantly lower than the control, while it took at least one month for pCO₂ to be similar to the 0 RPM control (Table A2).

Particulate analysis was conducted using the precipitates from the highest Mg(OH)₂ addition, i.e., +600 µmolkg-1, at both 0 and 125 RPM, to provide sufficient material to determine whether CaCO₃ precipitated onto Mg(OH)₂ particles. The Max. TA_loss values for the 0 RPM treatment were higher than those for the 125 RPM by about 194 µmolkg-1 (Table 1). The results obtained from F-AAS revealed higher Ca²⁺ and Mg²⁺ concentrations in the 0 RPM treatments, with values about 7.2 and 0.4 mg L⁻¹ higher, respectively. On average, 62.7±4.4 % and 58.2±5.2 % of the estimated particulate Ca was accounted for by the measured Ca²⁺, in the 0 and 125 RPM treatments, respectively. Finally, another major difference emerged from the measurements of Mg²⁺, where the 0 RPM treatment yielded nearly twice as much particulate Mg as in the 125 RPM treatment. Overall, the particles collected from the 0 RPM experiments contained 3.8±0.4 % Mg²⁺, while the particles collected from the 125 RPM experiments were about 3.0±0.3 % Mg²⁺.

From the same incubations, as well as from the highest Ca(OH)₂ addition experiments at 0 and 125 RPM, samples for SEM analysis were taken and analysed. Precipitate analyses revealed clear and strong signals for calcium, oxygen and carbon in all samples, with traces of magnesium, strontium, sodium and chloride, i.e., <1 % on average compared to the above-mentioned elements. While no clear difference in the composition of precipitates was observed, the morphology was clearly different. Precipitates from the Ca(OH)₂ experiments were all needle-like, as shown in Fig. 8. A major difference emerged between these precipitates and those from the Mg(OH)₂ experiments. In the Mg(OH)₂ treatments, the morphology of the crystals was clearly different, with crystals appearing denser and displaying a broccoli-like structure. Although individual crystals were more difficult to distinguish, these precipitates were generally larger than those observed in the Ca(OH)₂ experiments (Fig. 8).

The results indicate that hydrodynamics influence several physicochemical processes in parallel, affecting mineral dissolution, carbonate chemistry, secondary precipitation and realised CDR. A first clear outcome is that the alkalinity generated through Ca(OH)₂ and Mg(OH)₂ dissolution using a shaking table was substantially lower than reported in the literature (Hartmann et al., 2023; Moras et al., 2022, 2024; Varliero et al., 2024; Yang et al., 2023). Such an outcome was expected as the weaker turbulence generated by the shaking table compared with that produced by a stir bar, for example, could have decreased fluid exchange around the dissolving particles and enhanced their settling. Consequently, the dissolution was likely limited by mass transfer through the diffusive boundary layer surrounding particles (Lasaga, 2014; Levich, 1962; Saez Moreno et al., 2026; Sulpis et al., 2019). The reduced advection could have increased the thickness of the diffusive boundary layer surrounding particles, reducing the diffusive flux of dissolved products away from the particles surface. This in turn could have created micro-environments enriched in Ca²⁺, Mg²⁺ and OH⁻, increasing local saturation state and ultimately decreased dissolution rates. In our case, the low (125 RPM, Reynolds number of ∼1.7×104) or absent (0 RPM) turbulence coincided with a decrease in TA generation efficiency and suggest mass-transfer limitations (Drever, 1997; Saez Moreno et al., 2026). At 0 RPM, TA generation was consistently lower in static conditions compared to the advected conditions. At the same time, the persistence of such concentrated micro-environments increased the risk of heterogeneous and pseudo-homogeneous CaCO₃ precipitation (Marion et al., 2009). The competing processes of reduced dissolution and increased CaCO₃ precipitation likely explain the substantially lower measured alkalinity increase, at most half of reported values in literature, especially in the 0 RPM incubations. Although we did not directly measure such chemical gradients and micro-environments, the dissolution and precipitation patterns are consistent with the behaviour expected from thicker diffusive boundary layers surrounding dissolving particles (Lasaga, 2014; Levich, 1962). Taken together, these findings and previous studies using hydroxide and olivine-rich feedstocks suggest that hydrodynamic conditions are a first-order control on realised alkalinity generation (Flipkens et al., 2023; Fuhr et al., 2025; Meysman and Montserrat, 2017). Expected alkalinity generation should therefore not be simply inferred from the alkaline material properties but rather a combined result of material properties and surrounding hydrodynamic regime. Environments with strong advection and/or physical disturbance, such as strong porewater exchange or water flow generating high bed shear stress, may therefore be more suitable, as previously suggested for olivine-rich materials (Flipkens et al., 2023; Meysman and Montserrat, 2017).

A similar influence of advection was also observed in the NaOH treatments, in which the two advection regimes produced noticeably different TA increases. Under advected conditions, TA reached about 92 % of the expected TA within 1 d before consistently increasing to reach 100 % on day 108 (Fig. 7). The initially reduced TA generation was not accompanied by a decrease in DIC, suggesting that CaCO₃ formation was unlikely to be the source of the missing TA. Similar patterns were observed in the literature and attributed to Mg(OH)₂ formation following NaOH addition, both experimentally at pH above ∼10 (Haas, 1916; Kapp, 1928) and in the context of OAE (Hashim et al., 2025; Shaw et al., 2025; Suitner et al., 2024; Varliero et al., 2024). The progressive increase in TA to 100 % after three months is consistent with gradual dissolution of temporarily precipitated Mg(OH)₂ (Haas, 1916; Hashim et al., 2025; Kapp, 1928). Secondary mineral formation has recently been considered a loss in TA efficiency for OAE (Lunstrum et al., 2026). However, this transient Mg(OH)₂ formation may represent a temporary reduction in alkalinity generation rather than a permanent loss of CDR potential. A contrasting pattern was observed in the static NaOH-amended bottles. There, the maximum TA generated only reached about 55 % on day 3 before decreasing further. The concomitant decrease in TA and DIC suggests that CaCO₃ precipitation after NaOH addition became the dominant process, likely obscuring the signal of subsequent Mg(OH)₂ dissolution (Fig. 7).

Advection also directly affected CO₂ ingassing rates. In the advected bottles, pCO₂ was no longer significantly lower than the control within 3 d in most treatments, and at most within 6 d for the higher treatments using Mg(OH)₂ and NaOH. Without advection, the differences in pCO₂ between amended bottles and controls remained statistically different for at least 6 d, and up to 35 d in some instances. The faster ingassing rapidly increased DIC, which in turn rapidly reduced carbonate supersaturation (Zeebe and Wolf-Gladrow, 2001). This buffering likely decreased the risk of CaCO₃ precipitation. Together, these observations indicate that hydrodynamic conditions strongly influence dissolution kinetics, atmospheric CO₂ ingassing rates and secondary mineral formation, and therefore the alkalinity generation efficiency of OAE.

Although CaCO₃ precipitation occurred in nearly all solid-amended bottles, its pattern and temporal evolution differed between advected and static bottles. CaCO₃ formation in the context of OAE has been described as a runaway process when using such feedstocks, in which the amount of alkalinity removed from the system accelerates over time and may become higher than the amount of alkalinity added (Hashim et al., 2025; Moras et al., 2022, 2024; Suitner et al., 2026). Here, such behaviour was observed in most of the 0 RPM experiments. However, the pattern of alkalinity loss attributed to CaCO₃ precipitation in the 125 RPM treatments differed. A seemingly linear and constant loss was observed in the Ca(OH)₂ experiments with a target ΔTA of +600 µmolkg-1, whereas in the other 125 RPM treatments, the early loss attributed to CaCO₃ formation quickly stopped after a few days of incubation. To our knowledge, these patterns have not been reported in the literature for hydroxide-based OAE application before this study or the recent preprint by Saez Moreno et al. (2026).

The decrease and halt in the precipitation process coincided with CO₂ invasion. Under advected conditions, CO₂ invasion occurred more rapidly. When CO₂ dissolves in water, it reacts with a carbonate ion, CO32-, to form two bicarbonate ions, HCO3- (Zeebe and Wolf-Gladrow, 2001). This consumption of CO32- decreases ΩA, which directly influences CaCO₃ formation rates (Suitner et al., 2026). These observations suggest that rapid CO₂ ingassing may have reduced the risk of runaway CaCO₃ precipitation. Consequently, more rapid CO₂ ingassing likely shortened the CAP, reducing or, in some cases, preventing runaway precipitation. Conversely, slower CO₂ ingassing in the 0 RPM experiments was associated with a longer CAP and runaway CaCO₃ precipitation (Fig. 4). Hence, the risk of loss of TA through CaCO₃ precipitation appears to depend not only on the initial amount of alkalinity and form of alkalinity added, but also on the rate of CO₂ ingassing. These findings complement earlier studies showing that secondary CaCO₃ formation is influenced by mineral surface area, realised alkalinity, saturation state and precipitation inhibitors (Fuhr et al., 2022; Moras et al., 2022; Pan et al., 2021; Rigopoulos et al., 2018; Suitner et al., 2026). It also demonstrates that CO₂ invasion and hydrodynamic conditions regulate CAP duration and directly influence alkalinity generation.

The different rates of CO₂ ingassing resulted in a broader range of ΩA at which precipitation stopped. Previous OAE experiments with similar materials reported that precipitation ceased at ΩA ∼2 (Hashim et al., 2025; Moras et al., 2022, 2024). In our experiments, the observed final ΩA values were higher, particularly in the 125 RPM experiments, with values ranging between 2.2 and 3.6 (Fig. A2). This discrepancy most likely resulted from differences in experimental setups, particularly the fact that each bottle was open to the air. The continuous atmospheric CO₂ ingassing allowed in the system by the open-bottle design progressively reduced the CaCO₃ supersaturation. These observations suggest that OAE deployments may benefit from settings in which physical and/or air-sea gas exchange rapidly reduce local supersaturation. In natural systems, physical dilution and hydrodynamics in general are expected to reduce local supersaturation on shorter timescales than atmospheric CO₂ equilibration alone. Atmospheric CO₂ invasion should therefore be considered together with physical transport processes. Such settings could include areas with relatively rapid atmospheric CO₂ invasion or horizontal dilution, provided that alkalised water remains near the surface long enough for atmospheric equilibration (Jones et al., 2014; Lunstrum et al., 2026; Zeebe and Wolf-Gladrow, 2001). In the absence of advection, however, particle settling was likely more important, which may have reduced dissolution, increased the diffusive boundary layer around the particles and increased high saturation micro-environment at the bottom of the bottles. The smaller gradient of pCO₂ between air and water, coupled to the decreased invasion rate also likely enhanced the persistence of the CAP. Therefore, the buffering by CO₂ ingassing but most importantly the larger diffusive boundary layer resulting from the absence of advection eventually led to further CaCO₃ precipitation and lower final ΩA values, i.e., from 1.8 to 2.6.

SEM-EDS analyses were consistent with the formation of CaCO₃ and revealed differing morphologies depending on the alkaline material used (Fig. 8). All crystal morphotypes were morphologically consistent with previously reported aragonite precipitates formed in comparable seawater conditions, i.e., natural seawater at 25 °C, while EDS analyses did not reveal significant incorporation of other measured elements (Morse et al., 1997, 2007). Precipitates in Ca(OH)₂-amended seawater consisted of needle-like CaCO₃ crystals morphologically consistent with aragonite crystals described in Moras et al. (2022). However, in Mg(OH)₂-amended seawater, precipitates formed spherulitic, broccoli-shaped CaCO₃ structures, morphologically consistent with aragonite synthesised by Suitner et al. (2024). Synthesis of aragonite has been undertaken in a variety of experimental setups, with morphology influenced by factors including Mg:Ca ratios and dissolved organic matter (Chave and Suess, 1970; Morse et al., 1997; Pan et al., 2021; Pytkowicz, 1965; Rushdi et al., 1992). Because all of our experiments were conducted in parallel using the same batch of seawater, differences in the starting Mg:Ca ratio or dissolved organic matter concentrations were negligible. Instead, the differences were more likely related to the alkaline feedstock used. The dissolution of Mg(OH)₂ may have resulted in higher concentrations of Mg²⁺ in the diffusive boundary layer of particles compared to Ca(OH)₂. The role of Mg in delaying and restricting CaCO₃ formation is well-documented in the literature (Morse et al., 1997; Pan et al., 2021; Pytkowicz, 1965). Prolonged exposure to Mg-rich micro-environments has previously been associated with slower aragonite formation, favouring broccoli-shaped over needle-like crystals. Although these micro-environments were not directly quantified here, these observations are consistent with previously reported heterogeneous precipitation occurring within micro-environments with differing chemical compositions (Beck and Andreassen, 2010; Gránásy et al., 2005; Tracy et al., 1998; Zubovic et al., 2025).

The observations reported above have direct implications for estimates of the realised CDR potential of OAE, as both incomplete mineral dissolution and secondary precipitation determine realised alkalinity and therefore the total CO₂ uptake potential. Atmospheric CO₂ invasion is a major determinant of OAE efficiency, particularly its realised CDR potential, although it depends strongly on the material used. Ca(OH)₂ and Mg(OH)₂ are among the most widely investigated feedstocks for OAE (Eisaman et al., 2023; Moras et al., 2022, 2024; Shaw et al., 2025; Varliero et al., 2024; Yang et al., 2023). Thermodynamically, both minerals generate two moles of alkalinity per mole of mineral dissolved, corresponding to a theoretical CDR potential of 0.9–1.1 tCO2t-1 Ca(OH)₂ and 1.1–1.4 tCO2t-1Mg(OH)2, assuming a CO₂ uptake efficiency of 0.75–0.9 moles of CO₂ captured per mole of TA (Schulz et al., 2023). This potential decreases drastically if, for example, the CO₂ emitted during the calcination of limestone to produce Ca(OH)₂ is not captured (Caserini et al., 2021; Foteinis et al., 2022; Kheshgi, 1995). Our results also point to an additional limitation, the incomplete TA generation. Under a conservative OAE deployment scenario, e.g., ΔTA +100 µmolkg-1 (Bach et al., 2025), and moderate advection, e.g., 125 RPM, the realised TA only reached 55 %–58 % (Fig. 2). The realised CDR potential of solid-based OAE therefore depends not only on the chemistry and quantity of material added, but also on hydrodynamic conditions, the fraction of TA realised, and the amount of atmospheric CO₂ ultimately taken up.

Although these processes are more readily quantified in a controlled laboratory setting, they become increasingly difficult to distinguish under open ocean conditions. In principle, incomplete dissolution can be differentiated from CaCO₃ precipitation by measuring both TA and DIC in the system. Limited TA generation without a DIC decrease may indicate incomplete dissolution, while a concomitant decrease in TA and DIC in an approximate 2:1 ΔTA : ΔDIC ratio may indicate CaCO₃ precipitation. Once CO₂ ingassing occurs, the resulting DIC increase hides this stoichiometric relationship, making the quantification of realised TA more difficult. This represents an important challenge for monitoring, reporting and verification (MRV) efforts during field-scale OAE deployments. Current MRV protocols recognise loss terms, such as biotic calcification, additionality and secondary mineral formation, but the distinction between incomplete dissolution and CaCO₃ precipitation becomes more difficult to constrain once atmospheric CO₂ ingassing alters DIC (Bach, 2024; Isometric, 2026; Lunstrum et al., 2026). To date, the first independently verified carbon credits generated from coastal OAE deployment were issued under the Isometric protocol to the Halifax-based company Planetary (Isometric, 2026; Kauk, 2025). The protocol requires monitoring of effluent characteristics, particularly pH, to ensure compliance with regulatory thresholds that must not be exceeded. In the United States, historical Environmental Protection Agency aquatic life criteria established an upper pH guideline of 8.5 and 9.0 for saltwater and freshwater ecosystems, respectively. However, the European Union does not prescribe a single pH limit for surface waters (European Parliament and Council of the European Union, 2000; U.S. Environmental Protection Agency, 1986, 2026). Our observations support the need for such a process-based approach and identify processes that should be considered in a future iteration of MRV protocols.

As mentioned earlier, current MRV protocols account for loss of TA to secondary mineral formation through a loss-term (Isometric, 2026). Hence, reliably quantifying secondary CaCO₃ precipitation remains one of the major uncertainties for OAE efficiency and MRV. Although precipitated CaCO₃ may contribute to long-term CDR, its formation in surface waters consumes alkalinity that would otherwise remain available for atmospheric CO₂ uptake, and therefore reduces the near-term CDR potential of OAE (Lunstrum et al., 2026; Moras et al., 2022). Although secondary CaCO₃ is unlikely to redissolve rapidly in natural open-ocean surface waters, it may settle onto coastal sediments or be exported to the deeper ocean, where it may later dissolve below the carbonate compensation depth or become buried in sediments where further dissolution may occur (Dale et al., 2024; Fuhr et al., 2023, 2024, 2025; Suitner et al., 2026). This temporary loss of CDR potential could become a second source of alkalinity overtime, though it is yet to be fully addressed. Nevertheless, because secondary CaCO₃ in our experiments mostly formed heterogeneously on added particles, export of coated undissolved feedstock could reduce further alkalinity generation within the surface layer, constituting an additional near-term loss in CDR potential.

The morphology of secondary CaCO₃ precipitates may also influence such transport. Based on modelling considerations from Suitner et al. (2026), the porous needle-like structures observed in the Ca(OH)₂-amended bottles are unlikely to inhibit dissolution because seawater can seep between the needle structures and reach undissolved particles. In contrast, the slower dissolution of Mg(OH)₂ seems to favour slow-growing, denser broccoli-shaped structures. The higher Ca²⁺ and Mg²⁺ concentrations measured in the precipitates from the 0 RPM experiments support the interpretation that CaCO₃ coated Mg(OH)₂ particles (Table 1). Such coatings could simultaneously reduce dissolution efficiency while increasing particle density and thereby promoting export of undissolved Mg(OH)₂ particles from the mixed layer. Conversely, rapid particle export could significantly decrease the surface area available for further precipitation, limiting the loss of TA to CaCO₃ (Baltruschat et al., 2026; Suitner et al., 2026).

Although highly relevant to OAE applications, such physical processes are poorly represented in bottle experiments. In natural systems, vertical and horizontal transport, tidal mixing and dilution would further reduce local supersaturation and available surface area for precipitation, as may occur in point-source alkalinity release (Lee et al., 2026). While this was not tested here, these processes could shorten the CAP duration more rapidly than atmospheric CO₂ ingassing alone because CO₂ equilibration occurs over longer timescales in the open ocean and varies substantially with latitude and season (Zhou et al., 2025). Better constraining these coupled physical and geochemical processes should therefore be a priority for future OAE research. Dedicated experiments incorporating realistic hydrodynamics, atmospheric CO₂ equilibration and particle transport would improve estimates of realised CDR and provide valuable constraints for future MRV protocols.

A primary outcome of this research is the close coupling between the CAP duration and multiple geochemical processes. Hydrodynamics influenced mineral dissolution and CO₂ invasion, while CaCO₃ precipitation and CO₂ invasion affected carbonate chemistry and determined the duration of the CAP. During the first few days after alkalinity addition, DIC measurements yielded estimated net CO₂ fluxes inferred from DIC changes of -4.7±21.2 and 38.2±25.3 µmolkg-1d-1 for the 0 and 125 RPM experiments, respectively. However, these rates coincided with a decrease in TA during that time, most likely resulting from CaCO₃ formation, as indicated by the simultaneous decrease in TA and the negative apparent flux of CO₂ in the 0 RPM treatments. This simultaneous removal of DIC likely biased the inferred CO₂ fluxes. Estimated DIC loss to CaCO₃ can be inferred from TA loss, where for each mole of TA lost to CaCO₃, 0.5 mole of DIC is removed. This resulted in about 10.9±6.3 and 4.8±4.0 µmolkg-1d-1 in the 0 and 125 RPM treatments respectively, corresponding to about 1.1 and 0.5 mg CaCO₃ kg⁻¹ d⁻¹. Correcting for this loss substantially increased the estimated CO₂ fluxes, especially in the 0 RPM treatments, yielding 5.3±15.5 and 43.0±25.1 µmolkg-1d-1 for the 0 and 125 RPM experiments, respectively. This supports the need to combine carbonate chemistry measurements with direct CO₂ flux estimates, because in field applications, such uncertainty could lead to underestimation of the realised CDR.

The corrected fluxes correspond to gas-transfer velocities k0RPM and k125RPM of 0.8±0.6 and 2.8±0.9 cm h⁻¹, respectively. These values are lower than typical mean open-ocean gas-transfer velocities, commonly averaging between 14 and 20 cm h⁻¹, though records range from nearly 0 to roughly 50 cm h⁻¹. Although derived from simplified laboratory experiments, these gas-transfer velocities provide a useful first-order approximation for comparison with empirical air-sea gas exchange parametrisations. The derived wind speeds should therefore be interpreted as gas-exchange equivalents rather than direct hydrodynamic analogues of open-ocean conditions (Frew et al., 2007; Yu et al., 2013; Zhou et al., 2023). Using the parametrisation of Wanninkhof (2014), corrected equivalent wind speeds were estimated to be 1.7±1.0 and 3.1±0.7 m s⁻¹, respectively. Wind speeds in the open ocean vary greatly, from less than 1 to over 20 m s⁻¹ during heavy wind and storm periods, with a mean of approximately 6 m s⁻¹ (Ho et al., 2006; Miller et al., 2010; Risien and Chelton, 2008). Hence, the data presented here provide a first-order approximation to gas transfer velocities associated with low to moderate wind scenarios and may provide useful constraints for modelling low-wind or weak gas-exchange conditions. Stronger winds may enhance wave-driven turbulence, in turn enhancing mineral dissolution and air-sea CO₂ exchange. However, their overall influence on the CAP duration will also mostly depend on mixed-layer deepening, dilution, and vertical transport, which will determine the residence time of alkalised water near the surface (Cushman-Roisin and Beckers, 2011; England and Bach, 2025; Flipkens et al., 2023; Tokinaga and Xie, 2011).

Although the present study did not focus on biological responses to OAE perturbations, the observed carbonate chemistry changes may provide a useful framework to interpret previously published OAE-focused laboratory and mesocosm research. Biological responses to OAE are generally expected to arise primarily from associated changes in carbonate chemistry, including increased pH and reduced dissolved CO₂ concentration, rather than from TA increase in isolation (Bach et al., 2019). Several studies have explored the impacts of OAE on a wide range of organisms, from single-species laboratory incubations to large-scale mesocosm campaigns encompassing natural phytoplankton communities and fish larvae (Faucher et al., 2025; Ferderer et al., 2022, 2024; Goldenberg et al., 2024; Kousoulas et al., 2025; Paul et al., 2025; Ramírez et al., 2025; Schneider et al., 2026). Experiments conducted using an equilibrated approach, i.e., a TA increase accompanied by a DIC increase, resulted in low to negligible pH and pCO₂ changes (Kousoulas et al., 2025; Paul et al., 2025; Ramírez et al., 2025). Without depletion of the CO₂ concentration, biological responses were small to negligible in most cases. However, non-equilibrated approaches with materials such as Ca(OH)₂, Mg(OH)₂ and NaOH increased TA without an accompanying initial increase in DIC, resulting in longer periods of elevated pH and reduced pCO₂. Under such conditions, reduced growth of Gephyrocapsa (Emiliania) huxleyi was reported at pCO₂ values below ∼100 µatm following NaOH addition (Faucher et al., 2025). Similar thresholds were observed for other key phytoplankton groups, particularly diatoms (Ferderer et al., 2024, 2025). Optimal growth occurred at pCO₂ values above ∼120 µatm, with some observed negative effects at lower values. Together, these studies indicate that ecological responses to depend less on TA perturbation than on the magnitude and duration of high pH/low dissolved CO₂ perturbations.

Importantly though, these observations resulted from batch and mesocosm experiments with limited CO₂ ingassing. Limited CO₂ invasion likely prolonged exposure to low pCO₂ conditions for nearly 50 d in some cases (Ferderer et al., 2024; Goldenberg et al., 2024; Schneider et al., 2026). Such prolonged exposure may be representative of point-source alkalinity release, although dilution and CO₂ invasion will most likely decrease such duration. Our experiments support this assumption. In the absence of advection, reduced pCO₂ conditions persisted between one week and one month, although minimum pCO₂ values remained around 200 µatm following NaOH addition (Figs. 4 and 7). These concentrations remained close to reported optimal values for coccolithophore calcification in a large-scale mesocosm study (Schneider et al., 2026), suggesting, based on previously reported thresholds, that severe biological effects would have been unlikely under our experimental carbonate chemistry conditions. When advection was introduced, the CAP duration decreased substantially, which resulted in reduced pCO₂ conditions for only a few days. With pCO₂ returning to background values within 3 to 6 d, we hypothesise that, under higher gas-transfer velocities and corresponding equivalent wind speeds, atmospheric CO₂ invasion may decrease pH and replenish dissolved CO₂ concentrations more rapidly, substantially reducing environmental stress from OAE.

When considered together with previous (semi-)closed bottle and mesocosm experiments, our results indicate that studies conducted with limited CO₂ exchange may serve as an upper limit for quantifying biological responses to non-equilibrated OAE. In the open ocean, CO₂ ingassing and physical processes, particularly dilution and mixing, are expected to substantially decrease the CAP duration, thereby reducing the magnitude and duration of biological exposure to altered carbonate chemistry. Although this hypothesis is consistent with the observations reported here, it remains to be tested under realistic field conditions. Such testing will be important for assessing the environmental feasibility of large-scale OAE deployment. Several limitations should be considered when interpreting these findings. First, the experiments only reproduced simplified hydrodynamic conditions and did not fully capture the complexity of ocean turbulence, wave breaking, or horizontal and vertical transport. Second, repeated sampling progressively reduced the volume in each bottle, changing the headspace and potentially influencing gas exchange over time, although all treatments underwent the same sampling schedule. In addition, airflow rate and relative humidity were not measured at the individual outlets, and differences in headspace renewal among bottles therefore cannot be excluded. Nevertheless, disregarding biological processes, particle transport and large-scale dilution allowed physicochemical processes to be isolated. The results should therefore be interpreted as process-based laboratory observations that inform, rather than reproduce, field-scale OAE deployment.