With the exception of some short-term experiments conducted on Santa Catalina Island at the USC Dornsife Wrigley Marine Science Center (33°26^′40.52^′′ N, 118°28^′47.787^′′ W), all other experiments were conducted with water collected from ∼ 1 m depth in Los Angeles Harbor, at AltaSea Marine Center (33°43^′33.64^′′ N, 118°16^′11.999^′′ W). This region exchanges seawater with the Southern California Bight, which is a part of the California Current System. The microbial community structure in the Port of Los Angeles in the summer is dominated by diatoms, pico-eukaryotes, and heterotrophic bacteria. Dinoflagellates, Synechococcus, and photosynthetic nanoplankton are in less abundance. In the fall, the carbon biomass shifts to the following in order of abundance: Bacteria > Picoeukaryotes > Synechococcus > Diatoms ∼ Nanoplankton > Dinoflagellates; these proportionally decrease in the winter when Diatoms become the dominant species again (Bialonski et al., 2016; Caron et al., 2017; Connell et al., 2017). Regardless of the season, the carbon biomass is higher in the Port compared to the costal San Pedro Ocean Time series station (∼ 17 km offshore) or in the marine protected area near Santa Catalina Island (∼ 45 km offshore) (Connell et al., 2017).

Alkalinity perturbation experiments were conducted in 5 L Thermo Scientific Nalgene PETG bottles (total volume is ∼ 6 L with no headspace) suspended 1 m below the water surface off a dock at AltaSea Marine Center. Surface seawater conditions (temperature and sunlight exposure) during each experiment are recorded in Table 1. Temperature (°C) and Light Intensity (lux) were measured with HOBO™ Temperature and Light Loggers, and the lux was converted to photosynthetic photon flux density (µmol photons m⁻² s⁻¹) by multiplying lux with the standard conversion factor of sunlight (0.0185). Salinity of the control waters from all experiments were roughly 35 ± 0.5 ppt; salinity was measured with a refractometer.

In this study we induce an alkalinity perturbation that simulates two different methods of OAE at two target enrichments, (1) Base Addition: where a strong base is delivered to the surface of the ocean and is intended to increase oceanic uptake of atmospheric CO₂ by disrupting the pH balance. To simulate this, we use two common bases, NaOH (J. T. Baker, New Jersey. ACS Grade, Lot. No.: 99057), and CaO (Sigma-Aldrich, Missouri. Reagent Grade, Batch No.: MKCS6828). The other method considered here is (2) accelerated weathering of limestone (AWL) (Dong et al., 2025; Kheshgi, 1995; Rau and Caldeira, 1999) where atmospheric CO₂ and water are pumped through a limestone reactor to neutralize the CO₂. The outflow water is enriched in the byproducts of this reaction (Eq. 2) – Ca²⁺ and HCO3- ions. In this method the CO₂ is neutralized before it enters the seawater, as opposed to being neutralized in-situ during air-sea gas exchange with the base addition method. We mimic AWL with the addition of Calcium Chloride (CaCl₂, Sigma-Aldrich, Missouri. Assay ≥97 %, Batch No.: SLCL2754) and Sodium Bicarbonate (NaHCO₃, Sigma-Aldrich, Missouri. ACS Grade, Batch No.: SLCM3421) with a molar ratio of 1:2 for Ca2+:HCO3- (Eq. 3). 2CO2+H2O+CaCO3↔Ca2++2HCO3-32NaHCO3+CaCl2↔Ca2++2HCO3-+2Na++2Cl-

To achieve specific dosages/concentrations, the alkalinity needed to be fully dissolved, thus we avoided introducing any particles into the incubations. Also, aqueous alkalinity delivery is currently considered a favorable delivery method for OAE projects (Cai and Jiao, 2022; Savoie et al., 2025). To compare our results with other published regional studies (Bach et al., 2024; Ferderer et al., 2022; Kousoulas et al., 2025; Subhas et al., 2022) and evaluate impacts above current instrument detection limits (See Sect. 3.1.1), we chose target enrichments of ∼ 500 and ∼ 1000 µmol kg⁻¹. We created dissolved stock solutions of each alkalinity type (NaOH, CaO, CaCl₂, NaHCO₃) with minimal exposure to the atmosphere. A solid salt or base was precisely weighed, transferred to a pre-weighed volumetric flask, the flask was filled with Milli-Q water, and no headspace remained. After all solids were dissolved, the solution was siphoned through a 0.2 µm GF/F filter into an evacuated gas impermeable foil bag (Supelco, Pennsylvania) for storage. We took subsamples for TA and DIC from the stock foil bags during preparation and while dosing in the field (measurement details in Sect. 2.5) to back-calculate the final concentrations of the stock solutions. The target dosages, the predicted changes calculated through a dilution equation, and the actual measured changes in TA within bottle replicates (n=3) are reported in Table 2. Alkalinity was measured with an open-system Gran titration method at 25 °C (details in Lunstrum and Berelson, 2022; Subhas et al., 2015) and no precipitation was observed in the storage containers (stored at 4 °C until analysis). As the solubility of CaO in water is less than that of the other chemicals used in this study, a larger volume of a weaker CaO stock was required to make the alkalinity additions. Hence, an immediate drop in DIC (by <2 %) due to the expected dilution by the stock solution is observed in all CaO experiments.

To increase the sensitivity of our measurements for the carbon budget analysis, we added 1–2 mL of a ¹³C tracer in the form of NaH¹³CO₃ (Sigma Aldrich, Missouri. CAS No.: 87081-58-1, Assay = 99 %, Lot No.: MBBC0208V) to all controls and treatments in five experiments: 15, 22 July, 4, 18 November 2024, and 27 January 2025. The tracer solution (∼ 0.015 M) was prepared similarly to the stock solutions discussed above. The tracer changed the alkalinity by <5 µmol kg⁻¹ but increased the δ13DIC signal to 80 ‰–200 ‰ (precise start values are included in the associated dataset (Wani et al., 2026)).

There is long standing discourse on whether bottle experiments mimic natural conditions enough to be useful to answer questions of “ecosystem impact” (Daehler and Strong, 1996; Fraser and Keddy, 1997; Giesy, 1980; Ives et al., 1996; Lawton et al., 1993; Stachowski-Haberkorn et al., 2008). There are many experimental design factors that can influence the community assemblage, such as: material used for the structure, microcosm volume, filling mechanism, mixing, replication, carbonate chemistry, exclusion of larger organisms, and wall surface growth (Riebesell et al., 2010). In perturbation experiments conducted with small volume microcosms (<10 L) it is recommended to limit the timescale of study to days-weeks if the goal is to understand the impact of the perturbation on microbial or plankton assemblages (Parsons, 1982; Riebesell et al., 2010). The purpose is to limit unwanted effects such as biofouling on the walls of the enclosure, nutrient limitation, excess grazing or changes due to lack of predators – these are commonly referred to as “bottle effects”. Our aim was to compare multiple alkalinity sources; thus we opted for a replication method that incorporated sacrificial bottle timepoints over the span of four days (Iglesias-Rodríguez et al., 2023). We maintained a closed system experiment (with no headspace) to constrain the carbon budget by eliminating gas exchange as a potential variable for this study. This can be an analog for water that may temporarily lose contact with the atmosphere, e.g. when waters mix deeper into surface ocean or for times when gas exchange is minimal.

We collected seawater for each experimental treatment by pumping from 1 m below the surface from a dock at AltaSea, in the Los Angeles Harbor. The water passed through a 250 µm mesh (to remove large grazers) (Carter et al., 2005; de la Broise and Palenik, 2007) and 22 bottles were filled in a “round robin” manner to promote homogeneity between bottles filled at different times (filling occurred in <20 min). Once all bottles were full, with no headspace, half were randomly selected to receive a dosage of alkalinity (NaOH, CaO, or CaCl₂ + NaHCO₃). Bottles were opened, 50–150 mL of seawater was removed; stock water was pulled out of the stock solution foil bags into volumetric syringes and then pushed through 0.2 µm PES syringe filters into the bottles. Any remaining headspace was once again filled with seawater; bottles were sealed and then inverted five times before being tied to the dock. All treatment bottles were given a dose of alkalinity on Day 0, some were immediately sampled, others left in the water column to experience natural changes in temperature and light conditions until their sampling day. Only one treatment experiment was conducted per week; hence, we can only compare each alkalinity type to its respective control (no addition of alkalinity). Sampling occurred on days 0, 2, and 4 with three bottle replicates per time point. We also filled and added alkalinity to 2 bottles that were incubated devoid of light as well as 2 dark controls (wrapped in black tape, See Fig. 1). Results and discussion of all dark bottle experiments appear in the Appendix A.

Bottles pulled out of the ocean were transported back to the lab (on ice) on their sampling day and weighed to track water that was used for sub-sampling. Bottles were gently inverted 5 times and then opened to measure temperature (°C) with a probe and dissolved oxygen (µmol kg⁻¹) with PreSens Optical Oxygen Sensor Spots that were attached to each bottle. The oxygen measurements were corrected to the mean temperature recorded by the in situ HOBO™ Loggers. To avoid sampling from the surface layer of water in contact with the air, a ∼ 6 inch piece of Tygon^® F-4040-A attached to a 50 mL syringe via a one-way valve was inserted into bottom of the sample bottle. The first 15 mL of sample water was used to rinse the syringe, filter and needle that would be used to collect carbonate chemistry samples, ensuring no bubbles entered the sample storage containers as recommended by standard operating procedures (Dickson et al., 2007). Through subsequent sampling, 250 mL of sample water was filtered through a 25 mm 0.7 µm glass fiber filter (Whatmann GF/F) (∼ 15 min per bottle for filtration and subsampling). A portion of the filtrate was collected for 5 mL duplicates of DIC (injected into septum-top of pre-evacuated exetainers), alkalinity (15 mL Falcon tubes with no headspace), nutrients (15 mL with no headspace), dissolved organic carbon – DOC (40 mL pre-combusted amber glass septa vials), and 50 mL for archival purposes. We note that there could be some contamination in DOC samples because we used Tygon^® to pull the sample into the syringe (Dickson et al., 2007); however, this would affect the controls and treatments in the same manner. The GF/F filter was wrapped in aluminum foil and placed in a freezer (-20 °C) for later Chl a analysis (< a month) (using a Turner Trilogy Fluorometer and the EPA SOP LG405.v11 for “Chlorophyll-A Non-Acidification Method” (U.S. EPA, 2021)). One liter was filtered for future genomic analyses not discussed in this paper, on Supor 0.2 µm 47 mm PES filters, and stored at -80 °C. The remaining (4.5 L) seawater was filtered (detailed in Sect. 2.5.2) for particulate organic and inorganic carbon.

A carbon budget for each bottle was created by assessing the total concentration of carbon in the dissolved and particulate pools. Carbonate system parameters were derived from TA and DIC measurements (discussed below) using CO2Sys.v3 (Pierrot et al., 2021) (Parameters: pH Scale = Total Scale, CO₂ Constants = (Dickson and Millero, 1987), KSO₄ = (Dickson, 1990), KF = (Perez and Fraga, 1987), Boron = (Lee et al., 2010)). Ωaragonite values for treatments with added Ca²⁺ (CaO & [Ca²⁺] + [HCO3-] Experiments) were calculated by multiplying the [CO32-] output from CO2Sys and [Ca²⁺] accounting for the increased Ca²⁺. All calculations use Ksp′ for Aragonite in seawater (Ksp′ = 10^−6.19). All other Ωaragonite values derived from CO2Sys scale [Ca²⁺] to salinity.

Samples for Dissolved Inorganic Carbon (DIC) were collected in pre-evacuated 12 mL borosilicate glass Labco Exetainers^® and immediately placed at 4 °C for storage until analysis (<1 week). For analysis, samples were acidified with 2 mL of 10 % H₃PO₄ (Spectrum Chemicals, Gardena, California) and the evolved CO₂ was analyzed by a Picarro G2121-i Cavity Ring Down Spectrometer (as detailed in Subhas et al., 2015). Optical grade calcite, with a known ¹³C value (2.47 ‰) was used to create an eight-point standard curve (ppmCO₂ per mg of CaCO₃) for each run and a certified reference material (CRM-Batch 214) generated by Andrew Dickson's Lab at SIO, UCSD was used to correct all the final DIC concentrations (4 CRM replicates included per run). The total CO₂ standard curve produced with the Automate Sample Delivery System + Picarro CRDS had an R2 = 0.9999 ± 0.0001 and slope variability = 1.5 % for all runs in this dataset. Replicate CRM samples provided a measure of precision, ±20 µmol kg⁻¹. The isotopic values of replicate standards were within ±0.2 ‰.

Samples for Dissolved Organic Carbon (DOC) were collected in 40 mL pre-combusted amber glass vials, acidified to 0.1 % v/v with concentrated hydrochloric acid (ACS grade, Supelco, Massachusetts), and placed in cold storage (4 °C) until analysis. The samples were run on a Shimadzu TOC-L Analyzer (NPOC method; Injection volume = 150 µL; No. of Injections: Best 4 of 5; Max Variance of Replicates: 2.00 %). Within a run the standard curve had an R2 = 0.9998 ± 0.0002, between runs the reproducibility of the slope had 16.5 % error. Samples were initially purged with CO₂-free compressed air to remove inorganic carbon, combusted in a furnace via catalytic oxidation, and the generated CO₂ was measured by an infrared gas analysis. Standards for organic carbon were made with Oxalic Acid (dried at 70 °C for 2 h), and CRMs from the Hansell Lab (University of Miami, RSMAS) were used to correct final concentrations (DSR Batch 22 Lot # 10–22) with a replicate precision of ±10 µmol kg⁻¹.

Roughly 4.5 L of seawater was filtered on 1–2 filters (0.7 µm 47 mm Whatman GF/F) for particulate inorganic carbon (PIC) and particulate organic carbon (POC) analyses. After filtration, ∼ 10 mL of buffered DI water (NaOH added to Milli-Q, pH 9) was used to rinse any residual seawater, and then filters were placed in the oven to dry overnight (60 °C). Filters were hole punched (diameter = duplicates of 6–12.5 mm); the hole punched fractions (2 %–8 % of total filtered area) were wrapped in tin foil capsules (9 × 5 mm) for combustion in a CosTech Elemental Analyzer – which is connected to the Picarro for CO₂ and 12C/13C analysis. This analysis of Total Carbon had a reproducibility of ±1 %. The remaining fraction of the filters were rolled and placed into Labco Exetainers^® – these samples were acidified with 5 mL 10 % H₃PO₄and the evolved CO₂ measured similarly to DIC described above. Particulate Organic Carbon (POC) was calculated by subtracting the PIC (µgC L⁻¹) concentration from the Total Carbon (µgC L⁻¹) calculated for the whole filtered area.

Statistical analysis and data visualization was conducted in R and R Studio version 4.5.1. Means and standard deviations for each treatment by time point were generated using the data_summary function. Triplicate values were averaged, but data points that were found outside of three standard deviations from the mean were flagged as outliers and not included in the statistical analysis (Pukelsheim, 1994). The dataset was analyzed with two statistical approaches: (i) To test whether the effect of alkalinity addition had a statistically significant impact on PIC, POC, DOC and Chl a over time, we used a linear model with an interaction term and assessed if the p-value on the interaction term (i.e., the difference between the slopes of the control and the treatment) was statistically significant (significance threshold α = 0.05) (results summarized Table 5). (ii) Since the previous method assumes linear relationships and normally distributed data (which may apply to some but not all variables) we also used a nonparametric two-way ANOVA (Analysis of Variance) – Scheirer Ray Hare test – to assess the impact of alkalinity addition on the Control and Treatment groups (significance threshold α = 0.05) (results summarized in Table 6). The implications of these tests and any statistically significant responses are summarized in the context of ecological variability in the Results and Discussion sections.

We intended to recreate an alkalinity impact condition that resembles high dilution of AWL after a few hours; an example outflow of alkalinity from a shipping vessel fitted with an AWL reactor would be about 3000–5000 µmol kg⁻¹ (Dong et al., 2025). As the ship's wake would dilute the outflow, we chose to dilute a stock (∼ 4754 µmol kg⁻¹) by different proportion; 1:150 (40 mL of stock), 1:300 (20 mL of stock), and 1:600 (10 mL of stock) (Table 3) based on dilution calculations presented by Dong et al. (2025) (inspired by Caserini et al., 2021; Chou, 1996). These incubations that lasted 3 h were conducted on Santa Catalina Island.

As seen in Table 3, additions at what could be realistic deployment concentrations after several hours of mixing result in changes that are extremely close to replicate error and instrument measurement error. The control seawater had an alkalinity = 2223 ± 6 µmol kg⁻¹, DIC = 2015 ± 21 µmol kg⁻¹, Ωaragonite = 2.41 ± 0.02, pCO₂ = 536 ± 4 µatm, PIC = 7.1 ± 0.2 µg L⁻¹, and POC = 68.4 ± 7.7 µg L⁻¹. In the least diluted sample, the seawater alkalinity changes from 2223 ± 6 µmol kg⁻¹ to 2233 ± 5 µmol kg⁻¹ (the expected increase would be 2248 µmol kg⁻¹) but there is no statistically significant impact on the concentrations of PIC or POC (p>0.05). Since we saw no impact on PIC or POC at such small changes in alkalinity, and these additions were at the lower limit of our measurement capabilities, we focused the rest of the paper on experiments that perturb the alkalinity by larger amounts (25 %–50 %) and for longer durations (2–4 d).

All the experiments with a 25 %–50 % alkalinity amendment above the control seawater were conducted with water collected and incubated at AltaSea in LA Harbor. Table 4 summarizes the concentrations of PIC, POC, Chl a, DOC, Ωaragonite values of the control seawater on Day 0, and the change in Ωaragonite values after treatment.

The likelihood of observing mineral precipitation after OAE is controlled by several factors including supersaturation of the seawater with respect to a particular mineral, nucleation sites, temperature, time, and even interference from catalytic or inhibitor ions (Berner, 1975; Marion et al., 2009; Moras et al., 2024; Morse et al., 2007; Morse and He, 1993; Mucci and Morse, 1983; Naviaux et al., 2019; Nguyen Dang et al., 2017; Suitner et al., 2026). Here we summarize similar studies that used NaOH, NaHCO₃, and CaO to find the minimum thresholds for precipitation of CaCO₃ after OAE (detailed experimental conditions are included in Tables B1 and B2 in Appendix  B).

Several other studies that conducted similar manipulations of alkalinity additions to natural seawater (initial Ωaragonite=1–3) have reported no observed precipitation of CaCO₃ (Hartmann et al., 2023; Moras et al., 2022; Ringham et al., 2024; Suitner et al., 2024). Of these, experiments conducted with NaHCO₃ + Na₂CO₃or NaOHhave shown stable seawater alkalinity up to Ωaragonite≈15–19 for at least four days and longer (∼ 20 d) in closed systems (Hartmann et al., 2023; Moras et al., 2022; Suitner et al., 2024). In an open system that was allowed to equilibrate with atmospheric CO₂, no precipitation was observed after NaOH addition up to Ωaragonite≈27 for at least 20 d (Ringham et al., 2024). At present, we are the only study to use CaO in an aqueous form. The one study that does not see CaCO₃ precipitation but uses solid CaO and Ca(OH)₂ feedstocks to add alkalinity, only raised alkalinity by 250 µmol kg⁻¹ (raising Ωaragonite from 2 to 5), after which higher additions resulted in precipitation (Moras et al., 2022).

In these same experimental conditions discussed above, the threshold for CaCO₃ precipitation was surpassed after Ωaragonite increased, temperature increased, or the mineral induction period where alkalinity is stable had ended. In some cases, precipitation was favorable due to the presence of nucleation sites (Hartmann et al., 2023; Hashim et al., 2025; Moras et al., 2022, 2024; Ringham et al., 2024; Suitner et al., 2024; Varliero et al., 2024). These are all parameters that vary in a given oceanic environment. With NaHCO₃ + Na₂CO₃ additions or NaOH additions where no solid feedstock was used, CaCO₃ precipitated as early as 1 d when Ωaragonite>30 (Hartmann et al., 2023; Ringham et al., 2024; Suitner et al., 2024) or after 20 d when Ωaragonite exceeded 15 (Hartmann et al., 2023; Suitner et al., 2024). At warmer surface temperatures (27 °C) precipitation started at Ωaragonite≈10 (NaOH addition) within hours in a closed system (Hashim et al., 2025). In the experiments where nucleation sites were present, precipitation could be observed as low as Ωaragonite∼7 with CaO and Ca(OH)₂ (Moras et al., 2022; Varliero et al., 2024). The influence of parameters such as surface area for nucleation and oversaturation with respect to Ωaragonite on runaway CaCO₃ precipitation have been further constrained in observation driven modeling efforts (Suitner et al., 2026).

As our alkalinity additions only raised Ωaragonite to 2 (for [Ca²⁺] + [HCO3-] addition) and Ωaragonite to 7–8 (NaOH) for 4 d, which is within the range of no observed precipitation of other studies, we believe our results are consistent with existing literature. With our CaO addition we raised Ωaragonite to 8–14, and we expected CaCO₃ as our saturation values were above that of Moras et al. (2022). However, our experimental conditions did not have solid feedstock like Moras et al. (2022), which could explain the lack of precipitation. Given our results, OAE practioners in LA Harbour, perhaps analogous to other near-shore oceanic regions, could expect no CaCO₃ precipitation with aqueous NaOH, [Ca²⁺] + [HCO3-], or CaO addition at ∼ 500 µmol kg⁻¹ for at least four days. Thus far it is clear that the thresholds for heterogeneous precipitation (i.e., relevant if using mineral feedstock) versus pseudo-homogeneous precipitation (i.e., relevant if primarily using aqueous solutions for OAE in natural seawater) (Chen et al., 2005; Marion et al., 2009; Wolf et al., 2008) must be different, and may vary with different temperature, salinity, or even organic matter content (Moras et al., 2024; Suitner et al., 2026). Future studies could focus on comparing these precipitation thresholds by systematically changing parameters that influence the induction period before mineral precipitation.

The impacts to Chl a and POC from the CaO addition when compared to the unamended conditions show different outcomes. Of the two experiments conducted around ∼ 500 µmol kg⁻¹, one (27 January 2025; measured ΔTA = 612 ± 13 µmol kg⁻¹) shows that Chl a was statistically lower. This trend was not seen in Chl a in the experiment conducted at ∼ 1000 µmol kg⁻¹ (4 November 2024; measured ΔTA = 923 ± 12 µmol kg⁻¹). POC production was reduced by 100 µgC L⁻¹after 4 d from the ∼ 500 µmol kg⁻¹ CaO additions (but the trend is not statistically significant due to high initial variability). POC production is reduced by 200 µgC L⁻¹ after the ∼ 1000 µmol kg⁻¹ addition, which is statistically significant. This coincides with statistically different (lower) dissolved oxygen after a ∼ 1000 µmol kg⁻¹ CaO addition as compared to the control. The impacts of CaO addition at ∼ 500 µmol kg⁻¹ may not be statistically significant on LA Harbor organic matter cycling, but at ∼ 1000 µmol kg⁻¹ the impacts do become statistically significant.

Other studies that test the impact of ocean liming – either simulated by CaO, Ca(OH)₂, or even NaOH + CaCl₂ addition – have seen minimal effects on growth rates, grazing rate, and community composition when the resultant change has been below pH 9 (de Castro et al., 2025; Traboni et al., 2025). In our study, the starting pH_T of the seawater is around 7.68–7.85 and the CaO additions bring this pH_T to 8.57–8.91 (∼ 500–1000 µmol kg⁻¹ experiments). Several studies, that raised pH > 9 with CaO, have shown a decrease in organism quality (i.e., nutrient composition) and growth rate (Bhaumik et al., 2025) or decreased motility and even increased mortality (Camatti et al., 2024; Pedersen and Hansen, 2003). CaO addition, that reduces pCO₂ below ∼ 100 µatm, could cause issues in primary production from the CO₂ limitation. This has been shown in coccolithophore growth (Bach et al., 2011; Faucher et al., 2025; Riebesell et al., 1993; Sett et al., 2014) and for other marine phytoplankton (Hansen, 2002; Pedersen and Hansen, 2003). These results draw attention to how reorganization of community composition (with smaller phytoplankton that are better adapted to lower pCO₂ availability (Chrachri et al., 2018; Flynn et al., 2012; Wolf-Gladrow and Riebesell, 1997)) could result from alkalinity additions. This in turn could reduce the transfer efficiency of POM to the deeper ocean (Boyd and Newton, 1995). In this study we do not incorporate a characterization of size classes, but future alkalinity enhancement studies should also try to incorporate an analysis of changes in particle size class, among a suite of other ecological impacts (Marx et al., 2026).