We computed emission-driven simulations with the Alfred Wegener Institute Earth System Model (AWI-ESM-1-REcoM). AWI-ESM is based on the AWI Climate Model AWI-CM1; but includes dynamic vegetation on land . AWI-ESM-1-REcoM further includes the representation of the carbon cycle between land, ocean, and atmosphere. Ocean and sea ice are represented by the Finite Element Sea Ice-Ocean Model version 1.4 FESOM1.4;. Ocean biogeochemistry is described by the Regulated Ecosystem Model REcoM REcoM2;. The ocean and ocean biogeochemistry components are discretized on an unstructured mesh, allowing a variable grid resolution (12–147 km, mean 76 km, median 41 km). The carbonate system in REcoM is computed across the entire water column by the mocsy 2.0 routine . The atmospheric component of AWI-ESM is represented by the spectral atmospheric model ECHAM version 6.3;. Land dynamics are modelled by the land surface model JSBACH version 3.20, including dynamic vegetation and the soil carbon model Yasso .

The ocean biogeochemistry model REcoM describes the cycling of carbon, nitrogen, silicon, iron, and oxygen . In the control version (without carbonate system effects; hereafter called “NO-CSE”), the ecosystem consists of two phytoplankton groups (small phytoplankton, diatoms), two zooplankton groups (generic zooplankton, polar macrozooplankton), and two detritus groups slow- and fast-sinking;. Calcification is proportional to the gross photosynthesis rate of 2 % of the small phytoplankton group, with a fixed particulate-inorganic-to-organic carbon ratio (PIC : POC)ref=1. Calcite dissolution scales only with depth. Further, the 3D ocean carbonate system as well as the CO₂ flux between the atmosphere and ocean is computed by the mocsy 2.0 routines . A more detailed description of REcoM can be found in Appendix A.

Deviating from this control version, three major code changes in REcoM were used in this study (further on called “CSE”), as described in . Firstly, coccolithophores as a new group of explicitly calcifying phytoplankton were added to the ecosystem, replacing the fixed share of 2 % calcifiers in the small phytoplankton group. Calcification in CSE is not only a function of the gross photosynthesis rate, but also dependent on temperature and dissolved inorganic nitrogen (DIN) limitation, resulting in a variable PIC : POC ratio. Secondly, the CSE version accommodates for direct effects of OAE on calcification as well as on gross photosynthesis of all phytoplankton groups. Thirdly, the carbonate ion concentration, and not depth, determines the calcite dissolution.

To account for direct effects of alkalinity enhancement on phytoplankton, the gross photosynthesis and calcification functions in REcoM were supplemented by a CO₂ term f(CO₂) that scales between zero and three (i.e. maximal 3-fold increase in gross photosynthesis and calcification), depending on changes in the carbonate system. The term was initially developed to describe responses to ocean acidification , but the underlying carbonate system manipulations in the experiments also allow for use under OAE-relevant conditions, as realized by . As systematic assessments of OAE effects on phytoplankton growth and calcification are missing to date, we assumed that growth and calcification responses to changes in alkalinity can be described by the same function as responses to ocean acidification. Gross photosynthesis PS, which represents the increase in biomass over time without considering loss processes, is defined as 1PSi=f(T)i⋅f(PAR)i⋅f(N)i⋅fCO2i, where f(T)i, f(PAR)i, and f(N)i describe the effects of temperature, photosynthetically active radiation (PAR), and nutrient availability on the gross photosynthesis rate PS_i of the phytoplankton functional group i (more details in Sect. S2). Calcification (Calc) of coccolithophores (denoted by j) is defined as 2Calcj=PSj⋅Cj⋅(PIC:POC)ref⋅f(T)j,calc⋅f(N)j,calc⋅fCO2calc, where Cj is the biomass of coccolithophores and (PIC : POC)_ref a reference PIC : POC ratio of one. The temperature and DIN dependencies of calcification, f(T)j,calc and f(N)j,calc, follow . Calcification decreases linearly at temperatures below 10.6 °C. The dependence on DIN limitation modified from the original phosphate limitation; is described by a modified Michaelis–Menten equation. Both terms are explained in more detail in Appendix B and C. The CO₂ factor f(CO2) in PS_i and Calc_j (Eqs.  and  refeq:calc) is defined as 3fCO2iorcalc=ai⋅HCO3-bi+HCO3--exp⁡-ci⋅CO2(aq)-di⋅10-pH. The parameters a, b, c, and d (Table ) were derived from curve fitting to experimental phytoplankton growth data , and [HCO3-], [CO_2(aq)], and pH are the concentrations of bicarbonate, dissolved CO₂, and pH in the surrounding seawater. The CO₂ factor is zero at low alkalinity (or low HCO3- concentrations), plateaus at medium alkalinity, and decreases at high alkalinity (or low CO2(aq) concentrations) (Fig. a). Low temperatures (Fig. b) and low concentrations of dissolved inorganic carbon (DIC; Fig. c) narrow the window of maximum CO₂ factor values. In turn, it is high at high temperatures and DIC concentrations (Fig. d).

Model simulations for this study were branched off after an initial concentration-driven spinup (piControl) of 1051 years with a constant atmospheric CO₂ concentration of 278 ppm and a subsequent emission-driven spinup (esm-piControl) of 871 years. An additional 200 years of esm-piControl spinup with both the NO-CSE and the CSE model version (section ) was computed in parallel before starting the historical simulations (1850–2014; HIST-NO-CSE and HIST-CSE). For the subsequent future simulations, the SSP5-3.4-OS scenario was used. It follows the initial ramp-up of emissions equal to the SSP5-8.5 scenario (the unmitigated baseline scenario) from 2015 to 2039 before strong emission reduction from 2040 onwards, reaching zero emissions in 2070 and net negative emissions thereafter . Net negative emissions in this scenario are obtained by CDR methods other than OAE and additive to the OAE-caused atmospheric CO₂ reduction investigated in our study.

Starting in 2040 we added alkalinity with two different concentrations (“OAE-low”, “OAE-high”) in both model versions (Sect. ). We also computed simulations without alkalinity addition for each model version (“NO-OAE”). Hence, we ended up with six simulations for 2040–2100: NO-CSE-NO-OAE, NO-CSE-OAE-low, NO-CSE-OAE-high, CSE-NO-OAE, CSE-OAE-low, CSE-OAE-high (Table ). Further, we computed one simulation which builds on the CSE-OAE-high simulation but in which no alkalinity was added to a grid cell when the saturation state of aragonite exceeded 10 to avoid conditions that favour abiotic calcium carbonate precipitation (CSE-OAE-high-lim; Table ). We analysed differences between CSE and NO-CSE simulations as well as between OAE-high/OAE-low and NO-OAE simulations using independent two-sample t tests (significance level: α=0.05) either for the annual means of the entire time series 2040–2100 or for the annual means of the last 10 years of the simulation (2091–2100).

Alkalinity was added to the surface ocean in the exclusive economic zones (EEZs; up to 200 nautical miles away from the coastlines) of Europe, the US, and China (Fig. a). Subpolar regions (north of 67.5° N), the Baltic Sea east of 9.5° E and small marginal seas and remote parts of the EEZ (e.g. islands in the Pacific and Atlantic oceans, Greenland) were excluded from the mask. The amount of alkalinity added in the OAE-high simulations scales according to the availability of CaO from cement production , with increasing annual cement production and, hence, CaO additions from 2040 to 2100 (Fig. b). In the OAE-low simulations half of this amount was added to the EEZ. Given the vast growth of the cement production in China over the past decades (Spyros Foteinis, personal communication, 2025), the Chinese EEZ starts with the highest alkalinity addition in the beginning of the deployment time and begins to saturate by the end of the century, while the amounts added to the European and the US EEZs progressively increase over time. In each grid cell of the model, the addition scales with the relative sea-ice cover, which is assumed to prevent the distribution of alkalinity to the surface ocean. While the European and the Chinese EEZs are not affected by sea ice, the deployment in the US EEZ as reported here is 2.5±0.6 % (2.7±1.3 %) smaller in the first (last) decade of the deployment than in the initial deployment mask. In 2100, the amounts of alkalinity in the OAE-high (OAE-low) simulations are 29.7 (14.9) Tmol yr⁻¹ (Europe), 23.9 (12.1) Tmol yr⁻¹ (USA), and 49.6 (24.8) Tmol yr⁻¹ (China).

The efficiency of OAE (ηCO₂) is computed from excess volume-integrated DIC and alkalinity in the OAE relative to the respective NO-OAE simulations as 4ηCO2=ΔDICΔAlkalinity. The reduction in the CO₂ partial pressure in seawater (pCO_2(aq)) due to biological carbon drawdown, ΔpCO_2(aq,bio), is computed based on the biological sources and sinks to DIC (ΔDIC_bio) as well as the surface ocean carbonate system averaged over the upper 100 m of the water column with 5ΔpCO2(aq,bio)=ΔDICbioγDIC⋅pCO2(aq). The buffer factor γDIC=DICR , with R being the Revelle factor ratio of the relative change in pCO_2(aq) to the relative change in DIC;. Sources of DIC_bio are phytoplankton and zooplankton respiration, CaCO₃ dissolution, and remineralization of dissolved organic carbon, while photosynthesis and calcification are sinks of DIC_bio. Due to its strong seasonality, pCO_2(aq,bio) is calculated monthly and then averaged annually. Furthermore, ΔpCO_2(aq,bio) allows the effects of carbonate system states (buffer factor and pCO_2(aq)) and biological feedbacks of different simulations on the biological pCO₂ drawdown to be artificially combined. Both measures, ηCO₂ and ΔpCO_2(aq,biol), are computed both for the global ocean and in the three deployment regions as the mean over 2091–2100.

The reduction in atmospheric pCO₂ in the OAE-low and OAE-high simulations relative to the NO-OAE simulations ranges from 2.9 to 7.8 µatm (mean 2091–2100) and scales roughly with the amount of alkalinity added (Fig. a). Surface alkalinity (mean 2091–2100) in the OAE-high (OAE-low) simulation increases by 104–105 mmol m⁻³ (53–54 mmol m⁻³) in the European EEZ, 100–106 mmol m⁻³ (50–52 mmol m⁻³) in the US EEZ, and 619–649 mmol m⁻³ (313–321 mmol m⁻³) in the Chinese EEZ (range given for CSE and NO-CSE simulations). Relative to the simulation without alkalinity addition (NO-OAE), this is an increase in surface alkalinity of 5 %–30 % for OAE-high (2 %–15 % for OAE-low). Globally, pCO_2(aq) in the upper 100 m mainly follows the trajectory of the prescribed CO₂ emissions (increasing pCO_2(aq) by about 60 µatm until 2060, decreasing pCO_2(aq) thereafter), and differences between the CSE and the NO-CSE simulations without alkalinity addition are largely caused by the model setup (Fig. a). In the deployment regions, pCO_2(aq) in the NO-OAE simulations is additionally modified by regional dynamics in the air–sea CO₂ fluxes (Fig. b–d). Furthermore, by 2100 the simulations differ according to the amount of alkalinity added, with the highest pCO_2(aq) values in the NO-OAE and the lowest in the OAE-high simulations. Reductions in near-surface air temperatures and sea surface temperatures relative to the NO-OAE simulations are non-significant. This is probably due to both the relatively small reduction in atmospheric pCO₂ that may not result in temperature changes that go beyond natural variability as well as the time lag in the Earth system response, in line with .

OAE causes a significant increase in the global cumulative air–sea CO₂ flux in all simulations, ranging between 11.4 and 26.0 Pg C by 2100 (Figs. b and , Table ). About half of the OAE-induced CO₂ flux change occurs outside the deployment regions (Table ). Including areas of 1000 km distance from the deployment region, resulting in areas that are 3–4 times the size of the original deployment regions (Fig. ), covers 80 %–90 % of the excess CO₂ uptake (Table ). Especially in the US and the Chinese EEZ, a considerable share of excess ocean CO₂ uptake happens in the >1000 km surrounding area of the deployment region (Fig. a–c). Within the deployment regions, OAE is least efficient in the Chinese EEZ (0.34–0.45) and most efficient in the US EEZ (0.65–0.88; Table ). Efficiency in the Chinese EEZ increases when considering the surrounding 1000 km (0.55–0.67; Fig. f, Table ). Factors that can decrease efficiencies relative to theoretical values are feedbacks from the land carbon cycle, alkalinity losses by calcification, and the transport of alkalinity into deeper water parcels where CO₂ exchange with the atmosphere is impossible.

When accounting for carbonate system effects on phytoplankton, the ocean takes up 11 %–16 % more excess CO₂ than without these effects (12.6 Pg C versus 11.4 Pg C and 26.0 Pg C versus 22.4 Pg C, respectively; Table ), which is also reflected in higher efficiency values in almost all simulations (Table ). However, the reduction in atmospheric pCO₂ is smaller due to the weakened land CO₂ uptake (Table ), likely resulting from the different state of the climate system (radiative forcing and resulting effects on, for example, temperature, precipitation, and winds). The stronger CO₂ sink in the CSE simulations could be driven solely by lower initial alkalinity and DIC concentrations compared to the NO-CSE simulations (Fig. ) which chemically favour the uptake of atmospheric CO₂. To identify whether carbonate system effects on phytoplankton play an additional role in enhancing the ocean CO₂ uptake in the CSE simulations, we computed ΔpCO_2(aq,bio) (Eq. ).

The biological pCO₂ drawdown is a function of the strength of primary production and of the buffer capacity of seawater. Here, we combine results of different simulations to disentangle the roles of changing marine pelagic net primary production (NPP) in response to OAE and of different carbonate system states for the simulated ocean carbon uptake. The biological pCO₂ drawdown is consistently smaller in simulations with OAE than in those without, independent of whether carbonate system effects on phytoplankton growth are represented (for the CSE simulations: by 22 % in the European and US EEZs, by 62 % in the Chinese EEZ, and by 5 % globally; p value < 0.05; Fig. , Table ). This is because OAE increases the buffer capacity and thus reduces the imprint of NPP on pCO₂ drawdown . On top of that, two competing processes are responsible for differences between the CSE and NO-CSE simulations. Firstly, the CSE effects alter NPP and thus pCO_2(aq,bio). Globally, this biological response in the CSE simulation leads to a decrease in the pCO_2(aq,bio) drawdown (-1.4 µatm yr⁻¹ averaged over 2091–2100; Fig. ) and, hence, to a weaker ocean CO₂ sink. Secondly, however, the different carbonate system state leads to a slightly larger biological pCO₂ drawdown (+0.1 µatm yr⁻¹ averaged over 2091–2100; Fig. ). Note that both effects on the global biological pCO₂ drawdown are not significant. Thus, the stronger global ocean CO₂ sink in the CSE simulations must be fully driven by the state of the carbonate system, facilitated by a lower alkalinity-to-DIC ratio (1.118 in the CSE versus 1.123 in the NO-CSE simulation), a resulting lower buffer capacity, and a resulting larger efficiency of OAE and not by biological feedbacks. Consistent with the findings in , we find that the baseline alkalinity and DIC concentrations are pivotal for the surface ocean pCO₂ reduction after alkalinity addition, which highlights the need for a careful assessment of the initial carbonate system states in OAE model studies. However, we observe that carbonate system effects on phytoplankton can indeed locally enhance the ocean CO₂ uptake following OAE, for example in the Chinese EEZ, where the biological pCO₂ drawdown is significantly increased by +1.0 µatm yr⁻¹ (Fig. ).

In the deployment regions of the simulations with carbonate system effects on phytoplankton, marine NPP is lower with OAE relative to simulations without OAE, with a significant difference especially in the Chinese EEZ (up to -15 % in 2091–2100; Table ). Less well-pronounced anomalies can be seen on the global scale in the CSE simulations as well as in all NO-CSE simulations, where changes in marine NPP can only be caused by indirect OAE effects such as modifications of the radiative balance, winds, and mixed layer depth (Table ). In accordance with this, annual NPP anomalies are negatively correlated with surface alkalinity anomalies in the simulations with carbonate system effects on phytoplankton (p<0.05; Fig. a) but not those without (p>0.05; Fig. b).

Decreasing NPP in the CSE-OAE simulations within the deployment areas is mainly driven by diatoms (significant negative correlation of diatom NPP with amount of alkalinity added; Fig. a) and slightly dampened by small phytoplankton (significant positive correlation of small phytoplankton NPP and amount of alkalinity added; Fig. b). Enhanced small phytoplankton NPP cannot fully balance the lower diatom NPP because of its smaller contribution to overall NPP (according to a community analysis in the simulations averaged over 5 years prior to the alkalinity deployment, 3 % of small phytoplankton versus 97 % of diatoms contribute to NPP in the Chinese EEZ; 40 % versus 58 % in the US EEZ). Nevertheless, within the bounds of the phytoplankton community described in our model, this indicates a community shift towards fewer large cells (diatoms) and more small cells (small phytoplankton) assuming unchanged grazing pressure on each group. Modifications in small phytoplankton and diatom NPP in the NO-CSE-OAE simulations are not correlated with the amount of added alkalinity (Fig. a and b).

Both coccolithophore NPP and CaCO₃ of the CSE-OAE simulations decrease significantly with increasing alkalinity addition inside the deployment regions (Fig. c and d, Table ). In the simulation without carbonate system effects, where calcification is performed by a fixed share of small phytoplankton, changes in CaCO₃ and the amount of added alkalinity are not correlated (Fig. C, Table ). While the PIC : POC ratio in the NO-CSE simulations is defined to be constant, both coccolithophore PIC and POC can vary independently from each other in the CSE simulations. A significantly higher PIC : POC ratio in the European EEZ and globally in the OAE-high simulation with carbonate system effects on phytoplankton (+0.01, resulting in PIC : POC ratios of 1.18 and 1.14, respectively; Table ) points towards more strongly calcifying coccolithophores, while a decreasing PIC : POC ratio in the Chinese EEZ (-0.05, resulting in a ratio of 1.10; Table ) reflects lighter calcifying coccolithophores.

To assess whether the CO₂ factor is the primary driver of the negative correlation between OAE and NPP as well as CaCO₃ concentration, we examined changes in the factor over the time period of the OAE deployment. In some years before 2070, the CO₂ factor is smaller in the OAE compared to the NO-OAE simulations in the European and the US EEZ because of the small OAE signal caused by low alkalinity additions (Fig. a, b, d, and e). At the latest from 2070 onwards, the factor is always higher in the CSE-OAE simulations relative to the CSE-NO-OAE simulations (Fig. ). Hence, OAE is always beneficial for gross photosynthesis and calcification within the assumptions of our model, and strengthened CO₂ limitation does not play a role for marine NPP on a regional and global level under sustained OAE.

Instead, the inverse response of small phytoplankton and diatoms to OAE is likely caused by a stronger increase in the CO₂ factor of small phytoplankton, which alone would lead to a higher photosynthesis rate (Fig. a–c). Indeed, our parameter choice in f(CO₂)_i allows for a higher CO₂ factor for small phytoplankton in comparison to diatoms under increasing alkalinity concentrations (Fig. a), pointing towards a competitive advantage of small phytoplankton over diatoms. It was shown in that small modifications in the CO₂ factors can trigger considerable shifts in the phytoplankton community, even by a phytoplankton group that is less represented in the respective region. Hence, the unequal increase rather than a decrease in the CO₂ factor is the likely cause for decreasing diatom and increasing small phytoplankton NPP. Decreasing diatom NPP can be further enhanced by indirect OAE effects that are caused by OAE-induced modifications in the atmospheric CO₂, which changes the radiative balance, winds, and, finally, the mixed layer depth and other drivers that have the potential to modify bottom-up and top-down effects on phytoplankton similar to.

The stronger increase in the CO₂ factor of calcification relative to coccolithophore NPP (Fig. d–f) would suggest that the PIC : POC ratio should increase by 2100. We see this in the European EEZ and globally (Table ), but not in the other EEZs. Similar to our interpretation of decreasing diatom NPP, we hypothesize that indirect OAE effects (e.g. on the radiative balance, winds, mixed layer depth) as well as the competition with the other phytoplankton groups dominate here, which is supported by the fact that both coccolithophore NPP and CaCO₃ concentrations decrease relative to the NO-OAE simulations despite the increasing CO₂ factor. Furthermore, coccolithophore biomass (not NPP) can be modified by changes in the grazing pressure, resulting in changes in the PIC : POC ratio that deviate from modifications in NPP.

Our simulations show that OAE can decrease NPP significantly through the coupling of direct OAE responses (i.e. the CO₂ factor) to indirect feedbacks (e.g. competition, cascading effects on bottom-up and top-down drivers). Adding half of the amount of alkalinity (CSE-OAE-low versus CSE-OAE-high) effectively reduces the decline in total NPP in the Chinese EEZ by 60 % (3.96 g C m⁻² yr⁻¹ versus 9.18 g C m⁻² yr⁻¹), but the NPP decline in the European EEZ is 4 times higher than in the OAE-high simulation (2.48 g C m⁻² yr⁻¹ versus 0.61 g C m⁻² yr⁻¹; Table ). Moreover, the reduction in atmospheric pCO₂ by 2100 scales with the amount of alkalinity added, reaching only about 50 % in the CSE-OAE-low simulation compared to the CSE-OAE-high simulation (2.9 µatm versus 5.7 µatm; Table ). This poses the question of whether a more targeted limitation of alkalinity addition could mitigate OAE impacts on phytoplankton while preserving CDR effectiveness.

With the motivation to avoid conditions in which abiotic CaCO₃ precipitation could happen, we complemented a CSE-OAE-high simulation in which no alkalinity was added to a grid cell when the saturation state of aragonite exceeded 10 (CSE-OAE-high-lim). This threshold is only exceeded in the Chinese EEZ, reducing the amount of added alkalinity by up to 4 mol m⁻² yr⁻¹ (about 10 %) compared to the CSE-OAE-high simulation and dampening the increase in surface alkalinity to 469 mmol m⁻³ (compared to 649 mmol m⁻³ in the CSE-OAE-high simulation). The significant NPP decrease in the Chinese EEZ of the CSE-OAE-high-lim simulation is dampened to 48% of the CSE-OAE-high simulation (4.75 g C m⁻² yr⁻¹ versus 9.18 g C m⁻² yr⁻¹), similar to the CSE-OAE-low simulation. The significant NPP decline in CSE-OAE-high-lim in the European EEZ is still stronger than in the CSE-OAE-high simulation (2.01 g C m⁻² yr⁻¹ versus 0.61 g C m⁻² yr⁻¹), but less strong than in the CSE-OAE-low simulation (2.48 g C m⁻² yr⁻¹; Table ). NPP in the US EEZ decreases, however, more (up to 37 %) than in both CSE-OAE-low and CSE-OAE-high. The decrease in the PIC : POC ratio in the Chinese EEZ vanishes, while PIC : POC anomalies in the other EEZs and globally are comparable to the CSE-OAE-high simulation (Table ). Whereas the OAE effects on the ecosystem with limited alkalinity addition are often smaller compared to the CSE-OAE-high simulation, it even increases the CDR effectiveness: atmospheric pCO₂ in the CSE-OAE-high-lim simulation is reduced by 7 % more (6.1 µatm) than in the CSE-OAE-high simulation (5.7 µatm; Table ).

With up to 1 Pg C yr⁻¹ (in the CSE-OAE-high simulation, with a global alkalinity input of 96 Tmol yr⁻¹), OAE has the potential to store more than 3 times as much atmospheric CO₂ in the ocean than natural rock weathering 0.3 Pg C yr⁻¹; and about 40 %–60 % of the residual and hard-to-abate emissions 1.6–2.7 Pg C yr⁻¹;. Approximately 50 % of this excess CO₂ flux occurs outside the deployment regions, partly even in the periphery of 1500 km, likely depending on the prevailing surface ocean currents in the deployment region that transport the alkaline material away from its initial injection site. Water transport can also modify the time in which alkalinized waters are in contact with the atmosphere, thus allowing gas exchange. This reduced time for equilibration is likely what we observe in the US EEZ (Fig. e). A similar share of 50 % was observed in the coastal OAE model study of . These dynamics emphasize the need for large-scale “monitoring, reporting, and verification” (MRV) processes (quantify the efficiency of CDR activities) as they require tracking patches of artificially elevated alkalinity beyond the deployment region to fully assess the excess CO₂ taken up by the ocean . Further complicating MRV, our study shows that the enhanced ocean sink by OAE is partially compensated, or even overcompensated, by a reduced land sink, in line with previous studies e.g.. This leads to a smaller reduction in atmospheric pCO₂ than would be expected from monitoring the air–sea CO₂ flux alone this study;.

Globally, carbonate system effects on phytoplankton are not the reason for higher air–sea CO₂ fluxes in the CSE compared to the NO-CSE simulation, but rather dissimilarities in initial surface alkalinity and DIC caused by differences in representations of the CaCO₃ cycle in the two model versions. In the European and Chinese EEZs, however, we indeed see that carbonate system effects on phytoplankton increase the potential of the ocean to take up atmospheric CO₂ relative to the NO-CSE simulations. Additional investigations are needed for the effect of biological feedbacks on long-term organic carbon storage: we see higher export efficiency (POC flux at 100 m/total NPP; ) and transfer efficiency (POC flux at 1000 m/POC flux at 100 m; ) in the Chinese EEZ of the CSE simulation (0.6 % and 2.6 %, respectively) compared to the NO-CSE simulation (0.5 % and 2.4 %, respectively), but lower export and transfer efficiencies in the European EEZ (CSE: 6.2 % and 8.0 %, respectively; NO-CSE: 6.6 % and 8.2 %, respectively). This suggests more efficient organic carbon export to depth in the Chinese EEZ due to the biological feedbacks, but increased surface remineralization and, hence, less effective deep organic carbon storage as a result of biological effects in the European EEZ. This has implications for MRV as it suggests that biological feedbacks need to be taken into account when assessing the excess ocean CO₂ sink and export of organic carbon to the deep ocean resulting from OAE.

We found that OAE-induced anomalies in surface alkalinity correlate with a decrease in total NPP. As the CO₂ factor, the primary link between changes in the carbonate system and phytoplankton photosynthesis, does not imply reduced growth, other indirect OAE effects must diminish NPP. In fact, this is in line with the hypothesis of that shifts in the carbonate system by OAE are too small to trigger a significant effect on productivity. Yet, an imbalance in the change in the CO₂ factor between phytoplankton groups can result in a modified habitat competition in bottom-up and top-down factors at the expense of the dominant phytoplankton group. A similar finding for the CO₂ factor, termed as “cascading effects”, was described by . Thus, positive or neutral OAE effects on phytoplankton in laboratory studies must not necessarily result in positive or no effects of OAE on primary producers in the real ocean. Point observations of the plankton community before, during, and after OAE applications would increase the understanding of direct and indirect ecosystem responses to OAE.

Our simulations do not confirm the ecological realization of the physiologically beneficial effect of OAE on coccolithophores, which was hypothesized by . NPP of coccolithophores decreases with progressing OAE, and PIC : POC decreases in the Chinese EEZ, with the highest alkalinity deployment rates; both increasing coccolithophore NPP and PIC : POC ratios would have been a sign of the “white ocean” . Accordingly, our study could not confirm the reduction in surface alkalinity by enhanced calcification, which would reduce the efficiency of OAE part of the “leakage term”;. It has to be noted, however, that our alkalinity addition is only up to 10% of the addition required to trigger coccolithophore proliferation according to (1.1 Pmol yr⁻¹ versus a total of 0.1 Pmol yr⁻¹ in 2100 in our simulations). The authors also point out that this global response may be overridden by other environmental factors on a regional or local scale, which is likely the case in our study as well. Furthermore, coccolithophores constitute only a very small part (<0.1–0.8 %) of the phytoplankton community in the deployment regions of our model. In the real ocean, however, the deployment regions do host coccolithophore blooms . Repeating the simulations with OAE deployment in coccolithophore hotspots of our model (e.g. the North Atlantic and the equatorial Pacific) instead of the EEZs or improving the coastal coccolithophore representation in our model could provide further insights into the alkalinity leakage.

Deploying only half of the alkalinity reduces atmospheric pCO₂ by 50 %, as expected, but at the benefit of mitigating negative effects on the ecosystem. Hence, reducing the OAE deployment more locally (in this study, the reduction depends on the saturation state of aragonite) can be as or even more effective than deploying as much alkalinity as possible while minimizing the effects on the ecosystem. In the real ocean this would also reduce the risk of secondary mineral precipitation, a process that we do not parameterize in our model. Hence, in future OAE applications it is about finding the optimum between effectiveness and environmental impact. This optimum is likely unique for each deployment region, which hinders our ability to give quantitative suggestions for the ideal amount of alkalinity addition.

We only consider alkalinity effects on phytoplankton, but zooplankton may be equally sensitive to OAE. Although the study of reveals that the plankton food-web in a mesocosm experiment was relatively resistant to OAE, the authors list potential vulnerabilities that may appear in other plankton communities. For example, zooplankton could be affected by OAE-induced changes in the nutritional value of their prey . Studying community-level effects of OAE in ecosystem models would reveal a better understanding of its large-scale ecosystem effects but is currently impeded by lacking data on zooplankton–OAE interactions for model parameterizations.

Just as other Earth system models e.g., both the CSE and the NO-CSE version have a bias towards low surface alkalinity in comparison to observations (Fig. ). We especially consider the representation of calcium carbonate dissolution above the saturation horizon as well as the improved biogeographical representation of plankton calcification other than coccolithophores to be worthy of improvement. As shown by , biases in the surface alkalinity can indeed lead to an overestimation of the excess CO₂ uptake, which should be taken into account when transferring model findings to real ocean applications.

Ensemble simulations would help to quantify the effects of internal variability, thereby giving a better idea of the potential indirect effects of OAE on the ecosystem. However, since our study focuses on direct OAE effects in the upper ocean, where the signal-to-noise ratio is high, we consider our study to be a robust first step towards a better understanding of the mechanistic effects of OAE on phytoplankton.