Future Changes in Carbon Chemistry Under the Implementation of Artificial Ocean Alkalinization Based on CMIP6 Simulations

Abstract

Artificial ocean alkalinization (AOA) is one of the most promising marine carbon dioxide removal technologies, but its influence on marine carbon chemistry remains unclear. We applied data from the Sixth Coupled Model Intercomparison Project (CMIP6) to characterize the temporal and spatial variabilities of future marine carbon chemistry under the implementation of AOA. Our study shows that the marine carbon system varied significantly under the implementation of AOA, but some specific effects may be masked by the forcing of the high carbon emission scenario SSP5-8.5. Based on the CMIP6 protocol, which added 0.14 Pmol of alkalinity into the ocean annually, AOA promoted an increase in DIC, delayed the rise in pCO₂, and mitigated declines in pH and Ω, respectively. The temperate oceans in both hemispheres were the most significantly impacted basins, whereas the Southern Ocean was the least affected. During this century, the oceanic carbon sink is expected to intensify rapidly until around the year 2080, and then gradually weaken. The implementation of AOA merely changed the relative strength of the oceanic sink, rather than its overall variation pattern. Furthermore, in the deep ocean, the effect of AOA was present but quite limited in mitigating ocean acidification.

1. Introduction

According to the Sixth Assessment Report by the Intergovernmental Panel on Climate Change (IPCC), the global surface temperature increased by 1.1 °C during 2011–2020 compared to 1850–1900, which is unequivocally caused by human activities through the emission of greenhouse gases (GHGs) [1]. Climate research conducted in recent decades has found that carbon dioxide (CO₂), the largest contributor to GHGs, is at higher levels than at any time in the past 200,000 years of human history, having already surpassed a concentration of 420 μatm by the end of 2024. Moreover, the historical cumulative net emission of CO₂ from 1850 to 2019 is estimated to be as much as 2400 ± 240 Gt [1] (1 Gt = 10⁹ t). Consequently, limiting temperature increases to well below 2 °C, and ideally 1.5 °C above pre-industrial levels by the end of the century, as outlined in the Paris Agreement’s objectives, requires not only rapid reductions in greenhouse gas emissions but also removing billions of tons of CO₂ from the atmosphere [1].

Marine carbon dioxide removal (mCDR) technologies have recently gained extensive attention because it is estimated that about a quarter of the CO₂ emissions have been absorbed by the ocean since the Industrial Revolution [2,3]. Thus, the ocean is expected to have a great influence on the mitigation of climate change. As a whole, the ocean, serving as the largest pool of reactive carbon, contains about 38,000 Gt of dissolved inorganic carbon (DIC) [4]. The exchange of CO₂ between the atmosphere and the ocean mixed layer (roughly the top 100 m of the ocean), primarily regulated by the concentration gradient, temperature, and wind speed, is fairly rapid. The characteristic time scale for this air–sea exchange process is on the order of years. Equation (1) describes what CO₂ undergoes after it enters the ocean [5] (see Appendix A for details). The gaseous CO₂ firstly transforms into its aqueous form and then forms carbonic acid (H₂CO₃). (In fact, the aqueous form of CO₂ and H₂CO₃ are difficult to distinguish technically). Carbonic acid rapidly dissociates into free hydrogen ions (H⁺) and bicarbonate ions (HCO₃⁻). Then, the bicarbonate further dissociates into H⁺ and carbonate (CO₃²⁻) at a relatively slow rate. The dissolved species in Equation (1), predominantly in the form of HCO₃⁻, generally make up the carbonate alkalinity system of seawater. From a chemical perspective, total alkalinity (TA), which roughly refers to the excess of proton acceptors over proton donors (Equation (2)), is a crucial parameter that largely determines the buffering capacity for CO₂ in seawater [6].

CO₂ (g) ⇌ CO₂ (aq), CO₂ (aq) + H₂O ⇌ H₂CO₃, H₂CO₃ ⇌ HCO₃⁻ + H⁺, HCO₃⁻ ⇌ CO₃²⁻ + H⁺

(1)

TA = [HCO₃⁻] + 2[CO₃²⁻] + [B(OH)₄⁻] + [OH⁻] + [HPO₄²⁻] + 2[PO₄³⁻] + [H₃SiO₄⁻] + [NH₃] + [HS⁻] – [H⁺]F – [HSO₄⁻] – [HF] – [H₃PO₄] + … – …

(2)

Based on the above-mentioned processes of ocean carbon chemistry, the fundamentals of artificial ocean alkalinization (AOA) are proposed as follows (see Appendix B for details). Increasing alkalinity drives the consumption of H⁺ and the production of bicarbonate (HCO₃⁻) and carbonate (CO₃²⁻), which leads to a corresponding increase in pH and a following decrease in the partial pressure of CO₂ (pCO₂) in seawater, respectively. These changes would ultimately promote the uptake of gaseous CO₂ from the atmosphere via the air–sea exchange process.

AOA is considered one of the most promising mCDR methods, with a theoretical sequestration potential in the range of 3 to 30 Gt CO₂ yr⁻¹ [7,8,9]. Previous studies have explored some aspects of AOA. For example, based on an ocean carbon cycle model, Ilyina et al. [10] confirmed that intensive enhancement of ocean alkalinity has the potential to promote oceanic uptake of CO₂ from the atmosphere and could avoid further ocean acidification in the meantime. Feng et al. [11] ran AOA simulations in the Great Barrier Reef, Caribbean Sea, and South China Sea using an Earth system model of intermediate complexity, and found that alkalinization could counteract the local acidification changes expected in the 21st century in terms of both oceanic surface pCO₂ and surface aragonite saturation (Ω). However, Keller et al. [12] found that the ultimate effect of AOA was limited by the production and transport capacity of alkalinity material. Thus, the practical AOA-induced reduction of atmospheric CO₂ under current conditions is relatively small compared with the expected business-as-usual CO₂ emissions, and as a result, the atmospheric CO₂ will continue to rise. Zhou et al. [13] presented global maps of AOA efficiency, defined as the ratio between the change in DIC and the amount of added alkalinity. They found that the equilibration kinetics had two characteristic timescales: rapid surface equilibration followed by a slower second phase. These kinetics vary considerably with latitude and the specific season of alkalinity release.

At present, there seems to be little consensus on the climate impacts and marine response to AOA, especially in the context of continuously rising carbon emissions. Methodologically, AOA implementation could be obtained by both observation-based estimations, e.g., [14,15,16], and model simulation outputs, e.g., [17,18,19]. But, in a way, model simulations can overcome the inevitable temporal–spatial limitation of observation-based methods. The Earth system model (ESM) is the latest generation of the state-of-the-art climate models, which couples the carbon cycle processes among the atmosphere, land, and ocean, and simulates real Earth systems to the maximum extent. Therefore, the ESM could serve as a helpful and powerful tool to analyze and diagnose the implementation of AOA.

To address this, we apply the up-to-date datasets produced by the new version of ESM from the Sixth Coupled Model Intercomparison Project (CMIP6) to characterize the temporal and spatial variabilities of marine carbon chemistry under the implementation of AOA. This paper will provide fundamental information on the CMIP6 AOA experiment, present the variabilities of the four most important carbonate chemistry parameters (DIC, pH, pCO_2, and Ω), and estimate the long-term average and future tendency of air–sea CO₂ exchange flux (FCO₂). The remainder of this paper is organized as follows. Section 2 introduces the model, datasets, and analytical materials. The results and discussions are presented in Section 3, and Section 4 presents the conclusion.

2. Materials and Methods

We used the NorESM2-LM model in this study, which is the second generation of the fully coupled Earth system model developed by the Norwegian Climate Center [20]. For details, the atmosphere component of NorESM2-LM was built on the Community Atmosphere Model version 6 (CAM6) but with particulate aerosols and aerosol–radiation–cloud interaction parameterization, which is referred to as CAM6-Nor, consequently. The ocean component is the Bergen Layered Ocean Model (BLOM), which employs an isopycnic vertical coordinate with near-isopycnic interior layers and variable density layers in the surface mixed-boundary layer [20]. The ocean biogeochemistry component of NorESM2-LM was adapted from the HAMburg Ocean Carbon Cycle model and was converted to isopycnic coordinate (iHAMOCC) [21], which prognostically simulates five key ocean biogeochemical cycle processes: the inorganic seawater carbon chemistry, the NPZD (Nutrient Phytoplankton Zooplankton Detritus)-type ecosystem module, the air–sea gas exchange, the vertical fluxes of inorganic and organic particles, and sediment biogeochemistry. The sea ice model component is based upon the Community Ice CodE (CICE) sea ice model [22]. What is more, the NorESM2-LM employs the latest version of the Community Land Model (CLM5) as the land component [23].

Data analyzed in this study were derived from the monthly outputs of three CMIP6 experiments: the esm-hist, the esm-ssp585, and the esm-ssp585-ocn-alk [24,25]. Briefly, the CMIP6 esm-hist simulation, in which the atmospheric CO₂ concentration is calculated according to the historical anthropogenic CO₂ emissions forcing, was essential for the reliability testing before the ocean alkalinization experiment. This experiment was run from the years 1850 to 2014. The esm-ssp585, then, is driven by the SSP5-8.5 high CO₂ emission scenario and runs from the end year of the esm-hist simulation until the end of this century (2015 to 2100) and serves as the control run and branching point for the subsequent ocean alkalinization experiment. The esm-ssp585-ocn-alk simulation, forced by the SSP5-8.5 high CO₂ emission scenario, also adds 0.14 Pmol TA (1 Pmol = 10¹⁵ mol) to the upper ice-free ocean surface waters between 70°N and 60°S every year from 2015 to 2100. In general, the average differences between the AOA esm-ssp585-ocn-alk and the no-AOA esm-ssp585 of specific parameters are obtained as the net effect of the AOA implement. What is more, the key information about the CMIP6 experiments conducted by the NorESM2-LM is summarized in

Table 1. Ocean alkalinization experiment simulations conducted by the NorESM2-LM.

CMIP6 Experiment ID	Simulation Description	Run Time
esm-his	CO2-emission-driven historical scenario	1850–2014
esm-ssp585	CO2-emission-driven SSP5-8.5 scenario	2015–2100
esm-ssp585-ocn-alk	SSP5-8.5 scenario with 0.14 Pmol yr−1 alkalinity added	2015–2100

below.

The observational dataset employed in this paper is the climatological air–sea CO₂ flux from the Lamont–Doherty Earth Observatory of Columbia University, with the original resolution of 4° × 5° [26], which was widely used in the study of the global carbon cycle. The air–sea CO₂ exchange flux climatology of Lamont–Doherty Earth Observatory was based upon about 3 million measurements of surface water pCO₂, obtained from 1970 to 2007. Therefore, the model result was also calculated for the same period. We re-gridded the climatology data of Lamont–Doherty Earth Observatory into a 1° × 1° grid and referred to it as “Takahashi 2009” [26] for a comparison of the long-term average and spatial distributions of model bias. In this article, the trend of a specific carbon chemistry parameter was calculated on the gridded monthly data using the ordinary least squares regression method. Additionally, the time series of a parameter was calculated by the area-weighted mean value of global gridded data.

3. Results and Discussion

3.1. Performance of the Model Simulation

The simulated global air–sea CO₂ flux in the NorESM2-LM esm-his experiment was compared with the observational results by Takahashi 2009 [26] to assess the general performance of the model (Figure 1). The comparison indicates that the model could generally reproduce the dominant patterns of Takahashi 2009: the equatorial Pacific is the major oceanic CO₂ source for the atmosphere, and the temperate oceans in both hemispheres are the major oceanic sinks for atmospheric CO₂, with the North Atlantic being the most intensive CO₂ sink. What is more, the long-term average annual net uptake flux of CO₂ by the global oceans has been estimated to be 2.05 PgC yr⁻¹ for the NorESM2-LM model, which is exactly in accordance with the climatological results of Takahashi et al. [26,27] (1.6 to 2.2 PgC yr⁻¹). What is more, Qu et al. [28] had already assessed the long-term average and spatial–temporal variability of the global air–sea CO₂ exchange flux (FCO₂) since the 1980s, based on the results of 18 CMIP6 ESMs. They found that the NorESM2-LM performed particularly well compared with the other CMIP6 ESMs because its root mean squared error (RMSE), with respect to the observation results, was quite low among all 18 CMIP6 ESMs. These consistencies arguably indicate that the NorESM2-LM performs fairly well in simulating the mean state of the global air–sea CO₂ exchange process. As a result, we could apply its following results to esm-ssp585 and esm-ssp585-ocn-alk for further investigation.

3.2. Variations of the Marine Carbon System During AOA

Generally, the carbonate system can be described by six fundamental parameters in thermodynamic equilibrium: DIC, TA, [CO₂], [HCO₃⁻], [CO₃²⁻], and pH. The remaining concentrations of OH⁻ and pCO₂ can be readily calculated using the dissociation constant of water and Henry’s law [5]. Given the first and second dissociation constants of carbonic acid and the definitions of DIC and TA, the marine carbon system can be constrained by knowing two carbonate chemistry parameters, along with temperature, salinity, and pressure. In order to depict the variations in the marine carbon system under the implementation of AOA more comprehensively, we investigated four important marine carbon parameters, except TA, from the esm-ssp585-ocn-alk experiment, and compared them with those from the esm-ssp585 experiment; they are DIC, pCO₂, pH, and Ω. It is worth pointing out that DIC and pCO₂ are usually adopted to characterize the budgets of the oceanic carbon source/sink, and Ω and pH are used to reveal the features of ocean acidification.

Overall, the implementation of AOA indeed changed the response of the marine carbon system to the ever-growing carbon emission significantly. Simulated by the esm-ssp585-ocn-alk experiment conducted by the NorESM2-LM, the global-averaged concentrations of DIC and pCO₂ obviously increased, with rates of 2.54 μmol·kg⁻¹·yr⁻¹ and 6.80 μatm·yr⁻¹ during the period from 2015 to 2100 (Figure 2a and Figure 2b), respectively, while the pH and Ω clearly decreased with rates of −0.041 yr⁻¹ and −0.0156 yr⁻¹, respectively (Figure 2c and Figure 2d). The raised concentrations of DIC and pCO₂ were easy to explain because the added alkalinity contains abundant carbon in the form of [HCO₃⁻] or [CO₃²⁻]. Therefore, the DIC and pCO₂ would elevate to high levels through the chemical balance of the marine carbon system automatically [5]. However, the reduced results of the pH and Ω after AOA seemed perverse to some degree (with an enhanced TA but aggravated acidification?). We attributed these results to the relative inadequacy of alkalinity with respect to the high-carbon emission budget. In the CMIP6 project, the esm-ssp585-ocn-alk experiment was also driven by the SSP5-8.5 high CO₂ emission scenario. Shared socioeconomic pathways (SSPs) are climate change scenarios of projected socioeconomic global changes up to 2100, as defined in the IPCC’s Sixth Assessment Report on climate change in 2021. SSP5-8.5 is a very high greenhouse gas emissions scenario, where CO₂ emissions triple in 2075 and the concentration of atmospheric CO₂ is estimated to be as high as more than 1100 μatm in 2100 [29]. The CMIP6 esm-ssp585-ocn-alk experiment only adds 0.14 Pmol TA to the upper ice-free ocean surface waters between 70°N and 60°S every year from 2015 to 2100. The aforementioned results of the pH and Ω imply that AOA with the intensity of the esm-ssp585-ocn-alk experiment in CMIP6 cannot counteract the effects of the SSP5-8.5 high CO₂ emissions. As a matter of fact, the original intention of the esm-ssp585-ocn-alk experiment was not to test the maximum potential of AOA, which would be fairly difficult given the way in which ocean carbonate chemistry is simulated, but rather to compare the response of the models to significant alkalinity perturbations [12]. Nevertheless, if we compare the esm-ssp585-ocn-alk experiment with the esm-ssp585 experiment without the AOA measurements (blue dash line in Figure 2), it can be explicitly detected that AOA promoted an increase in DIC, delayed a rise in pCO₂, and restrained the aggravation of pH and Ω, actually (Figure 2).

As for the spatial features’ changes in surface distributions of the marine carbon system, we calculated the differences between the AOA esm-ssp585-ocn-alk experiment and the no-AOA esm-ssp585 experiment, and present their long-term average results in a global perspective in Figure 3. It was conspicuous that the implementation of AOA was able to induce increases in the DIC, pH, and Ω, and a reduction in pCO₂ globally (Figure 3a–d). It needs to be specifically pointed out that the above-mentioned “increases” and “decreases” were relative to the no-AOA esm-ssp585, of course. In terms of spatial distribution, the temperate oceans in both hemispheres were the most significant impacted basins, whereas the Southern Ocean was the less affected region. This spatial distribution pattern had also been confirmed by previous studies, e.g., [10]. To a great extent, this was because the alkalinity was only added to the ice-free ocean surface waters between 70°N and 60°S based on the CMIP6 protocols. Therefore, the Southern Ocean, which is generally located south of 60°S latitude, should be the least disturbed ocean for AOA before the seawater is affected by alkalinization and transported there by ocean flow and mixing. Nevertheless, some studies argued that adding alkalinity to surface seawater, which contacts the atmosphere for a much longer period, was more effective in lowering atmospheric CO₂ [10]. Therefore, in view of the circulation pattern and the prevailing Antarctic Circumpolar Current, alkalinization at the Southern Ocean site had the fastest and largest effect on regional surface carbon system chemistry, whereas alkalinity added in the surface waters in the North Atlantic had only a small effect on surface carbon system chemistry because the North Atlantic was the area where deep water formed. By reason of the foregoing, the esm-ssp585-ocn-alk experiment of CMIP6 simulated the overall variation pattern of the marine carbon system during AOA.

3.3. Effect of AOA on Air–Sea CO₂ Exchange Flux

Artificial ocean alkalinization is considered one of the most promising ocean-based carbon dioxide removal methods. By increasing surface TA, the seawater carbon chemistry equilibrium system moves towards being bicarbonate (HCO₃⁻) and carbonate (CO₃²⁻), and then the surface pCO₂ decreases subsequently, simulating the net transfer of CO₂ from the atmosphere to the ocean. This is the marine chemistry fundamental for AOA. In this section, we will present the influence of AOA on the air–sea CO₂ exchange flux.

The annual net global air–sea CO₂ exchange flux was adopted from the variable of “fgco2” in the no-AOA esm-ssp585 experiment and the AOA esm-ssp585-ocn-alk experiment, which means “surface downward mass flux of carbon dioxide expressed as carbon”. We obtained two instructive findings (Figure 4): in the present century, the oceanic carbon sink will intensify observably until the year around 2080, and then slow down gradually. The implementation of AOA merely changed the relative strength of the oceanic sink rather than the above variation pattern. To be specific, the annual air–sea CO₂ exchange flux enhanced more rapidly with a rate of (0.061 PgC·yr⁻¹) under the AOA measurement than those without AOA measurements (0.048 PgC·yr⁻¹). After 2080, however, the net sink of the ocean reduced at nearly equal rates of 0.012 to 0.014 PgC·yr⁻¹.

We also plotted the temporal evolution of the carbonate and bicarbonate systems, along with the DIC and Revelle factor, as shown below (Figure 5). The Revelle factor is considered a buffer factor that quantifies the response of ocean chemistry to changes in DIC and alkalinity [30]. The result shows that the global mean content of CO₃²⁻ gradually decreases (Figure 5b) along with increasing HCO₃⁻ (Figure 5c) and the Revelle factor (Figure 5d), indicating that CO₃²⁻ is continuously converted into HCO₃⁻ under the forcing of enormous atmospheric CO₂. With an ever-rising atmospheric CO₂, the buffer capacity of the ocean for atmospheric CO₂ will gradually decrease both in the SSP5-8.5 and the AOA scenarios. As for the time node at the year around 2080 (Figure 4), it is difficult to explain whether it is a “tipping point” for oceanic CO₂ uptake or merely a representation of internal oscillation of the oceanic carbon system, just based on the current data we obtained from the CMIP6 esm-ssp585-ocn-alk experiment.

Did the addition of 0.14 Pmol TA every year of the CMIP6 protocols lower the atmospheric CO₂ concentration under the high SSP5-8.5 scenario? Our results suggested that the effect of AOA did exist, but it was quite small. Figure 6 illustrates that the concentrations of atmospheric CO₂ increased continuously both for the no-AOA esm-ssp585 (blue) and the AOA esm-ssp585-ocn-alk (orange) from 2015 to the end of this century. It has been mentioned above that the SSP5-8.5 is a very high greenhouse gas emissions scenario, where the concentration of atmospheric CO₂ is estimated to be as high as more than 1100 μatm in 2100. Hence, mainly controlled by the ever-increasing carbon emission, the globally averaged atmospheric CO₂ concentration rose rapidly and exceeded 1000 μatm around the year 2100. The AOA project lowered the increase rate by about 0.5 μatm·yr⁻¹ (7.28 minus 7.78). Hence, in simple terms, the supplement of alkalinity material should be increased substantially if we are to realize the purpose of lowering the atmospheric CO₂ concentration under the high SSP5-8.5 scenario.

3.4. Effects of AOA on the Ocean Interior, Take Ω for an Example

By increasing the TA, the buffering capacity of seawater is enhanced, so that the seawater’s pH and CaCO₃ saturation degree should increase. However, the relevant results in Section 3.2 indicated that, controlled by the ever-increasing carbon emission, the surface pH and Ω actually decreased, e.g., (Figure 2c,d). Due to the pressure effect, the saturation degree of CaCO₃ usually decreases with the depth of seawater. When Ω is above 1, CaCO₃ tends to precipitate, and when below 1, CaCO₃ tends to dissolve. Therefore, the concentration of Ω was basically >1 in the surface ocean, but it could be <1 in the deep layer. In this section, we applied the “minimum depth of aragonite undersaturation in seawater”, a parameter simulated by the NorESM2-LM, as an appropriate index to evaluate the effect of AOA on the ocean interior.

Firstly, we calculated the globally mean values of the minimum depth of aragonite undersaturation and presented them in Figure 7. The results show that the minimum depth of aragonite undersaturation would decrease rapidly under the high SSP5-8.5 scenario, regardless of whether AOA was implemented or not. The implementation of AOA merely lowers the decreasing rate for about 0.26 m·yr⁻¹ (−5.04 minus −5.30, in Figure 7). In the year 2100, global oceans deeper than about 3900 m would be undersaturated with aragonite, implying a growing ocean acidification under the high SSP5-8.5 scenario. As a result, Figure 7 suggests that the effect of AOA did exist in deep oceans, but was quite small for the mitigation of ocean acidification.

4. Conclusions

Artificial ocean alkalinization (AOA) is one of the most promising marine carbon dioxide removal (mCDR) technologies, which has an enormous theoretical sequestration potential for atmospheric CO₂. In this article, we applied the results of the Earth system model (ESM) from the Sixth Coupled Model Intercomparison Project (CMIP6) to characterize the temporal and spatial variabilities of marine carbon chemistry under the implementation of AOA. Our research has shown that the marine carbon chemistry system varied widely under the implementation of AOA, but some effects were covered by the forcing of a high-carbon emission scenario, SSP5-8.5. Based on the CMIP6 protocols, the AOA experiment adds 0.14 Pmol TA to the upper ice-free ocean surface waters between 70°N and 60°S every year from 2015 to 2100, and could promote an increase in the DIC, a delayed rise in pCO₂, and restrained aggravation of pH and Ω, to various extents, actually. The esm-ssp585-ocn-alk experiment of the CMIP6 was able to simulate the overall variation pattern of the marine carbon system during AOA. The temperate oceans in both hemispheres were the most significantly impacted basins, whereas the Southern Ocean was the less affected region. In the present century, the oceanic carbon sink will intensify until around 2080 and then slow down. The implementation of AOA merely changed the relative strength of the oceanic sink rather than the variation pattern. Our results also suggested that the effect of AOA on the atmospheric CO₂ concentration under the high SSP5-8.5 scenario did exist, but was quite small, so the concentrations of atmospheric CO₂ increased continuously both for the no-AOA esm-ssp585 and the AOA esm-ssp585-ocn-alk from 2015 to the end of this century. In a similar way, the effect of AOA on the ocean interior did exist, but was quite small for the mitigation of ocean acidification under the high SSP5-8.5 scenario. In the year 2100, the global ocean shall be deeper by about 3900 m and will be undersaturated with aragonite, implying the growth of ocean acidification.