Considerations for hypothetical carbon dioxide removal via alkalinity addition in the Amazon River watershed

The Amazon River plume plays a critical role in shaping the carbonate chemistry over a vast area in the western tropical North Atlantic. We conduct a sensitivity analysis of hypothetical ocean alkalinity enhancement (OAE) via quicklime addition in the Amazon River watershed, examining the response of carbonate chemistry and air–sea carbon dioxide flux to the alkalinity addition. Through a series of sensitivity tests, we show that the detectability of the OAE-induced alkalinity increment depends on the perturbation strength (or size of the alkalinity addition, ΔTA) and the number of samples: there is a 90 % chance to meet a minimum detectability requirement with ΔTA>15 µmol kg⁻¹ and sample size >40, given background variability of 15–30 µmol kg⁻¹. OAE-induced pCO₂ reduction at the Amazon plume surface would range between 0–25 µatm when ΔTA=20 µmol kg⁻¹, decreasing with increasing salinity (S). Adding 20 µmol kg⁻¹ of alkalinity at the river mouth could elevate the total carbon uptake in the Amazon River plume ($\mathrm{15}<S<\mathrm{35}$) by at least 0.07–0.1 Mt CO₂ per month, and a major portion of the uptake would occur in the saltiest region (S>32) due to its large size, comprising approximately 80 % of the S>15 plume area. However, the lowest-salinity region (S<15) has a greater drop in surface ocean partial pressure of CO₂ (pCO₂^sw) due to its low buffer capacity, potentially allowing for observational detectability of pCO₂^sw reduction in this region. Reduced outgassing in this part of the plume, while more uncertain, may also be important for total additional CO₂ uptake. Such sensitivity tests are useful in designing minimalistic field trials and setting achievable goals for monitoring, reporting, and verification purposes.

1 Introduction

To meet the Paris Agreement goal of limiting global temperature change to well below 2 ^∘C (UNFCCC, 2015), reducing greenhouse gas emissions is urgently needed but insufficient on its own. Modeling results from the IPCC (2022) estimated a total remaining carbon budget of less than 500 Gt CO₂ for a >50 % probability of staying below 1.5 ^∘C warming by 2100. Staying within this small remaining budget is a formidable challenge, as it requires current emission rates, which totaled ∼36 Gt CO₂ yr⁻¹ in 2021 (Friedlingstein et al., 2022), to rapidly approach zero. Accordingly, even very optimistic emission reduction scenarios assume carbon dioxide removal (CDR) will be needed to remove 10–20 Gt CO₂ yr⁻¹ from the atmosphere by the end of the century (NASEM, 2019, 2021). With the ocean covering ∼70 % of the Earth's surface and providing the largest sink for anthropogenic CO₂ emissions to date, there is growing interest in CDR solutions in the marine environment.

Several ocean-based CDR approaches have been suggested over the past decades to reduce CO₂ in the atmosphere (NASEM, 2021). Enhanced weathering (EW) and ocean alkalinity enhancement (OAE) are related techniques with the primary goal of accelerating the weathering process that would naturally remove atmospheric CO₂ at a very slow rate (10 000 to 100 000 years; Gonzalez and Ilyina, 2016). EW techniques involve pulverization of weatherable rocks and their application on land, while OAE applies those materials to increase alkalinity at the ocean surface (Kheshgi, 1995; Renforth and Henderson, 2017). When mineral additions are made near the edge of a coastal watershed, the weathering process may happen on land and/or in the ocean, and the distinction between the two techniques blurs.

EW and OAE can have other co-benefits besides reducing CO₂ and mitigating global warming. Several early EW experiments via regional-scale liming focused on mitigating acidification in inland lakes and ambient streams. These studies found that the surface water pH could be elevated to desired levels and persist for years without causing deleterious effects to the environment (e.g., Wright, 1985; Porcella, 1989; Driscoll et al., 1996). EW deployments in agricultural settings have also been proven effective in reducing soil acidity, preventing soil erosion, and enhancing crop yields (Caires et al., 2006; Köhler et al., 2010; Kantzas et al., 2022). Similarly, OAE can increase the pH of seawater, alleviating ocean acidification, which is a major stressor for the marine ecosystem (e.g., Doney et al., 2009). Both EW and OAE approaches are still in early phases of conceptualization, and most OAE studies so far have focused on numerical simulations or efforts within laboratories (Köhler et al., 2010; Gonzalez and Ilyina, 2016; Moras et al., 2022; Wang et al., 2022). Understanding the feasibility, effectiveness, and ecological risks of OAE is required before any large-scale efforts should be implemented (NASEM, 2021). Though harmful ecosystem effects associated with highly elevated alkalinity cannot be ruled out (Bach et al., 2019), this risk must be weighed against a counterfactual in which carbon dioxide remains in the atmosphere.

Large river-dominated tropical oceans are potential test grounds for OAE for several reasons. First, mixing and subduction of surface waters into the ocean interior is minimized in tropical oceans relative to higher latitudes (Gonzalez and Ilyina, 2016; Lenton et al., 2018). Second, large rivers form surface plumes that extend thousands of kilometers offshore (Lentz and Limeburner, 1995; Coles et al., 2013). In combination, these two factors mean that added alkalinity would have a long time to absorb CO₂ at the surface ocean and impact a vast area along the plume path. During this time, the atmospheric CO₂ is continuously sequestered by the ocean until a new air–sea CO₂ equilibrium is reached. The plumes also have a heightened potential for observational tracking using surface salinity, which can be estimated from satellite observations. Finally, the carbonate-poor river waters (relative to ocean water) may help suppress secondary chemical precipitation of the added alkalinity, a risk that can reduce the efficiency of OAE (Bach et al., 2019; Hartmann et al., 2023). Overall, the transport of alkalinity in rivers to the ocean from EW deployments is seen as a CDR technique with the potential to scale to the gigaton level (Zhang et al., 2022).

Therefore, we examine the Amazon River–ocean continuum for its potential as a site of OAE. As the world's largest river by volume, the Amazon River represents ∼20 % of the global riverine discharge into the oceans (Salati and Vose, 1984). Its massive outflow (an average of ∼0.2 Sv, sverdrup; Fig. S1 in the Supplement) creates a thin surface layer of low-salinity (Fig. 1) and low-carbonate plume, extending up to 1.5×10⁶ km² at the ocean's surface (Molleri et al., 2010). As a result, the Amazon River plume has a profound influence on the carbonate dynamics and atmospheric CO₂ sequestration throughout the western tropical North Atlantic Ocean (Ternon et al., 2000; Cooley et al., 2007; Lefèvre et al., 2010; Ibánez et al., 2015; Mu et al., 2021; Olivier et al., 2022; Monteiro et al., 2022).

Figure 1 Sea surface salinity (SSS) for the Amazon River plume (15^∘ N–3^∘ S, 43–60^∘ W) during (a) September 2011 and (b) July 2012, derived from a remotely sensed diffuse attenuation coefficient at 490 nm. Area likely affected by the Orinoco River plume (<150 km off the coastline between 55–60^∘ W) is excluded from this study. Oceanic regions in light blue on the maps indicate that the satellite-derived SSS data are unavailable, mostly due to muddy nearshore waters or clouds. See Mu et al. (2021) for further details.

In this study of hypothetical alkalinity addition at the Amazon River mouth, we investigate the expected changes in air–sea CO₂ flux at the offshore Amazon plume waters and examine what perturbation size and sampling density would be needed to measure the alkalinity change and verify a resultant anomalous CO₂ flux. We make use of the knowledge from previous field studies of the carbonate chemistry in the Amazon River plume to address two main goals: (1) analyze the potential for detectability of OAE-induced alkalinity change relative to measurement precision, background variability, and sample size and (2) estimate changes in ocean pCO₂ and air–sea CO₂ flux in the Amazon River plume due to hypothetical alkalinity addition at the river mouth. Our effort aims to outline how one might consider the measurement, reporting, and verification (MRV) needed in the context of known background variability. We argue that this kind of sensitivity analysis is a first step that could lead to more realistic numerical simulations if the system does not fail basic tests of feasibility.

2 Method

2.1 Study site and mixing model

We use a river–ocean conservative mixing model (Cooley and Yager, 2006) informed by direct observations collected as parts of the ANACONDAS (Amazon iNfluence on the Atlantic: CarbOn export from Nitrogen fixation by DiAtom Symbioses) and ROCA (River Ocean Continuum of the Amazon) projects (Mu et al., 2021, 2023) to establish expectations of how alkalinity perturbations would influence the carbonate system in the Amazon plume (the gray and green pathways on the methods schematic shown in Fig. 2). The mixing model assumes no sources or sinks of the carbonate species modeled, and the deviations from this assumption are discussed in Sect. 2.2. In this model, dissolved inorganic carbon (DIC) and total alkalinity (TA) are treated as conservative tracers and used to describe pCO₂ variations as a function of salinity and temperature (Fig. 1). TA and DIC can be, respectively, expressed as

$$\begin{array}{}\text{(1)}& {\displaystyle }\begin{array}{rl}\text{TA}=& \left[{\mathrm{Na}}^{+}\right]+\mathrm{2}\left[{\mathrm{Mg}}^{\mathrm{2}+}\right]+\mathrm{2}\left[{\mathrm{Ca}}^{\mathrm{2}+}\right]+\mathrm{\cdots }\\ & -\left[{\mathrm{Cl}}^{-}\right]-\mathrm{2}\left[{{\mathrm{SO}}_{\mathrm{4}}}^{\mathrm{2}-}\right]-\left[{\mathrm{Br}}^{-}\right]-\mathrm{\cdots }\end{array},\text{(2)}& {\displaystyle }\text{DIC}=\left[{{\mathrm{HCO}}_{\mathrm{3}}}^{-}\right]+\left[{{\mathrm{CO}}_{\mathrm{3}}}^{\mathrm{2}-}\right]+\left[{\mathrm{CO}}_{\mathrm{2}}\right].\end{array}$$

We adopt the explicitly conservative form of TA here (Wolf-Gladrow et al., 2007) so that TA changes due to the addition of our hypothetical calcium-based alkalinity feedstock can be fully tracked by the increase in calcium ion concentration. To establish the baseline condition of the carbonate system in the Amazon River–ocean continuum, we adopted the river endmembers from Mu et al. (2023; also Table 1) and ocean endmembers (S=36, TA=2369 µmol kg⁻¹, DIC=2025 µmol kg⁻¹) from Mu et al. (2021). The conservative mixing model can be further expressed as

$$\begin{array}{}\text{(3)}& {\displaystyle }{\text{SSS}}_{\text{mix}}=S_{\mathrm{r}}\times f_{\mathrm{r}}+{\text{SSS}}_{\mathrm{o}}\times f_{\mathrm{o}},\text{(4)}& {\displaystyle }{f}_{\mathrm{r}}+f_{\mathrm{o}}=\mathrm{1}.\end{array}$$

SSS_mix is the salinity for a plume sample, S_r is the salinity of the river endmember (S_r=0), and SSS_o is the surface salinity of the ocean endmember. In Eq. (4), f_r and f_o are the proportions of river and ocean in the sample, both of which can be solved when SSS_mix is known. The theoretical TA and DIC in the plume, TA_mix, and DIC_mix at any given SSS_mix are then calculated from the known endmember properties and proportions via

$$\begin{array}{}\text{(5)}& {\displaystyle }{\text{TA}}_{\text{mix}}={\text{TA}}_{\mathrm{r}}\times f_{\mathrm{r}}+{\text{TA}}_{\mathrm{o}}\times f_{\mathrm{o}},\text{(6)}& {\displaystyle }{\text{DIC}}_{\text{mix}}={\text{DIC}}_{\mathrm{r}}\times f_{\mathrm{r}}+{\text{DIC}}_{\mathrm{o}}\times f_{\mathrm{o}}.\end{array}$$

Eventually, pCO₂^mix and pH_mix at a given salinity and observed temperature are calculated with CO2SYS (Lewis et al., 1998) from inputs of TA_mix and DIC_mix. In this calculation we use the carbonic acid dissociation constants K1 and K2 from Millero (2010) due to their eligibility across a wide salinity range suitable for this study.

Figure 2 Schematic diagram for the method used to calculate carbonate chemistry and air–sea CO₂ fluxes before and after a hypothetical alkalinity addition (OAE) in the Amazon River–ocean continuum. The baseline and OAE pathways (gray and green) use the conservative mixing model (Eqs. 3–6) and CO2SYS to calculate the baseline and hypothetical OAE pCO₂. The difference between the baseline and OAE mixed models (${\mathrm{\Delta }}_{\text{OAE}}{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix}}={p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix\_OAE}}-{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix\_base}}$) is added to the empirically derived pCO₂ at every SSS. The resulting ${p{\mathrm{CO}}_{\mathrm{2}}}^{\text{OAE}}$ is used to calculate the air–sea CO₂ flux across the surface of the plume for each hypothetical OAE experiment.

Table 1 Baseline total alkalinity (µmol kg⁻¹) determined at the N. Macapá gateway at the Amazon River mouth during September 2011 and July 2012 (Mu et al., 2023). The means represent the river's alkalinity endmember in that month.

We confined our study periods to two specific months, September 2011 and July 2012, during which measurements were made in both the river and throughout the Amazon River plume. During the river plume expeditions, surface seawater pCO₂, temperature, salinity, and chlorophyll fluorescence were continuously measured shipboard across the Amazon-influenced regions. An oceanic station was specifically sampled for TA and DIC during each cruise as the ocean carbonate endmember for each corresponding month (Mu et al., 2021). Details in quantifying the river endmembers are described in Ward et al. (2015) and Mu et al. (2023). Briefly, the Amazon River mouth region receives freshwater runoff from three channels: the north and south channels near Macapá and the channel of the Tocantins River near Belém (Fig. S2 in the Supplement). Sampling was conducted at the lower reaches of the three channels near the mouth (or “gateways”) in each month. Six TA and DIC samples were collected at three cross-channel locations of each gateway (left bank, center, and right bank) from both the surface and 50 % depth (5–10 m). For a concise demonstration on the detectability of the alkalinity addition in the river, we chose to target one gateway (North Macapá) as the perturbation site and consider how many samples would be needed to observe the TA perturbation relative to that gateway's background variability. However, to quantify the potential CDR induced by a TA perturbation just above detectability, we assumed the entire river mouth would be perturbed equally as in North Macapá. Because the outflow through both North and South Macapá gateways comprises >80 % of total discharge (Ward et al., 2015; Mu et al., 2023) and their chemical properties are almost identical seasonally (Mu et al., 2023), North Macapá is representative of the mouth's chemical composition. In an actual OAE field trial, perturbing one gateway will perturb the mouth region by an amount proportional to the percentage discharge there (relative to the combined discharge), and the natural variability in the unperturbed gateways will affect the background variability at the mouth. There are key unknowns in seasonal TA contributions from other gateways to the mouth which prevent us from combining all the gateways. Therefore, the mean and standard deviation of the six TA values at North Macapá, respectively, were then used as the river TA endmember and its natural background variability (Table 1).

2.2 Deviations from the conservative mixing model

The assumption that TA and DIC act as conservative tracers is not true in the presence of photosynthesis, respiration, bio-calcification, abiotic dissolution/precipitation of calcium carbonate, and/or air–sea gas exchange. In fact, plume pCO₂ values measured from an underway system on the research vessels surveying for the ANACONDAS program were systematically lower than predicted from the mixing model (Mu et al., 2021; Fig. S3 in the Supplement). This offset was assumed to be due to net community production, an assumption supported by the strong correlation between shipboard chlorophyll and the difference between the pCO₂^mix calculated from conservative mixing and the measured pCO₂ (Mu et al., 2021).

A linear regression between measured SSS ($\mathrm{15}<\text{SSS}<\mathrm{35}$) and measured pCO₂ showed a strong linear relationship (r=0.79 for September 2011 and r=0.97 for July 2012). Therefore, we use SSS to calculate an empirical estimate of pCO₂ (pCO₂^empirical) at every SSS level in each month (orange pathway in the schematic shown as Fig. 2). SSS fields in the Amazon plume were derived from the remotely sensed diffuse attenuation coefficient at 490 nm (K_d490, Fig. 1). The measured SSS vs. pCO₂^empirical regression is used to map pCO₂^empirical at every satellite-derived SSS value in the plume (as in Fig. 3a) and to calculate the empirical air–sea CO₂ flux, according to the following equation:

$$\begin{array}{}\text{(7)}& {\text{Flux}}_{\text{empirical}}=k\mathit{\alpha }({p{\mathrm{CO}}_{\mathrm{2}}}^{\text{empirical}}-{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{atm}}),\end{array}$$

where k is the gas exchange coefficient calculated from gridded reanalysis wind speed and the parameterization of Sweeney et al. (2007), α is the solubility of CO₂ in seawater (a function of water temperature and salinity, calculated from equations in Weiss, 1974), and pCO₂^atm is the monthly averaged pCO₂ in the atmosphere measured at a nearby NOAA monitoring station. Details in calculating the empirical CO₂ flux can be found in Mu et al. (2021).

Figure 3 pCO₂ at the surface of the Amazon River plume and its predicted response to alkalinity addition based on satellite-derived SSS. (a) Spatial distribution of unperturbed pCO₂^empirical (µatm) at the surface of the Amazon River plume ($\mathrm{15}<\text{SSS}<\mathrm{35}$) during September 2011, with SSS outside of the plume range removed from this map as the remotely sensed K_d490 vs. SSS regression is sufficiently robust only when $\mathrm{15}<\text{SSS}<\mathrm{35}$ (see Fig. 2 and Sect. 2.2 for details on how the mapped quantity is defined and Mu et al., 2021, for additional details). (b) Predicted changes in the plume surface pCO₂ (i.e., Δ_OAEpCO₂^mix from Eq. 8) due to addition of 20 µmol kg⁻¹ of alkalinity at the Amazon River mouth during September 2011.

2.3 Experimental design

The core of this sensitivity analysis is to manipulate the concentrations of TA at the river mouth by the addition of a hypothetical alkalinity feedstock. There are several feedstocks that could be considered for this purpose, each with a distinct set of considerations (Hartmann et al., 2013; Renforth and Henderson, 2017). Magnesium- or silicate-rich rocks such as olivine and other basalts (Hartmann et al., 2013), carbonate minerals (e.g., Rau, 2011; Renforth et al., 2022), mineral derivatives found in some industrial waste products, or even electrochemical production of alkalinity (Tyka et al., 2022; He and Tyka, 2023) are all candidates (NASEM, 2021). The stoichiometry of the reaction, dissolution rate, risk of secondary precipitation, and contamination with metals are all specific to the particular alkalinity source chosen. Quicklime (CaO) has been identified as being promising due to its high solubility in seawater and rapid dissolution rates (Kheshgi, 1995; Moras et al., 2022). Quicklime is made by heating limestone (primarily CaCO₃), an abundant rock at the Earth's surface. The heating of each molecule of CaCO₃ releases one molecule of quicklime and one molecule of CO₂ (Kheshgi, 1995). Therefore, carbon capture and storage of the released CO₂ would be necessary for OAE to be effective from the application of quicklime (Paquay and Zeebe, 2013; Renforth et al., 2013; NASEM, 2021; Moras et al., 2022). Hereafter, we assume that the alkalinity feedstock for the hypothetical OAE deployment is quicklime, which simplifies the discussion of stoichiometry, precludes contamination from metals and other industrial contaminants, justifies our assumption of rapid dissolution, and informs our discussion of potential unintended consequences.

Assuming the additional CaO converts 100 % to TA according to stoichiometry, the dissolution of CaO consumes CO₂ and releases ${{\mathrm{HCO}}_{\mathrm{3}}}^{-}$ through

$$\begin{array}{}\text{(R1)}& (\mathrm{CaO}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\rightleftharpoons \mathrm{Ca}(\mathrm{OH}{)}_{\mathrm{2}})+\mathrm{2}{\mathrm{CO}}_{\mathrm{2}}\rightleftharpoons {\mathrm{Ca}}^{\mathrm{2}+}+\mathrm{2}{{\mathrm{HCO}}_{\mathrm{3}}}^{-}.\end{array}$$

For every mole of additional CaO, TA is increased by 2 mol (Eq. 1), while DIC remains constant (i.e., consuming 2 mol of CO₂ while producing 2 mol of bicarbonate in Eq. R1) before any significant air–sea equilibration occurs. This assumption of a constant DIC during perturbation should hold over the timescale of the river–ocean mixing based on the following evidence: (1) CaO dissolution happens on hour scales (Moras et al., 2022), and (2) the time elapsed between the river outflow at the mouth and its mixing with ocean to relatively high salinity (SSS>15) is as short as 2 weeks (see Fig. 5 from Coles et al., 2013), while the air–sea CO₂ equilibration usually takes several weeks to months in the western tropical North Atlantic (Jones et al., 2014). We assume in this study that there is no secondary precipitation of CaCO₃ and that the additional CaO stays at the surface layer until fully dissolved. Another underlying assumption is that CaO is homogeneously mixed across the river mouth gateway, eliminating potential uncertainty/variability due to patchiness of the TA addition. We will address key unknowns associated with these assumptions in the Discussion.

We explore TA perturbations over the range 1–100 µmol kg⁻¹ (i.e., perturbation between 0.07 and 7 times the natural variability for September 2011 and 0.04 and 4 times the natural variability for July 2012). The mixing model is recalculated with each of the perturbed river endmembers to find the perturbed TA_mix and DIC_mix across the plume. We then use CO2SYS to calculate the theoretical pCO₂^mix_OAE at each salinity stamp with the perturbed TA_mix, DIC_mix, and observed temperature as inputs to compare to the unperturbed baseline. The difference of the TA-enhanced seawater pCO₂ and the baseline pCO₂ is

$$\begin{array}{}\text{(8)}& {\mathrm{\Delta }}_{\text{OAE}}{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix}}={p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix\_OAE}}-{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix\_base}}.\end{array}$$

In each perturbation scenario, the added alkalinity lowers the pCO₂^mix_OAE below the unperturbed mixing curve by the amount of Δ_OAEpCO₂^mix. To finally arrive at the surface ocean pCO₂ needed to calculate the air–sea fluxes according to Eq. (7), we add the Δ_OAEpCO₂^mix to pCO₂^empirical:

$$\begin{array}{}\text{(9)}& {p{\mathrm{CO}}_{\mathrm{2}}}^{\text{OAE}}={\mathrm{\Delta }}_{\text{OAE}}{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{mix}}+{p{\mathrm{CO}}_{\mathrm{2}}}^{\text{empirical}}.\end{array}$$

This approach (schematized in Fig. 2) implicitly assumes that the biological processes that lower the measured pCO₂^empirical below the theoretical pCO₂^mix would be unchanged from the deployment of additional alkalinity, an assumption that would demand careful testing in the field.

2.4 Sensitivity evaluation for detectability of hypothetical OAE deployments

To assess the TA detectability, we evaluate perturbations in one river gateway (North Macapá). We use a simple bootstrapping technique (or Monte Carlo simulation) to assess the minimum TA perturbation that would be detectable with a given number of measurements, taking into consideration the background variability during a given season and reasonable measurement precision.

We randomly generate 1000 baseline TA values under a normal distribution using the mean and standard deviation of the six measured TA values in September 2011 or July 2012. We then generate 1000 perturbed TA values for each perturbation scenario under a normal distribution, where the mean is the addition of the baseline mean and the perturbation strength, with the same standard deviation as the baseline. The range for the tested perturbation sizes is 1–20 µmol kg⁻¹ of TA. Within the 1000 TA pools from both the baseline and a perturbation scenario, we randomly select 1–100 values (i.e., varying sample size) from both pools, perform the student's t test between the selected sets, and calculate the p value to determine if the perturbed TA set would be seen as significantly different from baseline. The t test is repeated 200 times for each perturbation size and sample size, generating 200 p values. The mean of those p values is calculated for each perturbation scenario at each sample size (Fig. 4). A lower p value suggests a higher likelihood that the perturbed TA values would be considered significantly different from the baseline and therefore “detectable”. We use a threshold of p<0.1 to indicate statistical significance.

Figure 4 Detectability of hypothetical TA enhancement relative to background variability. (a and b) The theoretical shift of TA (µmol kg⁻¹) means and distributions for September 2011 and July 2012 due to the addition of 20 µmol kg⁻¹ of TA at the Amazon River mouth. Black lines indicate the in situ TA measurements at the mouth on which the baseline data distributions are based. The standard deviations in the perturbation scenarios are assumed the same as those from the baseline. (c and d) The p-value maps from t tests performed between the baseline and TA perturbation scenarios at various sample sizes and TA perturbation strengths for September 2011 and July 2012. Areas in yellow (p<0.1) indicate conditions where an analyst would conclude that the TA perturbation was detected relative to the baseline condition with 90 % certainty.

3 Results

3.1 Baseline carbon chemistry in the plume

The surface pCO₂^empirical in the Amazon River plume derived from remotely sensed SSS generally ranged from 210–400 µatm in September 2011 (Fig. 3a), with the lowest values in low-salinity regions near the French Guiana coast, which increase as the river mixes with ocean water (Fig. 1a). Figure 3a masks out regions with surface salinity below 15 and beyond 35 psu, as these waters are out of the range of a sufficiently robust algorithm between K_d490 and SSS (Mu et al., 2021).

The distribution of plume surface pCO₂^sw is primarily shaped by two processes: the Amazon River–ocean mixing and biological CO₂ consumption/release in the plume waters (Mu et al., 2021). For example, zero-salinity river water near the mouth is supersaturated with respect to atmospheric CO₂ in July 2012, where pCO₂>1000 µatm is observed (Mu et al., 2021, 2023) due to low carbonate buffer capacity and high microbial respiration. As the river waters mix with the ocean, the sharp increase in buffer capacity and shift towards a net autotrophic state lowers the surface pCO₂ towards a minimum (below pCO₂^atm in July 2012; Mu et al., 2021) before rising towards the open ocean levels closer to equilibrium with the atmosphere (i.e., pCO₂∼400 µatm) at higher salinity (Fig. S3). The nitrogen-fixing diatom–diazotroph assemblages and other phytoplankton that are active in the Amazon plume waters (Goes et al., 2014) further enhance the CO₂ undersaturation in mid- to mid-high-salinity portions of the plume (i.e., $\mathrm{15}<\text{SSS}<\mathrm{33}$) on top of the undersaturated CO₂ state caused by conservative mixing (Mu et al., 2021).

3.2 Detectability of TA perturbations

We propose that the minimum requirement for MRV in a watershed OAE experiment is that the perturbed TA in the river can be detected above background variability. To illustrate the challenge of detectability, we consider the background TA variability in the North Macapá gateway in September 2011, with the assumption that the variability measured at that time (standard deviation of 14.5 µmol kg⁻¹) is representative of that season and gateway. It is intuitive that a large enough perturbation (commensurate with the standard deviation) is detectable with a small number of samples (Fig. 4), and a small perturbation (a factor of 3 smaller than the standard deviation or smaller) cannot be detected against the background variability regardless of the number of samples. This exercise reveals the challenge of balancing the effort needed to detect the perturbation against the obstacle and risk of increasing the TA by a large amount.

As expected and illustrated in Fig. 4, higher sample sizes and TA perturbations lead to greater detectability of the alkalinity enhancement. For practical purposes, neither sample size nor perturbation strength can be increased infinitely, so OAE experiments would likely seek a balance between the two. For example, in September 2011 (Fig. 4a and c), a TA perturbation of ∼7–10 µmol kg⁻¹ at a background standard deviation of 14.5 µmol kg⁻¹ would have been readily detectable with 40 samples. For July 2012 (Fig. 4b and d), when background variability was higher (25.7 µmol kg⁻¹), 40 samples would detect perturbations only if they were to exceed +20 µmol kg⁻¹ of TA. In an actual field OAE trial, each season (and gateway) may have different background variability that would require measurement and characterization in advance of any perturbation to have a clear strategy for sampling.

Table 2 Summary table for air–sea CO₂ flux in the Amazon River plume ($\mathrm{15}<\text{SSS}<\mathrm{35}$) and the mass of CaCO₃ mineral required for each TA perturbation size (ΔTA) in each month. Positive fluxes indicate CO₂ outgassing, and negative fluxes (–) indicate ocean CO₂ uptake. Note that the post-perturbation CO₂ flux and CaCO₃ demand increase linearly with increase in ΔTA.

3.3 Impact on pCO₂ and air–sea CO₂ flux

Figure 3b shows the pCO₂^mix decrease due to a hypothetical TA addition in the Amazon watershed, which is highest in the low-salinity waters near the river mouth and attenuates at higher salinities due to mixing with the unperturbed ocean water. Figure 5 provides a quantification of the air–sea CO₂ flux at each salinity level for different TA perturbation strengths (20, 50, and 100 µmol kg⁻¹). While the air–sea CO₂ flux per unit area (or flux density) is greatest at low salinities (Fig. 5a and b), the large area occupied by the diluted plume (salinities greater than 32; Fig. S4 in the Supplement) means that more than half of the total integrated air–sea CO₂ exchange – as well as the perturbation to this flux – would occur in this salty part of the plume.

Figure 5 The impact of TA addition on ocean uptake of CO₂ across the Amazon River plume. (a and b) Air–sea CO₂ flux density ($\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$) in the Amazon River plume at different sea surface salinities for the baseline and multiple TA perturbation scenarios in (a) September 2011 and (b) July 2012. Standard deviations of the fluxes within each salinity band (1 psu apart) are represented by the shaded areas. Negative values indicate atmospheric CO₂ sinks. (c and d) Cumulative ocean CO₂ uptake (Mt CO₂ per month) in the Amazon plume due to TA additions in (c) September 2011 and (d) July 2012. SSS ranging between 15 and 35 is used, while SSS<15 and >35 is omitted because the remotely sensed K_d490 vs. SSS regression is sufficiently robust only when $\mathrm{15}<\text{SSS}<\mathrm{35}$ according to the Amazon River plume definition from Mu et al. (2021).

Air–sea CO₂ exchange at SSS=34–35 for July 2012 is only slightly greater than 0 (Fig. 5b), but due to the large area of plume at near-oceanic salinity level, the total plume CO₂ uptake (i.e., negative flux) in lower S regions is entirely offset by the CO₂ outgassing at $\mathrm{34}<\text{SSS}<\mathrm{35}$. Regardless of whether the baseline Amazon plume is a carbon sink or source, the change in air–sea CO₂ exchange due to OAE shifts toward more carbon storage by the ocean and increases linearly with the size of TA perturbation (Table 2).

4 Discussion

4.1 Unaccounted for plume region with great CDR potential

While we explored the OAE-induced additionality of the carbon uptake (or net CDR) at different salinities in the Amazon River plume (Fig. 5), it is important to note that the cumulative additional uptake in Fig. 5b excludes the freshest part of the plume (SSS<15). Due to high organic carbon remineralization in shallow waters (Mu et al., 2021), the quasi-linear SSS versus pCO₂ empirical relationship collapses at SSS<15 and prevents the establishment of empirical pCO₂ in this region. Therefore, we excluded SSS<15 in our analyses (e.g., in Fig. 3, coastal areas <50 m deep near the mouth are masked) while acknowledging that because of the strong CO₂ outgassing in this region, the baseline air–sea CO₂ flux in the entire plume ($\mathrm{0}<\text{SSS}<\mathrm{35}$) will be shifted towards more CO₂ release, should the SSS<15 water be included. However, once TA is added, surface pCO₂ will decline in the entire plume regardless of salinity, causing greater net CO₂ drawdown across the plume. In other words, the OAE-induced CO₂ uptake increase calculated for $\mathrm{15}<S<\mathrm{35}$ plume water could substantially underestimate the true additionality when SSS<15 is not considered. The area of the S<15 water is small compared to the $\mathrm{15}<\text{SSS}<\mathrm{35}$ plume (i.e., <15 % of the SSS>15 plume area), but its lower buffer capacity also means pCO₂ is much more sensitive to TA addition, and therefore could still contribute a considerable amount to CDR.

A basic scaling would suggest that 10 % of the plume area with an OAE-induced surface pCO₂ decrease of 30 µatm would have as large an impact on area-integrated ocean uptake as the $\mathrm{15}<\text{SSS}<\mathrm{35}$ portion of the plume with a 3 µatm decrease. Therefore, including the freshest part of the plume and calculating its reduction in outgassing could easily double our estimate of additional CO₂ uptake for the entire plume. Lacking both knowledge of the spatial distribution of salinity and a robust estimate for pCO₂ in this low-salinity region prevents us from quantifying its CDR potential in a rigorous manner.

However, we can still conduct a back of the envelope calculation to estimate the distribution of CDR between SSS<15 and $\mathrm{15}<\text{SSS}<\mathrm{35}$ regions of the plume. With an addition of 20 µmol kg⁻¹ TA and an averaged Amazon freshwater discharge of 0.18 Sv (Fig. S1), we can calculate the total transport of this TA addition of 9.3×10⁹ mol per month. Given sufficient time for air–sea re-equilibration and the subsequent DIC increase due to OAE, we can assume the maximum DIC increase would be 0.8 times the ΔTA (Wang et al., 2022). Therefore the maximum CDR resulting from TA addition would be $\mathrm{0.8}\times \mathrm{9.3}\times {\mathrm{10}}^{\mathrm{9}}\mathrm{mol}\mathrm{per}\mathrm{month}=\mathrm{7.5}\times {\mathrm{10}}^{\mathrm{9}}$ mol per month, or 0.3 Mt CO₂ per month. Given that the estimated enhanced CO₂ flux in the S>15 region is approximately 0.07–0.1 Mt CO₂ per month (Table 2), ∼0.2 Mt of CO₂ uptake per month would be expected to take place in the low-salinity region (SSS<15), which is consistent with the estimation based on direct air–sea CO₂ flux change. This calculation highlights the potentially important role of the freshest part of the plume in the net CDR by the entire plume, which could be disproportionately large relative to its small size.

4.2 Exploring the minimum TA detectability

Here we propose that the minimum MRV requirement for an OAE experiment in a river be that the TA increase can be detected in the fresh river endmember, and we showed what size perturbation and sampling density would meet that minimum for the Amazon River. We found that the minimum detectable TA addition at the Amazon River mouth ranges between 10–20 µmol kg⁻¹, with a reasonable sample size of 40 during field campaigns before and during the perturbation (Fig. 4c and d). We focus on detectability results for TA, since that is the variable directly targeted by any OAE approach. However, we present similar detectability analyses for the increase in pH (Fig. S5 in the Supplement) and decrease in pCO₂ in the river mouth (Fig. S6 in the Supplement), which shows that the perturbation to these variables from a 10–20 µmol kg⁻¹ is also readily detectable with a reasonable number of samples. With commercially available autonomous sensors for pH and pCO₂ (Kapsenberg et al., 2017; Sabine et al., 2020), these might be attractive proxies for the added TA.

One may further hope that MRV efforts for OAE deployed at the river mouth would document reductions in plume pCO₂. We rely on the observed pCO₂, a mixing model, and satellite-derived SSS to deduce where one might expect pCO₂ reductions to be measurable relative to background concentrations. Sensors for measuring ${p{\mathrm{CO}}_{\mathrm{2}}}^{\text{sw}}$ aboard ships and uncrewed surface vehicles like Saildrone and Waveglider have a target accuracy of 2 µatm (Sabine et al., 2020). Figure 3 shows that, if Amazon River water TA was continuously increased by 20 µmol kg⁻¹, a broad region of the plume would see pCO₂^sw decreases measurable with current sensor technology if they could sustain regular sampling over the SSS<30 region before, during, and after the OAE deployment.

However, like with the detectability of the TA in the river mouth, the pCO₂ perturbation must be detectable given background variability and does not depend only on sensor accuracy. Here, a measure of that background variability is the root mean square error (RMSE) between the pCO₂^empirical, calculated from the linear regression of the in situ observations of pCO₂ and SSS, and the actual observations of pCO₂ (pCO₂^measured, Fig. 2). For September 2011, the rms error of this regression is ∼22 µatm for $\mathrm{15}<\text{SSS}<\mathrm{35}$ (Fig. S7 in the Supplement). Thus, we would expect that a perturbation approximately this size would be detectable with sufficient samples at the river mouth (Fig. S6). Figures 3 and S7 suggest that a region of the plume near the river mouth approximately $\mathrm{15}<\text{SSS}<\mathrm{20}$ could meet this criterion due to (1) lower salinity and higher perturbed TA and (2) relatively lower rms error in pCO₂^empirical and thus low background pCO₂ variability. This speculation is nicely verified in a minimum OAE detectability test using pCO₂ in the plume waters as the proxy (Fig. S8 in the Supplement), where the reduction in pCO₂ at $\mathrm{\Delta }\text{TA}=+\mathrm{10}\mathrm{\mu }\mathrm{mol}{\mathrm{kg}}^{-\mathrm{1}}$ is readily detectable with 10 samples for $\mathrm{15}<\text{SSS}<\mathrm{20}$ waters (Fig. S8c), contrasting sharply with the poor detectability at higher salinity ($\mathrm{30}<\text{SSS}<\mathrm{35}$; Fig. S8d).

It is encouraging that SSS can be estimated from satellite measurements, allowing for field campaigns – including uncrewed vehicles piloted from operators onshore – to find and sample in the plume where an OAE-induced pCO₂ reduction may be detected. Nevertheless, Fig. 5 shows that nearly half the additional ocean CO₂ uptake due to the OAE experiment could happen in the part of the plume with $\mathrm{30}<\text{SSS}<\mathrm{35}$, where detecting the perturbation in TA and plume surface pCO₂ is impossible with the accuracy of existing sensors or the laboratory analysis of bottle measurements.

Finally, we consider the amount of limestone (CaCO₃) required to produce pulverized quicklime as the alkalinity source material that sustains a set TA perturbation size (Table 2). The limestone equivalents needed to create a detectable perturbation of +20 µmol kg⁻¹ of TA is 0.4 Mt per month using river discharge numbers from September 2011, which is approximately equal to 0.04 % of US industrial limestone production in the year 2021 (USGS, https://pubs.usgs.gov/periodicals/mcs2021/mcs2021-stone-crushed.pdf, last access: 24 May 2023). Using river discharge for July 2012, the requirement grows to 0.7 Mt limestone per month. In terms of OAE-induced additional CO₂ uptake, every 0.1 Mt CO₂ per month represents more than 30 times greater CDR than the CO₂ captured by one ORCA Direct Air Capture plant in a full year (https://en.wikipedia.org/wiki/Orca_(carbon_capture_plant), last access: 24 May 2023), yet amounts to less than what Brazil emitted on an average 2 h basis in 2021 (https://www.icos-cp.eu/science-and-impact/global-carbon-budget/2021, last access: 24 May 2023). It is worth noting that concerns have been raised over the feasibility of CaO liming considering the energy need and CO₂ release associated with the production of this alkalinity source (e.g., Paquay and Zeebe, 2013; Renforth et al., 2013; Voosen, 2022), but limestone calcination coupled with heat recovery and carbon capture and storage technologies is found to significantly reduce energy costs and CO₂ emission during CaO production, potentially making this liming approach more sustainable (Foteinis et al., 2022).

4.3 Constraints on OAE deployments

So far, we have focused most of our efforts on describing the minimum detectable OAE perturbation. It is also important to consider what might constrain the upper limit of TA addition (i.e., through liming), such as ecosystem disturbances and secondary precipitation of calcium carbonate. Ideally, OAE would not put the ambient marine ecosystems at risk. The minimum detectable perturbation described above is close to the background TA variability and so has a greater likelihood for a small impact on the ecosystem relative to larger deployments. Heavy metals in some source alkaline minerals are problematic (NASEM, 2021), a consideration avoided with the use of a 100 % pure CaO compound as the source of TA addition. The highest perturbation we explored was +100 µmol kg⁻¹ of TA added to the river endmember, which would lead to initial pH increase of 2.0 at the river mouth, a pCO₂^sw decrease of up to 100 µatm (data not shown) in the Amazon plume, and a total enhanced atmospheric carbon uptake of ∼0.35 Mt CO₂ per month across the plume ($\mathrm{15}<\text{SSS}<\mathrm{35}$, with considerable additional contribution likely in the freshest plume waters) during September 2011 (Table 2 and Fig. 5c). This level of initial pH increase at the TA addition site is likely concerning for local marine communities, although the effect would be dispersed by the river and diminish with increasing distance from the river due to mixing.

Another constraint on the upper limit of TA addition through ocean liming is secondary CaCO₃ precipitation and the resulting reduction of atmospheric CO₂ uptake. Moras et al. (2022) suggested that runaway CaCO₃ during OAE may be avoidable if the saturation state of aragonite (Ω_A) is kept below 5 for natural seawater. Assuming such a Ω_A threshold for barely avoiding CaCO₃ precipitation is also applicable in fresher Amazon plume and river water, we have calculated the corresponding maximum TA perturbation allowed to keep Ω_A below this threshold according to the method in Mucci (1983), with the results shown in Table 3. Because the Ω_A increases with salinity, TA perturbation that likely avoids secondary CaCO₃ precipitation becomes substantially smaller close to SSS=35 (i.e., ∼100 µmol kg⁻¹, compared to ∼260 µmol kg⁻¹ at S=0). Adding 20 µmol kg⁻¹ of TA in September 2011 would only increase Ω_A by a maximum of 0.3 (Table 3), and Ω_A stays below 5 everywhere in the plume for TA additions of 100 µmol kg⁻¹ or smaller. An important caveat on assuming a Ω_A threshold of 5 is that the Amazon River at the mouth contains heavy loads of mineral particles and will likely facilitate the formation of heterogeneous nucleation and precipitation for CaCO₃ (Renforth and Henderson, 2017; Moras et al., 2022), thereby lowering the maximum TA addition allowed. Further study is needed to ensure Ω_A is always kept below the proper precipitation threshold in all regions of the river plume.

Table 3 Theoretical aragonite saturation state (Ω_A) for different TA perturbation strengths (ΔTA) and salinity levels at the surface of the Amazon River and its plume during September 2011, according to the equation ${\mathrm{\Omega }}_{\mathrm{A}}=\left[{\mathrm{Ca}}^{\mathrm{2}+}\right]\left[{{\mathrm{CO}}_{\mathrm{3}}}^{\mathrm{2}-}\right]/K_{\text{sp}}^{\mathrm{A}(\ast )}$. [Ca²⁺] is calculated by adding the increment of [Ca²⁺] due to TA perturbation (i.e., $\mathrm{\Delta }\text{TA}/\mathrm{2}$) to the unperturbed [Ca²⁺]. The unperturbed [Ca²⁺] term is comprised of relative contributions from both the river and ocean [Ca²⁺] endmembers at a specific salinity. [Ca²⁺] of 0.15 mmol kg⁻¹ is used as the river endmember, approximating results from Drake et al. (2021), and the global ocean average [Ca²⁺] of 10.28 mmol kg⁻¹ is used as the ocean endmember. [${{\mathrm{CO}}_{\mathrm{3}}}^{\mathrm{2}-}$] is derived from theoretical TA and DIC in the river–ocean endmember mixing model (see Sect. 2.1 and Fig. 2) using a constant water temperature of 29 ^∘C (or 302.15 K). The temperature- and salinity-dependent stoichiometric solubility product for aragonite, $K_{\text{sp}}^{\mathrm{A}(\ast )}$, is calculated from Mucci (1983). The following equations in Mucci (1983) are involved in the determination of $K_{\text{sp}}^{\mathrm{A}(\ast )}$ for each salinity: $\log K_{\text{sp}}^{\mathrm{A}(\ast )}-\log K_{\text{sp}}^{\mathrm{A}\left(\mathrm{0}\right)}=(b_{\mathrm{0}}+b_{\mathrm{1}}\text{SSTK}+b_{\mathrm{2}}/\text{SSTK})S^{\mathrm{0.5}}+c_{\mathrm{0}}S+d_{\mathrm{0}}{S}^{\mathrm{1.5}}$ and $\log K_{\text{sp}}^{\mathrm{A}\left(\mathrm{0}\right)}=-\mathrm{171.945}-\mathrm{0.077993}\text{SSTK}+\mathrm{2903.293}/\text{SSTK}+\mathrm{71.595}\log \text{SSTK}$; SSTK is the sea surface temperature in Kelvin (i.e., SSTK=302.15 K); and b, c, and d are constants reported in Mucci (1983).

Adding TA to the land surrounding the Amazon watershed could have possible advantages over spreading across the open ocean, involving co-benefits to the terrestrial ecosystems and agriculture such as increased crop yield in agricultural watersheds (Caires et al., 2006; Hartmann et al., 2013; Beerling et al., 2020), reduced soil runoff (Taylor et al., 2016), and reduction of N₂O production (Beerling et al., 2018). Introducing the TA on land might also slow its dissolution in the river and allow its reaction with CO₂ to form bicarbonate in pore water, helping reduce the risks associated with large perturbations to pH and TA, like secondary precipitation or phytoplankton community perturbations. Basaltic rock with slower dissolution rates, but more favorable stoichiometry from CDR than carbonates (Beerling et al., 2018), might also provide an alternative TA feedstock for land-based deployment. However, deployment on land would likely increase the difficulty to quantify the effect of TA addition for MRV purposes.

4.4 Additional challenges

One of the underlying assumptions for this analysis is that CaO reacting with CO₂ takes place in a time frame faster than air–sea re-equilibration takes to occur, and the layer of plume water dwells at the surface for a relatively long period of time. If the TA-enhanced surface plume is not exposed to the atmosphere long enough to allow for full air–sea CO₂ equilibration (e.g., due to water mass subduction), CO₂ removal efficiency will decrease in these regions (Jones et al., 2014). A logical next step to characterize the spatial footprint of the hypothetical experiment and the total timescale for equilibration would be to simulate the time-evolving patch of the alkalinity anomaly, the corresponding pCO₂ perturbation, and the total excess DIC in an ocean general circulation model with the OAE experiment relative to a control simulation.

We argue that the TA-enhanced waters do not need to remain continuously within the surface layer to induce ocean CDR. The change in the air–sea flux can be realized whenever the perturbed waters influence the surface, even if that is far from the plume or years after the deployment. Though the implied dilution of the signal or delay in causing additionality of the CDR may not change the net impact of the OAE experiment if integrated over a sufficiently long time period, it would make MRV of the total impact essentially impossible through observational methods alone.

Another assumption in this work was that photosynthesis, which is known to drive pCO₂ beneath the conservative mixing curve (Mu et al., 2021), would be unchanged by the alkalinity addition. This assumption is premised on the argument that perturbations barely detectable above background variability are unlikely to strongly change the ambient ecosystem, an argument supported by a number of recent studies. For example, results from a mesocosm experiment with Tasmanian coastal waters show statistically significant, though relatively moderate, changes to phytoplankton community structure and function at high TA perturbation values of +495 µmol kg⁻¹, which are 25 times higher than the minimum detectable perturbation explored here (Ferderer et al., 2022). Another microcosm incubation study reveals the alkalinity enhancement inducing minor but largely negligible short-term effects on primary productivity and no measurable changes in biological CaCO₃ precipitation, even at a TA perturbation level of +2000 µmol kg⁻¹ (Subhas et al., 2022). Regardless, ecosystem impacts would require careful monitoring in any field OAE experiment.

If the alkalinity source for OAE is quicklime, managing the heat released through its exothermic reaction during dissolution in water at a rate of 64 kJ mol⁻¹ CaO is another important consideration. Taking the highest perturbation scenario explored in Table 2 that increases TA in the Amazon watershed by 100 µmol kg⁻¹, we estimate 310 MW total heat released during the duration of the dissolution. In order for the heat released from the CaO reaction to be lower than 10 % of the daily solar radiation at the Equator ∼400 W m⁻² (a level that would presumably be swamped by natural variability in the river's heat budget), the TA injection would have to occur over an area of river larger than about 8 km², e.g., 2 km of river width and 4 km along its length. Therefore, it is possible to minimize the environmental risk due to heat released during quicklime dissolution by spreading it over a large enough area at the river mouth.

5 Conclusions

We conduct a sensitivity analysis of alkalinity enhancement in the Amazon River watershed, evaluating the detectability of added TA and predicting its influence on the atmospheric CO₂ uptake in the offshore plume. Adding 20 µmol kg⁻¹ of TA in a month at the Amazon watershed could increase the CO₂ uptake by the river plume by at least 0.07–0.1 Mt CO₂ per month. The full CDR might be as much as 3 times as high as this estimate due to reduced outgassing in the freshest part of the plume, where we lack a robust empirical pCO₂^sw–salinity relationship and therefore cannot robustly quantify the anomalous air–sea exchange in this region. We found that a TA perturbation of +10–20 µmol kg⁻¹ in the river is readily detectable with at least 40 samples, given the background variability of 15–30 µmol kg⁻¹. By adding a TA perturbation barely detectable relative to the background TA variability, the likelihood of substantial environmental disturbance at the injection site would be minimized. However, observing the corresponding pCO₂ perturbation in the plume presents a greater observational challenge. Even with the highly sensitive pCO₂ sensors available today (e.g., Sabine et al., 2020), detectability of the pCO₂ anomaly is plausible only in the fresher part of the plume given the dilution of the signal at higher salinities and large background variability. Quantifying the total additional CO₂ uptake from this type of OAE cannot rely on observations alone; idealized conceptual models like the one presented here and more sophisticated circulation-biogeochemical ocean models will always be required to understand the total CDR because the additional CO₂ uptake is likely to occur over long time periods and at very diluted perturbation levels. It is also clear that the basic scientific feasibility we aimed to assess here does not reveal the full feasibility of such interventions: ocean-based CDR research and deployment require care and attention to ethical, political, and governance issues.