Reversing ocean acidification along the Great Barrier Reef using alkalinity injection

A large, multi-agency collaboration (https://research.csiro.au/ereefs/) has developed the eReefs coupled hydrodynamic, sediment and biogeochemical model that simulates at multiple scales the environmental conditions of the Great Barrier Reef (Schiller et al 2014, Steven et al 2019). This coupled model provides skilful predictions of water column concentrations of dissolved inorganic nutrients and particulate organic matter across the entire length of the Great Barrier Reef from 2011–present (Skerratt et al 2019, Baird et al 2020b). The hydrodynamic model is nested within a global circulation model, in order to provide accurate forcing along the boundary with the Coral Sea, and forced by atmospheric winds, solar radiation and 22 river flows. Carbon and alkalinity concentrations in freshwater inputs are prescribed as 900 mmol m⁻³ (Mongin et al 2016). The biogeochemical model simulates optical, nutrient, plankton, benthic organisms (seagrass, macroalgae and coral), detritus, chemical and sediment dynamics. The Ocean-Carbon Cycle Model Intercomparison Project (OCMIP2, (Orr et al 2017)) numerical methods are used to quantify air-sea carbon fluxes and the carbon dioxide system equilibria in seawater.

The OCMIP2 procedures quantify the state of the CO₂ system using two prognostic variables, the concentration of dissolved inorganic carbon and total alkalinity. The value of these prognostic variables, along with salinity and temperature, are used to calculate the partial pressure of carbon dioxide, pCO₂, The air-sea flux of carbon dioxide is calculated based on wind speed data and the air-sea gradient in pCO₂. Thus an injection of alkalinity will create and air-sea gradient, driving atmospheric carbon into the ocean. A diagram showing the pathways of alkalinity in the model is presented in the supplementary material (available online at stacks.iop.org/ERL/16/064068/mmedia).

Simulations using the eReefs coupled marine model have studied the exposure of the GBR to ocean acidification (Mongin et al 2016), the fate of anthropogenic terrestrially-derived sediment loads (Margvelashvili et al 2018), coral bleaching (Baird et al 2018), and the fate of nutrient loads from rivers (Wolff et al 2018). A data assimilating simulation (Jones et al 2016) has also provided environmental conditions of the GBR for reporting to national and international management bodies (Robillot et al 2018).

The GBR has more than 3860 individual reefs in the Great Barrier Reef Marine Park Authority Features Map. Many reefs are smaller than a single model cell and some cells contain multiple reefs. Thus only 2144 model grid cells contains one or more reef.

The dissolution of the silicate-containing mineral olivine is one of the proposed AOA approaches (Köhler et al 2013). Olivine is an abundantly-available magnesium silicate mineral. The dissolution of one mole of olivine leads to an increase in total alkalinity, $\textit{A}_\textrm{T}$, of 4 moles, and in silicic acid, $\textrm{H}_{4}\textrm{SiO}_{4}$, of 1 mole:

$$\begin{align} (\textrm{Mg},\textrm{Fe})_{2} \textrm{SiO}_{4} & + 4~\textrm{CO}_{2} + 4~\textrm{H}_{2}\textrm{O} \rightarrow 2(\textrm{MgFe}^{2+})\nonumber\\ & +4~\textrm{HCO}_{3}^{-} +\textrm{H}_{4}\textrm{SiO}_{4}. \end{align} \tag{ 1 }$$

In this study, we do not represent the process of substrate weathering and dissolution of olivine. Instead, we simply assume a surface flux of alkalinity calculated from the olivine molar mass (153.31 g mol⁻¹), using a conversion of 1 g of olivine = 1/153.31 × 4 = 0.026 mol of $\textrm{A}_\textrm{T}$. Further, we do not consider the kinetics of olivine dissolution. Thus, the $\textit{A}_\textrm{T}$ source can be considered directly as processed calcium hydroxide or slaked lime. Not considering dissolution makes our study a best-case scenario in what may be a wide range of possible outcomes.

We simulate alkalinity injection along the Gladstone-Weipa bulk carrier route (Gladstone Harbour, 23.84^∘ S, 151.26^∘ E to Cape York, with the route extending beyond the model boundary to Weipa, 12.65^∘ S, 141.84^∘ E). The route is a 2200 km path along the entire length of the GBR in the Designated Shipping Areas (figure 1(A)). Interestingly, suitable mineral deposits are mined in the Gladstone region, and are presently exported using port facilities in Gladstone Harbour.

Zoom In Zoom Out Reset image size

The typical transit time of bulk carriers along the Gladstone-Weipa route at 10 knots is 96 h. Ships presently using this route have a carrying capacity of 90 000 t. A complete unloading of the ship at sea during the 96 h transit translates into a deposition of $\textit{A}_\textrm{T}$ of 6795 mol s⁻¹.

With a model resolution of 4 km, and a 1 h integration time step, the ship travels across three model grid cells per time step. The alkalinity release is therefore spread over three model grid cells every hour. The ship release of alkalinity is repeated every 96 h, starting from Gladstone and moving northward. We simulated the release for 2 years, from 1 January 2014 until 31 December 2015.

The skill of the model's carbon chemistry simulations are assessed against observations from the bulk carrier whose route is used for the simulation. The observations include a custom-built sensor suite, installed on the RTM Wakmatha, to continuously collect ocean chemistry data along the length of the GBR during its regular voyages (Steven et al 2019).

The observations include semi-continuous sampling of pCO₂, temperature and salinity, as well as discrete bottle samples for $\textrm{A}_\textrm{T}$ and dissolved inorganic carbon ($\textrm{C}_\textrm{T}$) from 31 sites along the ship track. The samples are taken from 7 m depth northbound and 11 m southbound (fully loaded). In total more than 2000 discrete samples were collected from 80 transits from 2013 to 2019.

Three numerical simulations are run sequentially. The first simulates the carbon chemistry of the GBR without AOA. Here we designated this as the control simulation; it is referred to in the eReefs Project as GBR4_H2p0_B3p0_Cq2b_Dhnd. This simulation, begun on 1 December 2010, is described in Baird et al (2020a) and is compared to observations to determine the skill of the model (section 3.1). A second simulation uses the state of the control simulation on 1 January 2014 as initial conditions and is forced with the ship-based alkalinity injection until 1 January 2016. A third simulation, analysed in the Discussion, uses the state of the alkalinity injection simulation on 1 January 2015 and then runs until 1 January 2016 without further alkalinity injection. Thus by comparing the first and second simulations over a 12 month period (long enough so the alkalinity injection impact has reach a quasi steady state) we can determine the impact of the injection; and comparing the second and third we can determine the duration of impact of the injection following cessation.

The model is forced with realistic weather forcing and nested into a global ocean model. Thus a true steady state can never be achieved. The years chosen for the experiment correspond to a neutral El Niño–Southern Oscillation and low cyclonic activity in the GBR region.

The assessment of the skill of carbon chemistry predictions presented here was undertaken on the simulation without alkalinity injection described in Baird et al (2020b). For comparison with the observational data, we output model data at hourly intervals along the entire ship transit line. Each individual observation was then matched in time and space to the corresponding model output.

A summary of the skill statistics is provided in table 1, with detailed comparisons for each constituent in figure 2. A mathematical description of the skill metrics can be found in Baird et al (2020b) along with the skill analysis for a further set of biogeochemical parameters.

Zoom In Zoom Out Reset image size

Table 1. Summary statistics comparing the eReefs model with bottle samples taken along the Gladstone-Weipa bulk carrier route (alkalinity, $\textit{A}_\textrm{T}$, and dissolved inorganic carbon, $\textit{C}_\textrm{T}$) and underway sensors (temperature, salinity and pCO₂). $\Omega_\mathrm{ar}$ is calculated from $\textit{A}_\textrm{T}$ and $\textit{C}_\textrm{T}$ statistics assuming T = 25 ^∘C, S = 35.

	Mean absolute error	Bias	Root mean square error
Temperature (∘C)	0.78	0.06	1.18
Salinity	0.43	−0.29	0.65
$\textit{A}_\textrm{T}$ (mmol m−3)	32.11	−15.08	43.99
$\textit{C}_\textrm{T}$ (mmol m−3)	30.41	−22.70	39.21
pCO2 (ppmv)	5.5	−2.59	6.90
$\Omega_\mathrm{ar}$	0.08	0.14	0.13

For $\textrm{A}_\textrm{T}$, uncertainty (bias ± mean absolute error) is −15.08 ± 32.11 mmol m⁻³ and $\textit{C}_\textrm{T}$ uncertainty is −22.70 ± 30.41 mmol m⁻³ (table 1). These uncertainties are much less than the spatial-temporal variation seen along the track (of order 200 mmol m⁻³ for both $\textit{A}_\textrm{T}$ and $\textit{C}_\textrm{T}$, figure 2). Further, these uncertainties are 2-fold lower than that of current climate models (Zheng and Long 2015).

The model skilfulness (in relative errors) varies between pCO₂, $\textit{A}_\textrm{T}$ and $\textit{C}_\textrm{T}$ due to the underlying simulated processes and their interactions: pCO₂ is driven by large scale oceanic processes, equilibration with the atmosphere, carbon chemistry equilibrium as well as variability in temperature and salinity; $\textit{A}_\textrm{T}$ is driven by the calcification/dissolution balance of the reefs and the input of fresh water for the catchment (figure A1) which have greater uncertainties; and $\textit{C}_\textrm{T}$ is primarily driven by reef processes as well as the additional variability introduced by water column productivity.

We track the changes in $\textit{A}_\textrm{T}$ and $\Omega_\textrm{ar}$ following the commencement of AOA. The spatial footprint of the change in surface $\textit{A}_\textrm{T}$ after twelve months of AOA is shown in figure 3. The integrated surface footprint of AOA is 124 000 km² (red shading, figure 3); defined where values of $\Delta A_\mathrm T \gt$ 5 mmol m⁻³ ($\Delta \Omega_\textrm{ar} \gt + 0.05$). The area of elevated $\textit{A}_\textrm{T}$ occurred generally along, and inshore, of the GBR shipping channel (figure 1 green line). This arises from the generally alongshore coastal currents on the shelf (Baird et al 2017), with the greatest impact on shallow waters. The change in $\textit{A}_\textrm{T}$ is smaller in deeper water due to greater vertical ocean mixing (dilution). Similarly, filaments can be seen moving offshore, but are of a lower concentration due to mixing with surrounding waters. The largest increase in surface $\textit{A}_\textrm{T}$ due to AOA is located in the central GBR region. The widest footprint (in the offshore direction) is in the southern GBR (∼100 km wide), and the narrowest in the northern GBR (∼50 km wide).

Zoom In Zoom Out Reset image size

The aim of this study was to assess the impact on ocean chemistry of alkalinity injection using potentially existing infrastructure. In particular, to assess whether the AOA along the bulk carrier route effectively targets the GBR and is effective in ameliorating observed and projected changes. The change in $\textit{A}_\textrm{T}$ was spread relatively evenly across the continental shelf (figure 3) resulting in increases in $\Omega_\textrm{ar}$ on most inshore and midshelf reefs (figure 4). Without AOA, the inshore reefs have the lowest $\Omega_\textrm{ar}$ due to both freshwater addition along the coast and the downstream impacts of coral calcification (Mongin et al 2016). Here, we find that the injection of alkalinity along the shipping route does indeed target inshore reefs. While, in contrast, the reefs along the offshore boundary of the GBR are barely impacted by the alkalinity injection. However, this maybe less important as offshore reefs, while still impacted by global OA, are less exposed to low $\Omega_\textrm{ar}$ than inshore reefs.

Zoom In Zoom Out Reset image size

Over the the first 6 months of the AOA, $\Omega_\textrm{ar}$ increases along the GBR shipping lane (figure 4). The largest increases are located to the east of the shipping lane again impacting mostly the inshore reefs. After 12 months of AOA, the $\Omega_\textrm{ar}$ increase spreads further offshore impacting almost the entire GBR with the exception of the fringing reefs (which appear gray coloured dots due to small impacts). The reefs coloured purple on the central panel of figure 4 have had an increase of $\Omega_\textrm{ar}$ greater than 0.1 following the first year of AOA, corresponding to about 25% of the GBR (in reef numbers). If the AOA is stopped after 12 months, within 6 months the increase of $\Omega_\textrm{ar}$ reduces to less than 0.05 for almost all the reefs (figure 4 top right insert).

Carbon chemistry equilibrium kinetics dictate that the increase in $\textit{A}_\textrm{T}$ leads to a decrease in partial pressure of CO₂ in the ocean leading to an increase in carbon sequestration. In our experiment we are looking at the impact of alkalinity injection during a 6–12 months time period. At the regional scale the carbon chemistry equilibrium is a result of the interplay between the rate of vertical mixing in the water column, the depth of water, and the rate of air-sea exchange. An analysis of the $\textit{A}_\textrm{T}$ to $\textit{C}_\textrm{T}$ ratio after 1 year indicates that less than half of the potentially additional sequestrable carbon has been taken up by this ocean; the alkalinity is being mixed vertically in the water column before it could drive CO₂ exchange across the air-sea interface. The equilibration and sequestration of CO₂ will not be completed during the time the water resides within our model domain, therefore benefiting the reefs as CO₂ has not entered the ocean yet, so pH remains higher. We note that the carbon sequestered in the ocean following the AOA is 35 000 t, about 0.0001% of the current CO₂ global emissions (assuming 35 Gt of carbon being emitted annually).

When the AOA is stopped after twelve months, $\Omega_\textrm{ar}$ returns to pre-AOA values over 6 months (brown line continuing into the blue line in figure 5, figure 4, top insert). The time-scale at which the ocean carbon chemistry returns to the non-AOA state is approximately equal across all the reefs (as denoted by the small surface area of the contoured blue lines compared to the contoured brown lines). The reefs that have seen large increases in $\Omega_\textrm{ar}$ lose the AOA benefit on a timescale similar to the other reefs.

Zoom In Zoom Out Reset image size

The GBR-wide impact of the alkalinity injection is presented in figure 5. During the first 100 d, the mean change in $\Omega_\textrm{ar}$ ($\Delta \Omega_\textrm{ar}$) at the 3860 reefs reaches quasi-steady-state, with a mean value of +0.05 (median +0.03). This steady-state of +0.05 $\Omega_\textrm{ar}$ corresponds to an increase of 10 µatm pCO₂ at T = 25 ^∘C, S = 35, $A_\mathrm T$ = 2250 µmol kg⁻¹ and $C_\mathrm T$ = 1950 µmol kg⁻¹. An increase of 10 µatm pCO₂ in turn is equivalent to 4 years (∼4.2) years of OA when using the observed value of 2.35 µatm pCO₂ increase per year in the GBR (Fabricius et al 2020).

The 25 to 75 percentile (known as interquartile range) covers a range of $\Delta \Omega_\textrm{ar}$ between 0.0 and 0.08. We find that the lowest quartile (25%) of the reefs are not impacted significantly by the AOA while the upper quartile (25%) sees $\Delta \Omega_\textrm{ar}$ that is larger than the change required to offset 4.2 years of OA.

The mean change in $\Omega_\textrm{ar}$ following the AOA (+0.05) is on average similar to the variability in $\Delta \Omega_\textrm{ar}$ across reefs (interquartile range: 0.08). This result highlights the challenges we face in detecting changes due to AOA at an individual reef. The variability of the $\Delta \Omega_\textrm{ar}$ is slightly larger during the second year of AOA. This result indicates that under a continuous AOA some reefs will experience significantly higher values ($\Delta \Omega_\textrm{ar} \gt 0.10$). During the second year of the AOA (red lines and contours figure 5), the mean $\Delta \Omega_\textrm{ar}$ oscillates between 0.04 and 0.06, slightly above that of the first year. Throughout the 24 months of AOA, about 25% of the reefs have an increase in $\Omega_\textrm{ar}$ greater than the decrease predicted to occur at present anthropogenic carbon emissions over a 4.2 years period, meaning the proposed AOA could be effective at reversing some of the OA.

The variability in the response to AOA of $\Delta \Omega_\textrm{ar}$ occurs as a balance between the alkalinity injection, its dilution in the Coral Sea, and the uptake of atmospheric CO₂ driven by the reduced partial pressure of CO₂ in higher alkalinity waters. The variability highlights that predicting the efficacy of the AOA in a natural environment is complicated (as opposed to a controlled experiment or in climate style model simulations).

The lower panel on figure 5 shows the spatial distribution of $\Delta \Omega_\textrm{ar}$ after one year of AOA. About 1000 reefs, almost 40% of the GBR, have an $\Delta \Omega_\textrm{ar}$ larger than what is required to offset 4.2 years of OA. Although the number reduces to about 8% of the GBR if offsetting 8.4 years. To put our results within a reef resilience context (Hoegh-Guldberg et al 2018), we provide the location for the 50 (blue) and 10 (red) most impacted reefs by the AOA (map insert on figure 5). The most impacted reefs are all located within two clusters, one in the northern GBR just south of Princess Charlotte Bay where the shipping lanes are close to the reefs. A second cluster is located in the Mackay/Broad Sound region where dilution by offshore Coral Sea waters is reduced by the presence of the GBR lagoon. None of the 10 most buffered reefs (named in figure 5) have been identified as high value reefs or of special significance as yet, which highlights the need for a more targeted strategy for surveying reefs on the GBR.

Under a continuous AOA, the large amounts of alkalinity being released in the ocean may lead for some reefs to exposure to $\Omega_\textrm{ar}$ greater than experienced in the pre-industrial period. This in turn may have adverse impacts on the marine ecosystem (see review of Renforth and Henderson (2017) and references therein). Furthermore we anticipate the change to be amplified if the magnitude AOA were to be increased suggesting that any adverse effects would be likely found in the regions of largest changes in this study.

It is also important to consider the source of alkalinity used in AOA. The form of alkalinity employed in this study (slaked lime) acts through an abiotic physio-chemical process. However this form of alkalinity represents only one of a number of proposed sources and methods of extraction (Renforth and Henderson 2017). Many of the proposed techniques/methods use fine-crushed ultramafic rocks, or waste material such as mine tailings, as their source of alkalinity; which are known to contain trace metal and/or micro-nutrients. In many cases there will be an impact on biological productivity, for example iron in olivine can significantly enhance ocean productivity (Hauck et al 2016). In practise in future work the biological impacts of AOA will need to be accounted for, but are beyond the scope of this study.

We also note that this modelling study has focused on the feasibility of impacting water column carbon chemistry through AOA, rather than the environmental and ecological risks to the GBR (Mcleod et al 2019). The nature of these risks will depend partly on the delivery mechanism of alkalinity. If it is of a particulate form, the absorption and scattering of the particles, and their deposition on the seabed, are likely to result in changes in the benthic light availability (Bach et al 2019) which has already seen significant changes with anthropogenic sediment inputs (Fabricius et al 2017). Further refinement of the modelling system would allow some of these negative impacts to be modelled.

The most significant limitation of the 4 km coupled hydrodynamic-biogeochemical predictions of the impact of the alkalinity injection occur on small shallow reefs. The 4 km horizontal resolution of the model grid inevitably smooths the actual bottom topography. Individual grid cells containing small ($\lt$16 km²) reefs have a bottom depth that is a mean of the reef top, reef slope and surrounding ocean. With an unrealistically-deep model reef, air-sea and benthic fluxes will have a smaller impact on water column properties than they would above the actual reef. This error is greatest offshore, where shallow reefs reside on a 100+ m deep shelf (Baird et al 2018). In these cases, the model will likely under-predict the feedback between the alkalinity injection and the benthic calcification and dissolution fluxes. The model also does not simulate the wave-enhanced air-sea fluxes over shallow areas. These limitations suggest that shallow reefs in the model are likely to simulate less variance than an actual shelf-scale injection would cause.

Consequently, we have focused our analysis on the changes in the water column $\textit{A}_\textrm{T}$ and $\Omega_\textrm{ar}$ that characterise the impact of the injection and are primarily driven by regional scale processes and therefore not sensitive to the limitations mentioned above. Ultimately this study should be repeated using a much finer scale model, to better capture the physical and biogeochemical processes, however at present these high-resolution models are still under development.

Actively enhancing the role of the ocean in buffering the impact of ocean acidification in response to excess atmospheric CO₂ is starting to be widely considered but will require a new set of research studies and management tools (http://oceannets.eu) to be developed (Gattuso et al 2018). The potential efficacy of artificially increasing the ocean buffering capacity has been explored at the global scale in an Earth system context (Keller et al 2018, Lenton et al 2018). Researchers are exploring the impact at the regional and local scales, where any impacts of AOA will be felt (Albright et al 2016). This study is bridging the gap between these two scales, exploring how AOA could be undertaken at the scale of the GBR. This work also differs from large-scale geoengineering studies that consider AOA as a means to balance anthropogenic inputs of carbon (Köhler et al 2013), and instead focuses on reducing the impact of ocean acidification. In this way we can consider AOA at the scale of GBR not as geoengineering, but as Environmental Engineering with carbon co-benefit.