Iron fertilisation and century-scale effects of open ocean dissolution of olivine in a simulated CO₂ removal experiment

The additional consideration of iron fertilisation amplifies the alkalinity- and silicate-based CDR of olivine dissolution, leading to further increase of the potential oceanic CO₂ uptake, NPP and export production (figure 1, table 2).

Zoom In Zoom Out Reset image size

In the last decade of the simulation scenarios, the total effect of olivine dissolution with the 0.1% (1%) iron availability scenario leads to an additional CO₂ uptake of 1.4 PgC yr⁻¹ (1.6 PgC yr⁻¹) or a rise by +24% (+28%) compared to the CTRL. The relative contributions of alkalinity, silicic acid and iron in the 0.1% (1%) scenario amount to 69% (57%), 7% (6%) and 24% (37%), respectively. The effect of iron fertilisation saturates already at an availability of 1% with a maximum in Fe-driven CO₂ uptake of 0.6 PgC yr⁻¹ or 0.2 PgC per Pg olivine in the last ten years of the simulation (figure 1(a)).

The annual mean pH is 0.006 units higher than in CTRL at the end of the olivine dissolution simulation with 0.1% Fe-availability. The effects of alkalinity (0.009) and silicic acid (0.001) lead to a larger pH restoration, however, the iron effect leads to an increase of pH initially, but to a decrease of pH after approximately 50 years when the transport of remineralised carbon reaches the surface and outweighs the current carbon draw-down by iron fertilisation (figure 1(b)). A termination of the open ocean dissolution of olivine results in a lower pH relative to the unperturbed CTRL state only 10 years after the termination. The pH increase due to olivine dissolution is small compared to the pH decrease driven by anthropogenic carbon uptake from about 8.1 to $\lt$ 7.8 in our simulations over the 21st century.

Diatom NPP increases by 9.0 PgC yr⁻¹ or 50% after 100 years in the low (0.1%) Fe-availability scenario. Here, silicic acid contributes 58% and Fe-fertilisation 42% to the total rise in diatom NPP. Diatom primary production can be increased by a maximum of 11 PgC yr⁻¹ by olivine addition when an Fe-availability ≥1% is assumed (figure 2(b)). Nanophytoplankton NPP simultaneously increases by 6.8 PgC yr⁻¹ or 30% and saturates already at 0.1% iron availability (table 2, figure 2(b)). In total, the enhancement of NPP through the input of silicic acid and iron as biproducts of the olivine dissolution amount to +39% or +15.7 PgC yr⁻¹ in the 0.1% availability scenario and to +45% or +18.0 PgC yr⁻¹ in the 1% availability scenario with a relative contribution of ∼90% from iron (figures 1(c), 2(a)).

While silicic acid favours a shift towards diatoms (see table 2 and Köhler et al 2013), iron stimulates NPP of both diatoms and nanophytoplankton (figures 2 and 3 and Aumont and Bopp 2006).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In the last ten years of the simulation, diatom NPP changes based on Fe fertilisation are strongest in the Southern Ocean and in the equatorial and North Pacific, and close to zero in the remaining ocean basins. Nanophytoplankton NPP thrives outside the areas where diatoms profit: in a band to the North of the Southern Ocean and in the whole Pacific except the equatorial East Pacific (figure 3). Nanophytoplankton NPP is reduced in the Atlantic Ocean and in the subtropical Indian Ocean where iron is not the most limiting nutrient in the CTRL simulation (figure 4). The largest increase in CO₂ uptake is in the Southern Ocean and in a band north and south of the eastern equatorial upwelling area in the Pacific with minor contributions from the north Pacific and north Atlantic. In the eastern equatorial upwelling area itself a strong decrease of CO₂ uptake is detected (figure 3).

Zoom In Zoom Out Reset image size

Globally, the export production of particulate organic matter at 100 m water depth increases by 2.9 PgC yr⁻¹ (+28%) at the end of the 100 years simulation (0.1% Fe-availability) with a dominance of the iron fertilisation effect (+2.7 PgC yr⁻¹) over that of silicic acid fertilisation (0.2 PgC yr⁻¹). At maximum an additional export of 3.3 PgC yr⁻¹ can be obtained by olivine dissolution for an availability of iron of 1% or more (table 2, figure 1(d)).

The largest effects of iron fertilisation by olivine dissolution occur within the first ten years of the simulations (figure 1). When terminating the continuous dissolution of olivine, the dominant part of the benefit of the larger oceanic CO₂ uptake is lost within a few years and a state close to the control simulation without olivine addition is reached within approximately one decade (figure 1). Terminating open ocean dissolution of olivine dissolution leads to a lower pH relative to the CTRL simulation ten years after the termination (−0.002 in the last decade of the simulation) and to a loss of oceanic CO₂ to the atmosphere relative to CTRL 25 years after the termination (−0.2 PgC yr⁻¹ in the last decade of the simulation). However, the cumulative sum of CO₂ uptake is at every time higher in the SI+ALK+FE_1T simulation than in the CTRL, indicating that the release of the remineralised carbon to the atmosphere is rather slow. Net primary and export production show a slower recovery to pre-olivine conditions and are still enhanced relative to the CTRL simulation at the end of the simulations.

Nearly all of the iron-effect on production, pH and CO₂ flux occur in the Southern Ocean and in the Pacific (figure 3) where iron is the most-limiting nutrient in our CTRL simulation (figure 4). Despite the relatively short spin-up and a decreasing trend in global surface iron concentration (−6% over 100 years) that is caused by regions where iron is not a limiting nutrient (Atlantic and Indian Oceans), the surface iron concentrations in the Southern Ocean and in the Pacific vary little over time in the CTRL simulation and show no clear linear trend (figure S3). In addition, the surface iron perturbation by olivine dissolution is much larger than any temporal variability in these regions (figure S3). We can thus assume, that the response to iron addition does not change over time due to model drift. In the Pacific Ocean, models differ in the degree and extent of iron limited regions (Laufkötter et al 2015) and the response of primary production to iron addition will be model dependent. Traditionally, it is assumed that iron is the most limiting nutrient in high-nutrient-low-chlorophyll (HNLC) regions. Recent evidence, however, suggests that iron can also be a limiting factor in non-HNLC regions on a seasonal scale (e.g., Nielsdóttir et al 2009). Behrenfeld et al (2009) report strong evidence for iron (co-)limitation in large areas of the south Pacific Ocean.

The consumption of additional macronutrients along with the micronutrient iron in the Southern Ocean leads to nutrient depletion in Antarctic surface waters that get subducted to form northwards moving mode waters. The effects of this nutrient depletion become apparent only in the Atlantic and Indian Ocean in our model where NPP is reduced north of 40°S (figure 3). This is in line with the current understanding that the inefficiency of the biological carbon pump in the Southern Ocean due to the iron deficiency fuels present-day biological production in low latitudes (Sarmiento et al 2004). An enhanced efficiency of the biological carbon pump in the Southern Ocean therefore reduces NPP in lower latitude upwelling regions on multi-decadal to centennial time-scales (Rodgers et al 2003).

We also tested for synergistic effects of silicic acid and iron addition by comparing the iron effect in an iron-only scenario (FE_1) to the joint addition of iron with alkalinity and silicic acid (SI+ALK+FE_1). Synergistic effects favour diatom NPP, which is affected most strongly at the start of the simulation (+1.5 PgC yr⁻¹), and decreases over time to about +0.7 PgC yr⁻¹ in model year 2099 (figure S1). The simultaneous addition of silicic acid and iron has a negative effect on nanophytoplankton NPP (−1 PgC yr⁻¹ in year 2000, −0.1 PgC yr⁻¹ in year 2099). Synergistic effects on total NPP, export production and CO₂ flux are about 0.6, 0.05 and 0.03 PgC yr⁻¹ in year 2099, respectively, and are small in comparison to the iron effect itself.

All carbon fluxes saturate at an available iron input of 2.3 Tg Fe yr⁻¹ (or a one-time per year increase of 1.1 nmol Fe L⁻¹ in the upper 100 m; 1% Fe-availability scenario) in our set of simulations (figure 5(c)). The potential of carbon sequestration by olivine dissolution increases from 0.36 PgC per Pg olivine (SI+ALK) after 100 years to 0.57 PgC yr⁻¹ per Pg olivine when considering the iron fertilisation effect with an availability of 1%.

Zoom In Zoom Out Reset image size

The carbon uptake potential after a century from alkalinity and silicic acid effects is larger than the 0.28 PgC per Pg olivine that was found after 10 years in Köhler et al (2013). This difference can be explained by the steady increase of the CO₂ flux response to alkalinisation (32% increase to 0.33 PgC per Pg olivine from 2010 to 2099, figure 5(a)). We explain this temporal evolution of alkalinity-driven CO₂ uptake by the increased sensitivity of CO_{2 (aq)} and consequently CO₂ flux to alkalinity and carbon perturbations in a high-CO₂ ocean (Egleston et al 2010, Hauck and Völker 2015). We illustrate the four-fold increase of the sensitivity of CO_2(aq) to alkalinity input over 100 years in figure 5(d). This sensitivity is calculated as

$$\begin{eqnarray}&&{\rm{\Delta }}[{\mathrm{CO}}_{2(\mathrm{aq})}]=\displaystyle \frac{{\rm{\Delta }}\mathrm{Alk}}{{\gamma }_{\mathrm{Alk}}}[{\mathrm{CO}}_{2(\mathrm{aq})}].\end{eqnarray} \tag{ 2 }$$

with ${\rm{\Delta }}\mathrm{Alk}=1\;\mathrm{mmol}\;{{\rm{m}}}^{-3}$ and with output from CO2SYS which is run with model fields of surface DIC, alkalinity, temperature, salinity, dissolved silicic acid and phosphate (converted from the modelled nitrate fields by the Redfield ratio). ${\gamma }_{\mathrm{Alk}}$ is defined according to Egleston et al (2010):

$$\begin{eqnarray}&&{\gamma }_{\mathrm{Alk}}={\left(\displaystyle \frac{\partial \mathrm{ln}[{\mathrm{CO}}_{2({aq})}]}{\partial \mathrm{Alk}}\right)}^{-1}.\end{eqnarray} \tag{ 3 }$$

This rise in the CO₂ uptake of the ocean due to the alkalinity input depends on the assumed changes in the anthropogenic emissions and would be smaller in other RCP scenarios with lower atmospheric CO₂ concentrations.

The alkalinity effect on CO₂ uptake due to olivine dissolution can be considered permanent in the sense that even after a termination, CO₂ uptake will always be larger or equal to CTRL and the sequestered carbon will not be released to the atmosphere as long as alkalinity is not removed from the ocean. However, the iron effect of olivine addition is temporary, i.e. terminating an olivine dissolution experiment that releases significant amounts of iron would lead to a reduced CO₂ uptake capacity after the termination, i.e. in our model the CO₂ uptake is reduced relative to the CTRL simulation three decades after the termination (figure 1).

The CDR potential of iron fertilisation decreases over time from 1.6 to 0.2 PgC per Pg olivine (or from 4.8 to 0.6 PgC year⁻¹). Our estimate of a total CO₂ removal by iron fertilisation within one century of 113 PgC is almost twice as high as the 60 PgC obtained with the PISCES model in a study that explicitly simulates the ecosystem response to iron fertilisation (Aumont and Bopp 2006). Other studies that assumed complete macro-nutrient draw-down in simpler biogeochemical models gave higher numbers by design (e.g. 120–277 PgC Jin and Gruber 2003, Cao and Caldeira 2010b). Zeebe and Archer (2005) estimate a lower reduction of 30 PgC as they consider only 15 fertilisation events per year South of 55°S. When fertilising only the Southern Ocean for 92 years about 60 PgC (or 0.6 PgC yr⁻¹) are taken up by the ocean in a study by Oschlies et al (2010a). The higher response in our model compared to Aumont and Bopp (2006) is partly due to larger areas of iron limitation in the Pacific Ocean in REcoM2 (figure 4). Models differ widely in the simulation of nutrient limitation in the Pacific (Laufkötter et al 2015), and recent evidence suggests that iron limitation could be more wide-spread than traditionally assumed based on macro-nutrient concentrations (Behrenfeld et al 2009, Nielsdóttir et al 2009). Aumont and Bopp (2006) use a lower limit for iron concentration of 0.01 nM to which iron is restored. In addition the models differ in the CO₂ scenarios (RCP8.5 in REcoM2, SRES98-A2 in PISCES) and whether they prescribe atmospheric CO₂ concentrations (REcoM2) or emissions (PISCES). The fact that REcoM2 prescribes concentrations and that we use a higher emission scenario can likely explain parts of the higher CO₂ removal potential. Furthermore, the ecosystem models differ in various details, e.g. variable C:N but fixed Fe:N ratios in the version of REcoM2 used here, compared to fixed C:N and variable C:Fe ratios in PISCES.

Our model simulates an export efficiency of 20% (export at 100 m relative to vertically integrated NPP) independent of Fe-availability. This is slightly higher than the 17% reported by Aumont and Bopp (2006), but lower than the 50% sequestration efficiency (at 1000 m) that Smetacek et al (2012) observed in an open ocean iron fertilisation experiment, in which the formation of fast sinking diatom aggregates accelerated carbon transport to the deep ocean.

The ratio of ΔCO₂ uptake to Δexport (relative to SI+ALK) decreases with time from approximately 85% at ≥1% Fe-availability (75% at 0.1% Fe-availability) to about 20% (12%) after 100 years (figure 5(b)) as expected due to remineralised carbon being transported back to the surface by the ocean circulation (Aumont and Bopp 2006). These numbers compare well with Aumont and Bopp (2006, 60% at start, 25% at end).

Our study aims to estimate the carbon dioxide reduction potential (i.e. its upper limit) of the olivine dissolution approach by assuming total and instantaneous dissolution of the distributed olivine. Any real-world application of this approach might lead to a smaller molar ratio of oceanic CO₂ uptake per dissolved olivine simply due to an incomplete dissolution of the material within the surface mixed layer (Köhler et al 2013).

The iron contained in olivine increases the CDR potential of olivine dissolution by roughly 60% (figure 5(a) + (c)). At least 28% of the additional NPP and 40% of the increase in export production occurs in the Southern Ocean (table 2). The relative dissolution rate of olivine depends on temperature and on the depth of the mixed layer. Although the deep mixed layers in the Southern Ocean might lead to long residence time of distributed particles in the surface ocean, the low temperatures likely lead to slow dissolution rates (Köhler et al 2013), and a large fraction of the particles would probably sink undissolved to the abyss. Olivine dissolution rates were not yet derived for polar conditions (e.g. Hangx and Spiers 2009, Rimstidt et al 2012). Köhler et al (2013) estimate relative olivine dissolution rates around 50% between 40 and 60°S and between 10% and 40% south of 60°S. Olivine dissolution rates from long-term laboratory experiments are an order of magnitude smaller than from shorter experiments due to the growth of a biotic community at the surface of the rock (Oelkers et al 2015). Dissolution kinetics estimated in Köhler et al (2013) might therefore be faster than what might be achievable in nature. In addition, there are few ships going to the Southern Ocean (Köhler et al 2013). Therefore a distribution of olivine from ballast water of ships of opportunity is only possible north of 40°S. To gain the full CO₂ uptake potential calculated with our model, about 300 large ships (net tonnage of 300 000 t each) would be needed with year-round commitment for an annual distribution of 3 Pg of olivine (Köhler et al 2013). Note that ice coverage and weather conditions will not allow transporting olivine to polar regions all year round.

Olivine dissolution leads to a maximum DIC increase of 40 mmol m⁻³ at 300–400 m depth. Assuming a typical C:-O₂ ratio of 117:170 (Anderson and Sarmiento 1994) gives an estimate of maximum oxygen consumption by iron fertilisation in the Southern Ocean of about 58 mmol m⁻³ (or 1.3 mL L⁻¹). This is on the order of <25% of the background oxygen concentration (Garcia et al 2014), but it could occur in addition to an oxygen reduction of up to 20 mmol m⁻³ in the mixed layer of the Southern Ocean as a response to climate change (Cocco et al 2013).

The radiative benefit of the iron and silicate fertilisation effects with olivine dissolution might be partly offset by N₂O production that accompanies organic matter remineralisation. This effect is smallest (compensates 13% of radiative benefit by iron fertilisation) in the Southern Ocean where 30% of the fertilisation occurs, but could amount up to 40% of the radiative benefit in the equatorial Pacific, the second largest region of iron fertilisation in our simulations (Jin and Gruber 2003).

Since olivine contains also a large fraction of other trace metals, such as nickel and chromium (De Hoog et al 2010), their input in the ocean and their potential harmful impact on marine ecosystems need to be assessed. Also note that undissolved particles in the surface ocean might reduce downwelling light intensities, and even modify ocean albedo. While the first of these side-effects has briefly been discussed in Köhler et al (2013), for the second a coupled atmosphere-ocean model would be needed. Further uncertainties arise from possible ecosystem shifts, shading of benthic ecosystems (Powell 2008) and phytoplankton-light-feedbacks that could lead to a warming of the sea surface (Manizza et al 2008) and a solubility-driven reduction of CO₂ uptake.

Geoengineering methods such as olivine dissolution are nowadays only investigated in model simulations. Any efforts to either test or implement ocean-based geoengineering approaches need to concur with existing international agreements such as the 'Convention on the Prevention of Marine Pollution by Dumping of Wastes and Other Matter', known as the London Convention (1972) and the later 1996 protocol, known as London Protocol (1996). The London Protocol has indeed extended its regularities on the issue of geoengineering on its 8th meeting in 2013 (reports can be found at https://webaccounts.imo.org) and any future ocean fertilisation experiments (performed by the contracting countries of the London Protocol) are regulated and need to fulfil the requirements of the protocol.