3

Nutrient Fertilization

3.1 OVERVIEW

Ocean-based carbon dioxide removal (ocean CDR) via nutrient fertilization refers to the addition of micronutrients (e.g., iron [Fe]) and/or macronutrients (e.g., phosphorus [P], nitrogen [N], silica [Si]) to the surface ocean with the deliberate intent to (1) increase photosynthesis by marine phytoplankton, and thus enhance uptake of carbon dioxide (CO₂) from surface waters, and to (2) enhance the transfer of the newly formed organic carbon to the deep sea away from the surface layer that is in immediate contact with the atmosphere. Step (1) can be accomplished wherever growth of phytoplankton is limited by nutrients, which is the case for some or all phytoplankton over large regions of the ocean, except eutrophic regions such as often found close to continental margins. Achievement of step (1) has been demonstrated for a number of fertilization experiments via in situ measurements and remote sensing of ocean color. There is larger scientific uncertainty about achieving step (2). Depending on the location, export depth, and remineralization rates of sinking particles, carbon can be sequestered for 100- to 1,000+-year timescales in the deep ocean. As such, nutrient fertilization aims to enhance locally the magnitude and efficiency of the natural ocean biological carbon pump (BCP; Figure 3.1) using energy from the sun and nutrients either from within the ocean (see Chapter 4) or from outside the ocean. In the case of fertilization with micronutrients such as Fe, relatively small amounts of iron may be needed relative to potential (C) C sequestration, whereas the amount (mass and volume) of nutrient required for fertilization with nitrogen, phosphorus, or silicate will be many orders of magnitude higher.

According to criteria described in Chapter 1, the committee’s assessment of the potential for ocean nutrient fertilization as a CDR approach is discussed in Sections 3.1–3.5 and summarized in Section 3.6. The research needed to fill gaps in understanding of ocean fertilization, as an approach to durably removing atmospheric CO₂, is discussed and summarized in Section 3.7.

3.2 KNOWLEDGE BASE

Ultimately the BCP sets the vertical gradient in dissolved inorganic carbon (DIC) through the depletion of inorganic carbon in the surface waters due to incorporation into biomass during photosynthesis, and the net remineralization of organic forms of carbon below the euphotic zone. This gradient is maintained as the strength of the BCP balances the vertical components of ocean mixing that work to homogenize these gradients. Ultimately, the strength of the associated surface CO₂ depletion affects the partitioning of CO₂ between ocean and atmosphere (Takahashi, 2004). Models suggest that if we were to turn off the BCP globally, net atmospheric CO₂ levels would increase by 200 parts per million volume (ppmv) on timescales of many hundred years (Sarmiento and Toggweiler, 1984; Maier-Reimer et al., 1996). Likewise, if the depth of C remineralization on sinking particles were to deepen by 24 meters on average, atmospheric CO₂ could, at least on long timescales, decrease by 10–25 ppmv (Kwon et al., 2009) due to steeper vertical gradients of inorganic carbon and enhanced vertical exchange at shallower depths. As an ocean CDR approach, the main goal would thus be to strengthen, or increase, the net transport of organic carbon out of the euphotic zone, and thereby increase the efficiency of the BCP, thus decreasing the carbon content of the surface waters in contact with the atmosphere while boosting the fraction of carbon that is transported to the deep sea where it can be sequestered on timescales >100 years.

Iron Fertilization

Historically the evidence for a control on atmospheric CO₂, via changes in the supply of iron, comes from the geological record and the glacial-interglacial cycles and correlations between CO₂ as captured in ice cores and dust, a primary source of iron to the ocean (e.g., Martin et al., 1990). This correlation was popularized by John Martin in his famous quip first made at Woods Hole Oceanographic Institute in 1988: “Give me half a tanker of iron and I’ll give you the next ice age.” The implied potential for a high leverage in terms of a much higher mass of carbon removed per mass of iron applied spurred interest in a potentially efficient way to remove atmospheric CO₂ by oceanic Fe fertilization. What followed next were several fundamental experiments in the lab and in bottles at sea demonstrating this connection between Fe limitation and phytoplankton growth in high-nutrient, low-chlorophyll waters (HNLC) where phytoplankton growth was shown to be limited by iron rather than macronutrients (Martin and Fitzwater, 1988; Morel and Price, 2003). This was followed by more than a decade of purposeful open-ocean Fe addition experiments (Figure 3.2).

The goal of prior artificial ocean iron fertilization (aOIF) experiments was largely focused on assessment of the primary response to added iron, and not to track C sequestration and its impact on deeper ocean layers. Several manuscripts and reports have been written on the results of these aOIF studies (de Baar et al., 2005; Boyd et al., 2007; Yoon et al., 2018; GESAMP, 2019), and a consensus has been established that an increase in photosynthetic CO₂ uptake can generally be achieved. As an ocean CDR approach, this open-ocean testing of the impact of Fe enrichment puts this method far ahead of others in terms of the knowledge base. Consequently, it has also put this method at the forefront of public concerns regarding all forms of “geoengineering” and has led to many groups having already formed strong opinions for or against OIF. These social acceptance issues are often focused on OIF, yet this is only one ocean CDR approach, and many of the same acceptance issues would be common to at least all biotic ocean CDR approaches and in many cases abiotic ocean CDR as well, especially if deployed at scale (see Chapter 2). Also given that these early field experiments were conducted largely without international oversight, they prompted the establishment of guidelines for future ocean fertilization (OF) research under international agreements (see Section 3.4 and Chapter 2). However, at the time of writing, those guidelines were not legally binding.

Although the increase in photosynthetic CO₂ uptake (step i) via aOIF is well established, there is less consensus about the transfer and subsequent storage of carbon at depth (step ii). In summary, the two main questions that arise from deliberate aOIF experiments, and are common to all CDR approaches in this report, are: (a) Will it work (to remove carbon from the surface ocean and impact atmospheric CO₂ for some period of time?) and (b) What are the biogeochemical consequences

(both intended and unintended)? These issues are explored in more detail below, particularly regarding use of aOIF for ocean CDR. We consider its efficacy and permanence, possible consequences when done at scale, and the ecological and geochemical impacts and future research directions. Note also that any attempt to deliberately alter the oceans’ BCP will have consequences that should be considered relative to the status quo of doing nothing.

In addition to aOIF studies, natural systems with episodic or local high Fe delivery have improved the knowledge base for OIF. For example, a natural analog for natural nutrient fertilization is the atmospheric deposition of volcanic ash that leaches trace metals in seawater, generally promoting primary productivity (see Fisheries, below, and e.g., Duggen et al., 2007; Jones and Gislason, 2008; Hamme et al., 2010; Browning et al., 2014; Zhang et al., 2017). Study of Fe sources and impacts around islands in the Southern Ocean has also provided many clues as to the impacts of OIF at larger scales (see Export Efficiencies, below, and e.g., Blain et al., 2008; Pollard et al., 2009). Another source of nutrients to coastal environments is the deposition of ash from wildfires, a phenomenon that appears to be increasing in frequency and intensity as a result of anthropogenic perturbation (Jolly et al., 2015; Cattau et al., 2020). Only a few studies have considered the effects of fires on coastal marine ecosystems when increases in atmospheric deposition of metals or macronutrients are observed (Young and Jan, 1977; Sundarambal et al., 2010; Kelly et al., 2021). One example is an unusual bloom and coral reef die-off during 1997 in Indonesia that has been explained by Fe deposition into the surface ocean by nearby wildfires. Also, an unusual phytoplankton community composition in the Santa Barbara Channel (Kramer et al., 2020) appears to be the result of atmospheric deposition of ash leaching metals and carbon following the Thomas Fire in California in 2017 (Kelly et al., 2021).

Macronutrient Fertilization

The global carbon cycle, marine biogeochemistry, and Earth’s climate are thought to have been affected by the supply of macronutrients from continental weathering and on timescales of tens of thousands to hundreds of thousand years. Compared to OIF, ocean macronutrient fertilization (OMF) has received less attention in the scientific community (but see Harrison, 2017). It has the obvious disadvantage of much larger amounts of material required per ton of carbon removed (see Costs and Energy, below). One possible advantage of OMF compared to OIF is the fact that low-nutrient, low-chlorophyll (LNLC) regions are easier to access than the Southern Ocean, the prime candidate region for OIF. Fertilization with inorganic nitrogen has been investigated and suggested as a CDR measure in N-limited LNLC regions (Lawrence, 2014) where sufficient phosphate is available. The few available cost estimates have been low (Jones and Young, 1997, estimate $20/t CO₂). While inorganic N fertilizer can, in principle, be fixed from the atmosphere, albeit at substantial energetic costs, a marine application of phosphate will have to consider that phosphate is a nonrenewable resource also needed in agricultural food production.

While the increase in photosynthetic CO₂ uptake (step i) is widely assumed uncontested, an unexpected decrease in chlorophyll biomass was observed in response to phosphate addition to the ultraoligotophic eastern Mediterranean, indicating that complex food web dynamics and ecosystem responses have to be carefully accounted for when making inferences on C fluxes induced by OF (Thingstad et al., 2005). It is not yet clear whether this is an issue for OMF and less so or not so for OIF. For OMF, the transfer of carbon to the deep ocean has received little attention and will face the same issues as for OIF. The following sections will therefore concentrate on OIF, for which a larger number of theoretical and experimental studies have been performed and the knowledge base is considered advanced compared to OMF.

3.3 EFFICACY

Export Efficiencies

Thirteen open-ocean aOIF studies were conducted between 1993 and 2009 by the oceanographic community as research experiments, resulting in a significant body of literature and several reviews comparing them (Boyd et al., 2007; de Baar et al., 2008; Yoon et al., 2018; GESAMP, 2019). In these field experiments, from 350 to 4,000 kilograms (kg) of iron was added in the form of Fe sulfate dissolved in acidic waters and released in the propeller wash of a moving ship more than 25 to 300 square kilometers (km²) in one or multiple additions resulting in initial Fe concentrations between ≈1 to 4 nanomoles (nM). These experiments resulted in variable growth response (net primary productivity [NPP] increased by <400 to >1,700 milligrams (mg) C per square meter per day) and shifts in community structure, largely driven by the growth of diatoms of several types (Table 1 in Trick et al., 2010; Tables 2 and 4 in Yoon et al., 2018). Observations from ships extended from as short as 10 days to 30–40 days, and in most cases, the fate of the enhanced growth was not studied due to the limited time on site, the lack of appropriate sampling and measurement tools for particulate organic carbon (POC) fluxes, and in many cases, continued addition of iron that kept the bloom in progress.

To use these results to address the C sequestration efficiencies in response to iron, we need to consider not just the molar ratio of iron added to carbon incorporated into algal growth (C:Fe of 150,000:500,000; Sunda and Huntsman, 1995; de Baar et al., 2008), but also the ratio of iron to carbon that is exported. In summarizing early aOIF experiments in the Southern Ocean, Buesseler and Boyd (2003) noted that two studies—Southern Ocean Iron RElease Experiment (SOIREE) and EisenEx (Eisen is iron in German)—showed no increase in C export in the form of sinking POC to depth within 13 to 23 days after fertilization. A third aOIF study, Southern Ocean Iron Experiment (SOFeX)-South, showed a measurable increase in POC flux between the control and fertilized patch after 30 days, with a C:Fe molar export ratio of 8,000 at 100 meters. These authors noted that the aOIF observations were too short to determine the ultimate fate of the Fe-induced POC export, but these data did not support some of the more optimistic claims surrounding the low cost and small amount of iron needed for ocean CDR (Buesseler and Boyd, 2003). Using the same 100 meters boundary for POC export, de Baar et al. (2008) reported C:Fe export ratios ranging from 650 to 6,600 in three aOIF studies, including those reported by Buesseler and Boyd (2003). De Baar et al. (2008) attributed this relatively modest efficiency compared to algal growth needs, as being due to 75 percent of the added iron being rapidly associated with colloidal forms and subsequently quickly lost via scavenging and hence unavailable for algal growth.

In support of this proposed Fe loss mechanism, de Baar et al. (2008) summarized several natural OIF studies where there was a nearby island source of natural iron in the Southern Ocean resulting in long-standing, yet locally variable bloom and export responses (Blain et al., 2008; Pollard et al., 2009). During one of these, CROZET, C:Fe export ratios ranging from 5,400 to >60,000 were found. Even higher natural C:Fe export ratios were found off the Kerguelen Plateau (up to 174,000). De Baar et al. (2008) compared different estimates of the C:Fe efficiency made using several methods, from looking at POC determined by traps and radionuclide methods, to quantifying export by calculating upward diffusive fluxes of iron and calculating a C balance. Suffice it to say that natural OIF studies showed higher C export ratios in response to iron than aOIF. Presumably, in the natural system, the community response is more likely to reach a steady-state or at least seasonal balance between sources and losses and is less impacted by the episodic nature of aOIF experiments as conducted to this point. One area of research and development (R&D) would thus be looking at the forms of iron added, increasing the Fe-binding ligands in an attempt to minimize losses (see Research Agenda, below), varying the input from pulse to continuous, as well as extending observations to full growth cycle including the bloom demise (several months).

For OIF to sequester carbon from the atmosphere, we need to consider not only the C:Fe ratios leaving the surface, but also the extent to which carbon associated with sinking particles (or other pathways of the BCP) is attenuated with depth, as it is only with C transport below at least the depth of annual winter mixing that carbon can be considered sequestered in terms of a CDR approach (see discussion of durability, below). Few of the aOIF experiments had depth-resolved C export production (EP) measurements, but one that did, LOHAFEX, observed a factor of 8 decrease in POC flux between 100 and 450 meters (using neutrally buoyant sediment traps; Table 5 in Yoon et al., 2018). This is not dissimilar to the expected range in POC attenuation associated with sinking particles and the natural BCP. In a summary of shallow POC flux attenuation below the euphotic zone in the natural BCP, Buesseler et al. (2020) found that up to 90 percent of the sinking POC flux can be lost in the first 100 meters below the euphotic zone, though in some settings, essentially no attenuation can be measured in those first 100 meters, depths over which POC flux attenuation is typically the greatest. This flux attenuation is the result of combined processes that convert sinking forms of carbon to nonsinking forms, such as occurs with “sloppy feeding” by zooplankton on large organic aggregates, and by heterotrophic consumption of sinking particles and conversion to dissolved organic and inorganic carbon by resident zooplankton, microbes, and other animals in the mesopelagic.

If attenuation efficiencies can be controlled or altered during purposeful additions at sites where the communities are more likely to sequester carbon, such as after the sinking of intact diatom cells, then the effectiveness of CDR would be directly affected, or at least the amount of iron needed greatly reduced. Looking again at natural systems, this total loss of carbon starting with NPP and export out of the euphotic zone and transferred 100 meters below, varies from 1–50 percent (export efficiency - C export/NPP). So in estimating the effectiveness of OF for ocean CDR, there remains a large uncertainty in these factors, which determines costs and potential biogeochemical impacts below a purposeful event, as well as its permanence (see below). Also of importance, is that the depth of remineralization for carbon, iron, and other macronutrient remineralization will differ. For example, the depth of remineralization typically follows the order of P < N < C < biogenic silica, from shallowest to deepest, but little is known about the remineralization depth of sinking particulate iron (Lamborg et al., 2008), which is presumably shallower for biogenically incorporated iron and deeper for detrital iron, which would track more closely the lithogenic fraction of the particle flux. More recent studies confirm the importance of particle composition and type in regulating Fe remineralization (Bressac et al., 2019). R&D directed at measuring and purposefully changing these export ratios is needed and cannot be answered by these initial 13 aOIF experiments. As such, current cost estimates for OIF (see below) are limited by the variations in export ratios, but compared to other CDR methods, particularly abiotic ones, OIF would require only a small amount of iron to have a large impact on C sequestration.

Durability or Permanence of CDR

Similar to the permanence issue for land-based CDR, any ocean-based CDR is only as effective as its durability, or timescale over which carbon is removed from and then returned to the atmosphere. This would hold whether using a C capture and storage method, where CO₂ was deliberately injected into the deep ocean, or as discussed in this section, carried into the deep ocean via sinking organic matter, such as in response to stimulation of phytoplankton due to OF. Likewise, biotic methods that deliberately sink organic matter from macrophytes would face a similar issue with durability depending upon where and to what degree the carbon degraded during sinking (see Section 6.2 on Macroalgae). Here we use 100 years to define what is considered “durable” (or “permanent”) C sequestration, similar to several land-based options such as enhanced management

of forests. This sequestration time frame in the ocean is largely determined by depth and location and is set by the mixing and circulation properties of the ocean.

Primeau (2005) characterized a “first-passage time” as the time when a fluid element at depth in the ocean will make its first return contact with the surface ocean, and thus CO₂ would be able to leak back into the atmosphere. While the Atlantic Ocean in his model had generally younger water mass ages than the Pacific Ocean (difference by about a factor 2 in the deep basins), first-passage times were found to be more uniform over different latitudes and ocean basins. In a model-based analysis employing a steady-state assumption for ocean circulation, Primeau (2005) found that these times were generally greater than 200 years for depths below 500 meters and about 600 years at 2,000-meter depth. A more recent study that illustrates the global pattern versus depth over the 100-year time horizon for C injections can be found in a model by Siegel et al. (2021). The shallow retention times are quite short, with less than 50 percent of the carbon retained more than 100 years in large parts of the ocean if carbon is introduced above 200–500 meters, but carbon is largely retained in most areas when introduced below 1,000 meters, with retention times of centuries, except in the North Atlantic Gyre, along the Southern Ocean polar frontal regions, and in the Southern Indian Ocean east of Africa (Figure S2 in Siegel et al., 2021a).

Another way to consider the timescale for C sequestration is to consider the fraction of CO₂ retained given variations in the attenuation of sinking POC flux from the surface to the 100-year sequestration depth. Siegel et al. (2021) show this as a map of the fraction of carbon retained for 100 years that leaves the surface euphotic zone (Figure 3.3), using the Martin et al. (1987) POC attenuation power-law exponent b which is a best-fit parameter for POC flux versus depth between generally 100 and 1,000 meters. A larger b signifies faster POC flux attenuation and thus less carbon brought to depth where it is sequestered. For example, with the global average b of 0.8 for POC flux attenuation, around 30 percent of the carbon leaving the surface would reach a depth of >100-year sequestration (Figure 3.3, center). This is not surprising since it is well known that much of the sinking POC flux in the ocean is lost due to natural processes that remineralize carbon in mid-waters. Using a faster carbon attenuation (b = 1.0) results in less carbon being sequestered (Figure 3.3, right), whereas slower C attenuation (b = 0.6) results in many regions exceeding 50 percent retention over 100-year timescales or longer (Figure 3.3, left). In practice, b varies from >1 to <0.5 (Buesseler et al., 2020), but the response to aOIF has been the generation of diatom blooms that in the natural ocean are more often characterized by a lower b, hence the map showing b of 0.6 may be a better predictor of regional patterns of C sequestration for a surface source of fresh POC following aOIF. The issue of deliberately reducing or selecting for low POC attenuation efficiencies is an area of further research since the overall effectiveness of OF as a CDR approach will depend greatly on the fraction that reaches the deep ocean (see Export Efficiencies, above, and further discussion below).

A region particularly well suited for long C sequestration might be the Southern Ocean south of the biogeochemical divide separating the Antarctic from the sub-Antarctic (Marinov et al., 2006) where surface waters and sinking matter enter the deep cell, or “unproductive Southern Ocean circuit” according to Toggweiler et al. (2006) of the global overturning circulation (Ferrari et al., 2014). Besides reaching long first-passage times of the deep waters entering the deep overturning cell, a second advantage compared to other regions is that the supply of macronutrients originates from the shallower cell and is thus not affected by OIF-induced changes in macronutrients. Removal of macronutrients from waters south of the biogeochemical divide would also have less deleterious effects on biological productivity elsewhere. Such effects would otherwise lead to slow saturation of the OIF-induced global mean air–sea fluxes under continuous fertilization, whereas no such saturation and sequestration timescales exceeding 100 years are seen in a modeling study south of the biogeochemical divide (Sarmiento et al., 2010).

Of significance to ocean CDR is the need for (1) thorough measurements and models, to quantify the permanence for a given site, and (2) deliberately selecting sites and enhancing export efficiencies to optimize for maximal sequestration time. Enhanced C removal and efficient transport via the BCP will depend on the pathway that carbon takes to reach the deep sea. For example, the physical pumps that transport suspended POC or mix dissolved organic carbon below the surface will not lead to long-term sequestration in regions other than those where deep waters are formed. Neither will most active migrations, such as zooplankton diel vertical migration, because these processes are largely limited to <1,000 meters (Boyd et al., 2019). However, the gravitational settling of POC does reach the seafloor, with on average ≈10 percent of the carbon fixed via NPP in the euphotic zone reaching 1,000 meters (Martin et al., 1987) and of that, <1 percent is buried in ocean sediments. So it is the gradient in sinking POC remineralization that will set BCP sequestration efficiencies for a given location. These remineralization rates vary with the speed of particle settling and are likely modulated at least by temperature and oxygen (Devol and Hartnett, 2001; Van Mooy et al., 2002; Boscolo-Galazzo et al., 2021). These BCP efficiencies vary widely in natural settings (e.g., Buesseler et al., 2020), and could potentially be altered if one could select for formation of blooms with fast-sinking pellets (e.g., salps), carcasses, and/or sites with less microbial degradation (e.g., colder waters and low O₂). Ultimately, this return flow of carbon to the surface from the site of export sets the time frame for the permanence OF as a CDR approach, and this must be considered and compared relative to other CDR approaches.

Monitoring and Verification

Effective CDR requires C accounting that is transparent and verifiable and requires ways to monitor the ecosystem responses in the upper ocean where OF is applied and in the deep sea where carbon is intended to be sequestered. In the case of OF, the area of treatment for the demonstration projects could be relatively large (>1,000–10,000 km²) with a timescale of several months to years. The broad synoptic images (swath widths ≥2,500 km), high spatial resolution (~1 km), and rapid resampling (nearly daily global coverage) make satellite ocean color observations well suited to document the enhancement of surface ocean productivity created via OF (e.g., Westberry et al., 2013). Previous OIF studies have used satellite maps of phytoplankton chlorophyll concentrations to map the extent of and changes in phytoplankton due to trace nutrient addition (e.g., Abraham et al., 2000; Westberry et al., 2013). Existing satellite data products can also be used to assess changes in the phytoplankton Fe stress either via changes in the chlorophyll-to-C ratio or the solar-stimulated chlorophyll fluorescence line height (Behrenfeld et al., 2005, 2009; Westberry et al., 2013; Xiu

et al., 2014). In an exciting new development, the National Aeronautics and Space Administration’s (NASA’s) upcoming Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission¹ will provide global, hyperspectral (5-nm resolution) observations of ocean color reflectance spectra (Werdell et al., 2019), improving the quantification of phytoplankton composition from satellite data (e.g., Uitz et al., 2015; Xi et al., 2017; Kramer et al., 2021). Merged satellite altimetry data products will also be useful for assessing the trajectory of surface water parcels and stirring of patches, which may in turn affect the efficiency by which Fe additions are utilized by the surface ocean (e.g., Abraham et al., 2000). Thus, satellite observations provide a suite of data products useful for monitoring and verifying the effectiveness of OIF for ocean CDR.

However, satellite data only measure surface ocean properties and do not provide estimates of C export and other necessary biogeochemical determinations. The biogeochemical Argo profiling float network (which would continue to operate after the start of the treatment) can be used to track surface productivity and variations in plankton biomass (Yang et al., 2021) as well as provide needed subsurface data. A suite of autonomous surface vehicles and water column gliders (the number would depend on the aerial extent), outfitted with abiotic sensors (temperature, salinity, pressure, pCO₂, O₂, bio-optics, and nutrients) would be deployed on and beneath the treated bloom area. These autonomous sensor platforms profile between the sea surface and ocean depths every few days, measuring water properties and relaying data via satellite. At least one area outside of the bloom area, but in a similar water depth and ecosystem, could be monitored as a control, with similar sensors and sampling systems. These sensor systems would deliver an important subset of essential ocean variables (see Danovaro et al., 2020) to document the resulting bloom size and its impact on ocean physical and chemical properties. To quantify organic C uptake and the strength and efficiency of the BCP, methods that sample and/or optically characterize particles are needed (McDonnell et al., 2015). To quantify the extent of C storage below known reference depths for C sequestration (see section above on permanence), sediment traps (e.g., Buesseler et al., 2008), bio-optics (e.g., Giering et al., 2020), or other radionuclide-based particle flux tracers (e.g., Waples et al., 2006) would be used to measure the POC and associated elemental fluxes to the deep ocean. Consideration in these monitoring strategies for assessment of sinking POC also needs to consider the horizontal displacement of surface particle sources and the eventual location of C sequestration in the deep sea, that is, consideration of particle source funnels (Siegel et al., 2008). Examples of several monitoring technologies are shown in Figure 3.4.

Management of deep ocean ecosystems used for ocean CDR requires establishment of benchmark conditions and monitoring regional-scale threats, such as marine heat waves and ocean deoxygenation, in addition to site-specific monitoring of the C removal. Ships of opportunity (e.g., Smith et al., 2019) could also be coordinated and engaged to corral the floats to remain in the regions of interest or reseed an area with floats and other monitoring systems. The ships themselves could also provide data and data products. Data and data products from decades-long “Line” surveys in some regions (e.g., Line P in the Northeast Pacific) provide a historic framework for understanding benchmarks and/or long-term ocean change in an HNLC area suitable for OIF (Wong et al., 1995; Timothy et al., 2013).

Given that the potential for the area to be fertilized is extremely large, DARPA’s (Defense Advanced Research Projects Agency’s) Ocean of Things,² currently under development, will be useful for monitoring impacts. Such systems are formed by an interconnected network of small, inexpensive, and potentially biodegradable sensors and floats. Their sensor suites include measuring sea-surface temperature over a large region, so that it can be mapped with sufficient density to better understand ocean currents and mixing. The premise is that this dense measurement network

___________________

¹ See https://pace.gsfc.nasa.gov.

² See https://oceanofthings.darpa.mil/.

can be combined with remote sensing data and models to merge observations and assess the fate of carbon. The program is in development, so there is an opportunity to incorporate small sensors specific for OF monitoring, such as ocean MINIONS³ a small, inexpensive isopycnal float with onboard sensors and an upward-looking camera that quantifies POC export associated with sinking particles (Melissa Omand, University of Rhode Island, ongoing personal communication).

To assess the potential positive or negative impacts on regional fisheries, information on trends in species diversity and fish abundances is a key metric. Imagery from fixed-observatory underwater cameras and regular ship-based video transect surveys could be used to develop data products on species diversity and abundance. Manual, machine vision, and crowdsourcing tools are several approaches that extract biological information from video and photo archives to reveal trends in species numbers (see, e.g., Matabos et al., 2017). Understanding impacts higher up the food chain would require sampling with traditional nets and use of new techniques such as eDNA methods (e.g., Closek et al., 2019) to catch trends that occur over multiple years of growth and adaptation to purposefully altered conditions.

Additionality and Downstream Effects

The aim of OIF is to stimulate photosynthesis, the production of biomass and uptake and redistribution of carbon. In this process, nutrients other than iron are taken up and will be redistributed as well. The larger the amount of carbon sequestered is, the larger will be the redistribution of nutri-

___________________

³ See https://twilightzone.whoi.edu/work-impact/technology/minions/.

ents. This will have local and remote effects on nutrient fields and therefore on nutrient limitation, biological production, biological diversity, and, eventually, the marine BCP.

One effect that has received attention is the so-called nutrient robbing (Shepherd, 2009), whereby macronutrients utilized during aOIF-induced biological production are not available for biological production and associated C uptake elsewhere. Besides its ecological implications, this represents a nonlocal CO₂ leakage and presents difficulties for appropriate accounting of CO₂ sequestration achieved by aOIF (Oschlies et al., 2010a). This effect is of particular concern for aOIF in the Fe-limited surface waters of the tropical and subpolar North Pacific, where no deep waters with significant amounts of unutilized (preformed) nutrients are produced, and hence all macronutrients in the surface layer will essentially be used up anyway, and aOIF would predominantly lead to a relocation of the areas of biological production. The net CO₂ sequestration inferred from individual patchy fertilization experiments may thus be considerably overestimated (Aumont and Bopp, 2006).

Nutrient robbing would also occur for aOIF in the Southern Ocean, where surface waters tend to be replete upon subduction, thereby forming nonzero preformed nutrients. Model studies indicate that Southern Ocean OIF will lead to nutrients being trapped in the Southern Ocean and less nutrients will be exported to regions farther north, eventually leading to a reduction of biological production north of the Southern Ocean (Oschlies et al., 2010a). Relieving the Fe stress on diatoms via Southern Ocean OIF may also lead to changes in the Si:N ratio of the organic matter export and, consequently, to a change in silicic acid leakage from the Southern Ocean (Holzer et al., 2019). Nutrient robbing by Southern Ocean aOIF is likely to be accompanied by eventual reduction in biological production in much of the “world ocean” outside the fertilization region. Models suggest that these Southern Ocean nutrients currently fuel up to three-quarters of the biological production in the global ocean north of 30°S (Sarmiento et al., 2004; Marinov et al., 2006).

3.4 SCALABILITY

Like all ocean CDR approaches, models are used to assess the scale at which OF would affect the global carbon cycle. These models have focused on a particular region (e.g., Southern Ocean) or HNLC regions globally and often use the complete drawdown of surface ocean macronutrients to simulate enhanced primary production and the amount of potential C removal. There are important differences between models, however, including the extent to which deep C sequestration is considered versus shallow C export; whether nutrient co-limitation is included; the timescale of removal and reequilibration of CO₂ with the atmosphere; and, for example, ignoring the impact of OIF on LNLC regions via stimulation of N₂ fixation. These are just some examples of why the estimates thus far on the total scale of OIF alone range widely, from a fraction of a Gt C/yr to up to 3–5 Gt C/yr (Table 3.1), with a recent GESAMP (2019) report settling on 1 Gt C/yr (3.7 Gt CO₂) as the maximum theoretical potential. Practical consideration for engineering such large-scale deployments is also not considered. Deliberate alteration of ocean ecosystems to this extent would have many impacts and feedbacks not included in these models, but certainly the potential exists to augment the natural BCP of 5 to 12 Gt C/yr (C flux at the base of the euphotic zone; Siegel et al., 2014) by a Gt/yr or more.

One outcome of these models is that regional differences in ocean CDR capacity for OIF are large. For example, numerical models generally show a maximum C sequestration potential when OIF is applied to the entire Southern Ocean, the largest HNLC region of the world ocean, during the growing season when growth is not limited by light. This would lead to substantial net air–sea CO₂ fluxes (Aumont and Bopp, 2006). In contrast, OIF was found to have limited impact when applied in the equatorial Pacific (Gnanadesikan et al., 2003).

TABLE 3.1 Ocean Iron Fertilization Global Sequestration Potential

Over longer timescales, a model applying OIF everywhere south of 30°S found that OIF-induced air–sea flux of CO₂ is largest during the first year, reaching 5 Gt C/yr in Keller et al. (2014), but drops to less than 2 Gt C/yr within 10 years and about 1 Gt C/yr on centennial timescales. The large C uptake during the first year can be explained by the large macronutrient reservoir that becomes accessible upon the relaxation of Fe limitation. Export of organic matter and subsequent remineralization at depth leads to trapping of much of these nutrients (and carbon) in the Southern Ocean. Upwelling of the nutrients and respired C trapped in the Southern Ocean offsets a substantial fraction of the OIF-mediated downward flux of POC, leading to a substantial return flux of respired CO₂ to the atmosphere and an atmospheric uptake efficiency, defined as the ratio of air–sea CO₂ flux to export production (Jin et al., 2008), of less than 0.5. The uptake efficiency is also affected by the lowering of atmospheric pCO₂ by successful OIF (or other CDR schemes). This will shift the CO₂ air–sea partial pressure difference toward a net efflux of CO₂ from the ocean to the atmosphere. A similar efflux might also occur for the net C flux between the terrestrial biosphere, where photosynthetic CO₂ uptake is often stimulated by elevated atmospheric CO₂ concentrations. The compensating effect of such effluxes due to CDR-induced changes in atmospheric CO₂ changes from a few percent in the first year of CDR operation to about 10 percent on decadal and 50 percent on centennial timescales (Oschlies, 2009).

3.5 VIABILITY AND BARRIERS

General Considerations

The intention of OF is to stimulate photosynthesis and the production of organic matter. Similar to any biological CDR method, this intentional perturbation of natural ecosystems will change species composition, food web structure, and biodiversity, and will generate winners and losers until a new ecosystem is established.

Fertilization-induced enhancement of biological production will also lead to enhanced remineralization and oxygen consumption. While oxygen levels will thus decline below fertilization areas, the trapping of nutrients is expected to lead to a decline in biological production and eventually oxygen consumption in other regions of the world ocean. For Southern Ocean OIF, the volume of low-oxygen waters located in the tropical oceans may thus even shrink despite a global decline in the marine oxygen inventory (Oschlies et al., 2010a).

The remineralization part of the nitrogen cycle also involves nitrification, during which nitrous oxide (N₂O) is produced. A second pathway for enhanced production of N₂O is associated with anaerobic remineralization in low-oxygen environments that may expand in response to OF. Detailed understanding of the rates of N₂O production and possible consumption is still lacking, but direct measurements during the aOIF SOIREE measured increased N₂O emissions from the fertilized patch that would offset 6 percent to 12 percent of the OIF-induced CO₂ uptake (Law and Ling, 2001). Similar offsets were inferred from models (Jin and Gruber, 2003; Oschlies et al., 2010a). It is important that this offsetting of enhanced biological CO₂ uptake by N₂O produced from enhanced remineralization will likely occur for any biological marine CDR scheme as well as for biological terrestrial CDR schemes.

Another non-CO₂ greenhouse gas (GHG) that has been observed in nine aOIFs is production of dimethyl sulfide (DMS), which can lead to the formation of cloud condensation nuclei above the ocean and thus provide additional positive co-benefit in terms of reducing global temperatures (e.g., Law, 2008). But field results are variable, with larger DMS increases seen in the Southern Ocean versus the North Pacific.

In summary, a number of trace gases could be affected by OIF, not only N₂O and DMS, but various halocarbons, methane (CH₄), and isoprene (see Figure 1 in Law, 2008), and accounting for their positive and negative feedbacks on climate needs to be included in research studies of OIF as an ocean CDR approach.

Harmful Algal Blooms

All aOIFs are intended to produce changes to community composition as a consequence of adding iron. Of concern to many has been the possible increase in the abundance of Pseudonitzschia, a diatom genus known to produce the harmful neurotoxin domoic acid (DA) (Silver et al., 2010; Trick et al., 2010). This unintended consequence is often put forward in the public media as a reason not to continue with OIF as an ocean CDR approach (Allsopp et al., 2007; Harris, 2012; Tollefson, 2017). Looking more closely, there are few data to support this concern based upon direct measurements of DA, including studies by Marchetti et al. (2009), who did not detect increased DA production by Pseudonitzschia in response to the aOIF Subarctic Ecosystem Response to Iron Enrichment Study (SERIES) in the northeastern Pacific. Trick et al. (2010) point out that 6 of the 11 aOIFs produced increases in Pseudonitzschia abundances, so roughly half of the experiments had the potential to cause unintended harmful algal blooms (HABs). However, Trick et al. (2010) measured DA/cell in natural conditions, from a single northeastern Pacific profile (Ocean Station Papa) and saw little difference in the DA quota for incubations of Fe- and non-Fe–enriched cells (3–4 × 10^-6 picograms [pg] DA/cell). But using these estimates of low DA per cell and assumptions about transfer to the DA in water (not measured), they postulate a possible toxigenic response to an aOIF deployment, if conducted at a scale 100 to 1,000 times larger than any aOIF experiment thus far. In other words, with enough Pseudonitzschia, one might see a harmful response.

In another study to consider DA production, Silver et al. (2010) measured DA in stored cells from two natural settings and two aOIF studies (Southern Ocean Iron Experiment–South and FeExII). The results show highly variable DA per cell and only elevated DA in two settings, the natural northwestern Pacific (K2 site DA/cell = 0.9 ± 07 pg/cell for four samples) and in the aOIF

site SOFeX-S (0.9 ± 0.2 for two samples). Values at these two sites were much higher than after the aOIF experiment FeExII (0.04 ± 0.02 for two samples) with values equally low as at a natural site in the Gulf of Alaska (0.03 ± 0.07 for four samples). The point is that based upon these 10 samples, there is no clear evidence of additional DA per cell after aOIF relative to natural systems. It is only when the abundance of Pseudonitzschia increases in response to OIF that there may be conditions where harmful responses are possible. Silver et al. (2010) noted that “neurotoxin impacts at higher trophic levels, well known in shelf and coastal regions, have not yet been reported in open ocean systems.’’ They conclude that caution is warranted, but as with any ocean CDR approach, there will be unintended consequences that will be important to study, and thus be able to predict, if one were to move from aOIF research to large-scale implementation.

Co-benefits

Fisheries

OF has increasingly been proposed as a method for fisheries enhancement, in addition to, or in place of using it as a method for ocean CDR. At the most basic level, enhancements to the base of the food chain should lead to increases in fish stocks, at least if other variables remain similar. This concept was put forward early in the framework of reducing global hunger, based upon the addition of nitrogen to the ocean to increase production and thus conversion to seafood for human consumption (Jones and Young, 1997). In part, given the controversy surrounding aOIF and lack of a C credit market, commercial interest has shifted in several recent cases to this “ocean seeding” idea. For example, in 2012, the Haida Salmon Restoration Corporation (HSRC) asked the Haida Nation village of Old Massett in British Columbia to fund a commercial venture to deliberately release 100 tons of iron off Haida Gwaii as a means to enhance the local salmon fishery. Controversy remains about the legality of this effort (Tollefson, 2012; Wilson, 2013), and it was also lacking in the public release of data or peer-reviewed studies documenting the impacts. While after-the-fact study of remote sensing images and plankton sampling did document a bloom within the study area (Batten and Gower, 2014; Xiu et al., 2014), no links could be made to enhanced fisheries.

We are thus left with no evidence on the potential positive or negative impacts on fisheries of the 2012 event, though follow-on proposals for aOIF have been put forward with the specific goal to enhance the local fisheries. In one such case, Oceanos⁴ is proposing an Fe addition in the Humboldt Current area in the territorial waters off Peru. Whereas links further up the food chain may be impossible or at least difficult to demonstrate with commercial-scale OIF, natural OIF events may provide clues as to the possible link between OIF and fisheries enhancements.

One of the best-documented natural OIF events that has been tied to fisheries is the 2008 Kasatochi volcanic eruption off the Aleutian archipelago. Hamme et al. (2010) documented a large-scale biogeochemical response of a doubling of surface chlorophyll over an area of 1.5 to 2 × 10⁶ km² and an observed decrease in surface pCO₂ by 30 µatm (8 percent) and increase in pH at Ocean Station Papa from 8.08 to 8.13. This productivity enhancement and decrease in CO₂ was attributed to the response to the addition of iron from Kasatochi and resulted in what they estimate to be a C export event on the order of 0.01 Gt C (0.04 Gt CO₂). Olgun et al. (2013) further studied the release of iron from the volcanic ash and supported their findings that enough iron would have been added to support the enhanced productivity seen by remote sensing.

The link between the Kasatochi event and enhanced fisheries, however, remains controversial. Parsons and Whitney (2012) were the first to suggest that the volcanic iron induced a massive diatom bloom in the Gulf of Alaska that enhanced the food supply for adolescent sockeye salmon,

___________________

⁴ See http://oceaneos.org/.

leading to one of the strongest sockeye returns on record in 2010 for the Fraser River. Olgun et al. (2013) noted that this diatom bloom could support a larger zooplankton copepod food source for these juvenile salmon. From the timing of the bloom and magnitude of the response, they thought it was “very plausible” that the eruption enhanced salmon survival, though they point to several other factors that can affect ocean survival of salmon. McKinnell (2013) looked more broadly at Sockeye salmon spawning success in the Fraser River and challenged whether the volcanic event was the cause. They present a case that the survival was unremarkable in the historical record for the Fraser River, and that several other factors refute this idea, such as that the region with the anomalous chlorophyll enhancement is not where the juveniles migrate, and that no other salmon from that feeding region had unexpectedly high returns.

It is not surprising that a link between short-term OIF enrichments and fisheries are hard to document given the episodic supply of iron during a volcanic eruption and the subsequent enhancement of fish stocks years later. That decoupling in time and the wide range of processes that impact fisheries will make it difficult to attribute a positive fisheries co-benefit to a local event or sustained regional OF. And finally, if the goal of OF is enhanced fisheries for human consumption, then the ocean CDR benefits decline as carbon is returned to the atmosphere via respiration of food supply on land.

Ocean Acidification

In terms of other co-benefits, if the consequences of OF are a reduction in surface ocean DIC, as seen in 11 of the 13 aOIF experiments (Yoon et al., 2018), this would result in a pH increase and thus a decrease in surface ocean acidification (OA), at least temporarily during the drawdown period. Since OA is considered to be detrimental to carbonate-producing marine life in particular (Doney et al., 2020), this would be a co-benefit by maintaining or increasing pH over scenarios without OF. Using a simple ocean model, Cao and Caldeira (2010) predict that the impact of OIF given an extreme scenario of complete surface phosphate removal would reduce atmospheric CO₂ by 130 ppm, but increase surface ocean pH by only 0.06, relative to the same emissions without OIF.

Interestingly Cao and Caldeira (2010) further emphasize that this surface OA decrease, or co-benefit, would be accompanied by further acidifying the deep ocean, as also expected for other approaches of moving atmospheric CO₂ into the deep ocean (Reith et al., 2019). This could have negative impacts on the growth of deep-sea corals, as well as the metabolic processes of deep-sea biota in general (e.g., Siebel and Walsh, 2001). In effect, OF might be a co-benefit for surface corals and shell fisheries, but a shift to less favorable conditions in the deep ocean, similar to the arguments for potential negative impacts of mid-water and deep-ocean oxygen decreases in response to OF (see General Considerations). Oschlies et al. (2010a) inferred a pH decline by more than 0.1 units over large parts of the mid-depth Southern Ocean after simulated multidecadal OIF.

Lacking longer or larger-scale aOIF studies to directly examine deep-ocean impacts, studies of natural OF systems may provide another line of evidence regarding the impacts of OIF on deep-sea biota. When comparing a naturally Fe-enriched setting versus nearby controls off the Crozet Plateau in the Southern Ocean, the Fe-induced increased supply of organic carbon to the seafloor led to greater densities and biomass of deep-sea animals (Wolff et al., 2011). In fact, a similarity in deep-sea ecology in the Fe-enriched site and the productive northeastern Atlantic was noted, with the suggestion that aOIF could similarly increase the benthic biomass and species composition.

No matter what the impact of OF on the deep sea, it should be noted that what deliberate and large-scale OF would do is essentially speed up the natural processes that are already happening, under any current scenario of enhanced CO₂ in the atmosphere. For example, in one emissions scenario, 40 percent of fossil-fuel CO₂ would be stored in the ocean with OIF by 2100 versus 27 percent without OIF (Cao and Caldeira, 2010).

Cost and Energy

Ocean fertilization approaches leverage mass ratios between nutrients and organic carbon (Table 3.2) such that the costs of raw materials could be relatively low when normalized to the mass of CO₂ removed. The costs in Table 3.2 do not include other parts of the supply chain (i.e., transport, loading, and addition to the ocean) or monitoring. It is clear that for macronutrient fertilization (N, Si, P), the amount of macronutrient added would be much greater than OIF, and hence the raw material costs are greater as reflected in the cost per ton of CO₂ removed. Similarly, to stimulate CO₂ removal at a meaningful scale (~1 Gt CO₂/yr), then the N and P production would be equivalent to 30 percent to 40 percent of current markets, whereas OIF would consume <0.1 percent of the current Fe market. Thus the challenge of obtaining iron and its land-based impacts from mining would be far smaller than for other macronutrients and ocean CDR methods such as alkalinity enhancement that have far greater material needs and hence impacts on land (see Chapter 7). Furthermore, phosphorus is a nonrenewable resource and its use in large-scale OF would compete with its use for agriculture.

The deployment costs for spreading nutrients in the ocean is also relatively low, especially in the case of iron where relatively small amounts are needed. For instance, the HSRC in its 2012 project chartered a fishing vessel to put 100 tons of iron a few hundred kilometers off the coast of Haida Gwaii at a reported cost of $2.5M (e.g., Biello, 2012). Early estimates for OIF that include both materials and delivery were as low as $2/t C ($0.5/t CO₂; Markels and Barber, 2002).

It is clear that costs for OF, however, are very sensitive to (1) the efficiency of nutrients added to stimulated C removal and (2) the ratio between carbon removed and that which is permanently sequestered at depth. On the basis of different Fe:C_seq efficiencies, Boyd (2008) illustrates that the cost per ton for OIF can vary from <$3/t C_seq to >$300/t C (<$1 to >$80/t CO₂, Figure 3.5), providing a best estimate based on aOIF experiments of $30–300/t C ($8–80/t CO₂). Other estimates at a larger scale suggest that the costs of aOIF could be as low as <$10/t C (Harvey, 2008; Renforth et al., 2013). The HSRC project mentioned earlier did not include monitoring and verification of C storage, but using a modest Fe:C ratio of 1,000, one can estimate a cost of $25/t C sequestered. A key research question is therefore to better predict and quantify these Fe:C_seq ratios, which will be

TABLE 3.2 Ratios of Nutrients to CO₂ Removed and Market Pricing and Production Comparisons

^a Expressed as t CO₂ for comparability (assuming a molar ratio of C:Si:P:Fe of 106:15:16:1:0.1-0.001 [Brzezinski, 2004] to derive maximum removal rates per ton of nutrient added).

^b These represent material costs per theoretical ton of CO₂ removed only, not the levelized cost of net C removal (see below for more on costs).

^c See https://www.intratec.us/chemical-markets/ammonium-nitrate-price.

^d See http://www.microsilica-fume.com/silica-fume-price-per-ton.html.

^e See https://www.indexmundi.com/commodities/?commodity=rock-phosphate&months=60.

^f See https://tradingeconomics.com/commodity/iron-ore.

set largely by both the bioavailability of the added iron and the extent of shallow remineralization of sinking POC flux that is stimulated in response to the Fe addition.

As with any ocean CDR approach, monitoring intended and unintended consequences to include changes in geochemistry beyond carbon and changes in ocean ecology requires additional costs that are rarely quantified. Perhaps the best way to estimate these costs is to scale costs based on prior research studies and aOIF field programs. These field studies looked at changes beyond the C balance, including shifts in plankton productivity and community structure; the consequences of other limiting macronutrients; and in the waters below the in situ Fe addition and in control sites, changes to oxygen, N₂O, CH₄, DMS; and other potential consequences, such as noted above, and including the presence of HABs. No official accounting is available, but budgets for SOFeX, for example, the last U.S. experiments in 2002 were on the order of $10M for two aOIF field deployments.

A more recent cost accounting for a research study of the BCP can be taken from NASA’s EXport Processes in the Ocean from Remote Sensing (EXPORTS) program.⁵ In the North Atlantic in 2021, EXPORTS used three ships and several of the latest autonomous platforms similar to that which would be needed for new in situ demonstration projects (see Monitoring and Verification). Although EXPORTS did not add iron, it looked at the fate of carbon in natural settings and ecological and community structure, with shipboard measurements largely over the course of 1 month. Using this more recent field experiment as an example, costs ran into $15M to $20M per field site, which included measurements of physical, biological, and geochemical processes, including C stocks and rates, such as C uptake, export, and remineralization. This was still of relatively modest duration and size (1 month, 10,000 km²) but it was conducted as a Lagrangian time series much like would be needed for CDR studies. The cost would be similar to scale appropriately for more comprehensive, CDR-focused aOIF science experiments and pilot-scale demonstration projects. No one is suggesting that an operational CDR approach would take the same level of detail and hence cost

___________________

⁵ See https://oceanexports.org/.

as much as a research program, but it is clear that a research program to track the consequences and fate of C, including ecological impacts, will readily exceed the cost of simply deploying the iron.

Governance

The legal framework for ocean CDR is discussed in Chapter 2. Many of the international and domestic laws discussed in that chapter could apply to nutrient fertilization.

Nutrient fertilization is the only ocean CDR technique for which a specific governance framework has been adopted at the international level. Specifically:

• In 2008, the parties to the Convention on Biological Diversity (CBD) adopted a nonbinding decision recommending that governments take a “precautionary approach” and refrain from engaging in nutrient fertilization, except for “small scale research studies within coastal waters” (Para. C4, 2010). The decision further states that small-scale research studies may only be authorized “if justified by the need to gather specific scientific data,” and should be subject to “thorough” review and “strictly controlled.” This was reaffirmed in a separate decision issued in 2010 (Para. 8w, 2010).
• Also in 2008, the parties to the London Convention and Protocol adopted a nonbinding resolution stating that nutrient fertilization projects “should not be allowed,” unless they are undertaken for the purposes of “legitimate scientific research” (Art. 3-5, Resolution LC-LP.1, 2008). The parties subsequently adopted a framework for reviewing research proposals to evaluate their “scientific attributes” and potential environmental impacts (Resolution LC-LP.2, 2010).
• In 2013, the parties to the London Protocol agreed to an amendment that effectively prohibits nutrient fertilization, except for research purposes (Resolution LP.4(8), 2013).

Note that neither the 2008 decision under the CBD nor the 2008 resolution under the London Convention and Protocol is legally binding. The 2013 London Protocol amendment will become legally binding once ratified by two-thirds of the parties to the Protocol, but that has not yet occurred and appears unlikely in the near future. Even if the ratification threshold is met, only parties to the London Protocol that have ratified the amendment would be bound by it. Notably, the amendment would not bind the United States, which is only a party to the London Convention and not the London Protocol (see Chapter 2).

Researchers (e.g., Webb et al., 2021) have concluded that the United States could, at least in some circumstances, undertake or authorize ocean CDR projects involving the addition of materials to ocean waters (presumably including nutrient fertilization) without violating the terms of the London Convention. Projects would, however, have to comply with applicable domestic law. Table 2.3, in Chapter 2, identifies key U.S. federal environmental laws that could potentially apply to nutrient fertilization. No detailed research has been conducted on the application of those and other U.S. laws. We understand that such research is being conducted as part of an ongoing project, led by Columbia University researchers, but they had not published their findings at the time of writing.

Ocean CDR is likely to face social acceptance challenges given its divisive history. With open-ocean activities, stakeholder analysis and defining relevant publics can be challenging. Moreover, if OF receives social license on a local or community scale, it may not receive social license globally; conversely, global social license will not translate to community-scale acceptance. One way scientists can provide the basis for robust societal debate is to provide research that aligns with open-science principles. For the research to assess OF as a CDR approach, there would need to be transparency in planning and open access to all data; results disseminated at open meetings and in peer-reviewed journals in a timely manner; full compliance with international laws; study of

intended and unintended ecological effects in both surface and subsurface waters; and assessment of impacts beyond the study area and extrapolation to global scales. If academic, private, and government scientists are involved in collective experiments, there needs to be the ability to maintain independence in their ability to report data and interpretations thereof. Finally, if research activities are funded by commercial interests, there is even a greater need for a clear and transparent code of conduct to ensure that results are considered unbiased and accepted by the public.

3.6 SUMMARY OF CARBON DIOXIDE REMOVAL POTENTIAL

The criteria for assessing the potential for ocean nutrient fertilization as a feasible approach to ocean CDR, described in Sections 3.2–3.5, is summarized in Table 3.3.

3.7 RESEARCH AGENDA

While OF, and OIF in particular, has a longer history of scientific study than all other ocean CDR approaches, these studies were not intended as a test of the feasibility and cost of OIF for large-scale CDR and climate mitigation, or to fully assess environmental impacts at deployment scales. Modeling studies, on the other hand, often focused on the sequestration potential, environmental impacts, and, sometimes, cost estimates of large-scale deployment. Efforts to bridge local experimental scales and global modeling scales (e.g., Aumont and Bopp, 2006) should be encouraged to help maximize the information gained. The earlier OIF studies do serve as a pilot-scale work that can be used to pose several key questions that would be answered with additional laboratory, field, and modeling studies as part of a portfolio of ocean CDR research activities. These research questions can be grouped broadly by the ones on “will it work” related to C sequestration effectiveness and “what are the intended and unintended consequences” related to changes to ocean ecosystems that are an intended part of responsible ocean CDR of any type.

These pilot studies taught us that aOIF experiments would need to be significantly longer and larger than earlier ones that used 0.3–4 tons of iron (II) sulfate (FeSO₄) and covered 25–300 km² with ship-based observations lasting 10–40 days. A demonstration-scale aOIF field study might need to add up to 100–1,000 tons of iron (using planes, or autonomous surface vehicles), cover up to 1 million km² (1 percent of HNLC waters), and last for at least an entire growth season with multiyear follow-up. This would be a scale similar to the Kasatochi volcanic eruption in the Gulf of Alaska (see Fisheries) that caused no permanent harm, but was of a size that it could be readily tracked and pH and CO₂ impacts could be measured, and it provided a regional C loss out of the surface of 0.01–0.1 Gt C (0.04–0.4 Gt CO₂) (Hamme et al., 2010; Longman et al., 2020).

If these demonstration projects were conducted in different HNLC settings (Southern Ocean, Gulf of Alaska, Equatorial Pacific) and LNLC as well, one could document and reach a predictive understanding of the differing ecological and biogeochemical responses. Documenting the CDR impact and understanding and minimizing any long-term ecological damage would be key to the success of any aOIF demonstration project. Several recent large-scale studies of the BCP, such as NASA EXPORTS, can be used to estimate costs ($15M–$25M per site) and duration (3–5 years) of any one such field study, resulting in a demonstration-scale research program and its synthesis and modeling thereof, with a total budget surpassing $200M over 10 years (see summary of research costs in Table 3.4).

On the path to such demonstration projects there is also work to be done in labs, mesocosms, and on smaller scales. In addition, model improvements to better capture the cycling of not just carbon but also iron (e.g., Black et al., 2020) are needed, with careful attention to permanence issues and downstream impact that may only or best be captured by models, and include realistic export and all pathways of the BCP. Observations and models of naturally enriched OIF settings (islands

TABLE 3.3 CDR Potential of Ocean Nutrient Fertilization

and volcanic events primarily in HNLC regions) have also proved useful in gaining an understanding of sequestration efficiencies and longer-term biological responses and should continue. Consideration of co-benefits (e.g., fisheries) and other non-CO₂ GHGs (N₂O, DMS, CH₄, and O₂) would also be needed. Finally, the technology needed to monitor OF is growing, but investment in new designs of autonomous vehicles with biogeochemical sensors, such as on bioARGO floats and gliders, and new optical/camera systems and particle collectors would allow for better tracking of C and other responses. On the research side, new ways to query the genetic shifts in the marine food web could be quite informative. All of this work would require resources on the scale of the current Ocean Observatories Initiative nodes, or Long-Term Ecological Research sites, and/or be put in place as enhancements to the decadal surveys already under way in ocean sciences.

Also necessary are research activities into the social costs and public acceptance of the deliberate manipulation of the ocean commons. While the legal framework under which this research could be conducted has been set in place by the London Convention and London Protocol, it has not been tested, and there are several unresolved questions about its application. Thus we do not know whether the current international agreements would work to allow research but limit unwanted practices in terms of study design, transparency, claims of C credits, and ecosystem enhancements or detriments. Again, these social and legal issues would only grow in importance with any large-scale ocean CDR deployments and will affect the ability to move forward with funding and permitting, even for research. As noted above, while some of these activities are small scale and can be done by individuals, many of the outstanding research questions will require an emphasis on demonstration-scale studies that are larger and longer than done previously. The outcome of studies on these demonstration scales is essential if we are to deploy any or a combination of different ocean- and land-based CDR approaches. Finally, studies are needed to better define costs and benefits of OF so that we can reliably predict the consequences and scales over which the benefits outweigh the costs relative to doing nothing, and against other land- and ocean-based CDR approaches.

In summary, some specific examples of research needs include:

1. C sequestration efficiencies. This is a key factor in setting impact on C storage and hence climate. These efficiencies are set by ability for a given amount of nutrient (iron) to enhance C sequestration. New topics for studies would include the following: Can we increase bioavailability of iron and reduce removal by Fe scavenging and thus enhance phytoplankton growth? Can we enhance C and Fe export to the deeper ocean to increase permanence? How can we observe or estimate or model sequestration times in a manner robust enough for accounting purposes? What level of robustness or reliability is required? Are there better methods to track carbon and added iron? Can we engineer designs for improved Fe delivery at larger scales? How do we optimize the deployment to increase production and export or durability (location, season, duration, and continuous versus pulsed delivery)? What are the consequences of multiple nutrient and other limitations to C sequestration responses? How do we manipulate conditions to get food web response that maximizes C sequestration (fast and efficient sinking, low grazing)? How do we improve our monitoring technologies to track consequences of OF to not only carbon, but also full biogeochemical responses and through the food chain?
2. Ecological responses. In addition to the intended additional C sequestration, unintended and unexpected consequences have occurred that we need to know more about: What would the impact of OF be on planktonic food webs? Would fisheries be enhanced? What are impacts on higher marine trophic levels and how would one recognize them? Would responses include HABs that threaten open-ocean or coastal systems? Would production of other GHGs enhance (DMS) or reduce (N₂O) the climate impact of CDR? Would downstream ecosystems be limited by intended macronutrient removal? What are responses to OF in low-nutrient settings such as the rates if N₂ fixation? What are consequences to geochemical conditions in the deep ocean that may alter deep-ocean ecosystems (O₂, changing DIC or pH)?
3. Social acceptance, governance, and deployment costs. For OF, some of the pressing advancements needed are the following: Are the London Convention and London Protocol sufficient to regulate research and demonstration-scale experiments (possibly), and eventual larger-scale deployments with potential downstream impacts (not likely)? What is the best code of conduct that should be followed for research on OF (and other CDR approaches)? Would OF research be considered acceptable and reversible by a society that is experiencing climate change consequences very differently (i.e., benefits of OF may be separated greatly in time and space from negative consequences of climate change)? If OF was included in C removal markets or platforms, where and to whom would the benefits go, and what would the risks be? What could be the harms and benefits of different policy models for OF deployment?

Many of these research agendas will need to include modeling, whether it is of the localized field experiments or global ecosystem–biogeochemical models to assess the long-term and remote consequences. These models will need to include the cycling of nutrients, including iron. However at present there is low confidence in model projections of Fe distribution and fluxes (Tagliabue et al., 2016; Black et al., 2020). Full earth systems models will be needed to link changes in ocean physics and biogeochemistry to atmospheric CO₂ and climate (e.g., Bonan and Doney, 2018). High enough resolution will be needed to include at least mesoscale physical interactions. Long-term models are needed to assess the full consequences of downstream impacts. Finally, the need for experimental and observational data to validate and verify models will be essential if we are to use models to extrapolate to scales that are not readily measurable in ocean sciences.

3.8 SUMMARY

For the purposes of reducing atmospheric CO₂, this chapter outlines the state of our knowledge (Table 3.3) and key remaining questions that need responsible and transparent study to advance OF research, in particular OIF (Table 3.4). Given that OIF mimics natural systems, much can be learned from studying Fe-enriched “hot spots” near islands or after volcanic events. But studying natural systems is not sufficient to predict outcomes of deliberate OIF as a CDR approach. Unlike many other CDR methods, an international framework has been proposed for evaluating demonstration projects, but it is not legally binding, and there remain many unresolved questions and gaps in the governance framework. Future projects could be 10–100 times larger than prior aOIF experiments, adding hundreds to thousands of tons of iron and resulting in blooms of 10⁵–10⁶ km² that could be tracked and studied for longer than a single annual growth cycle. The potential for net C sequestration of OIF is large enough (Gt C sequestered for >100 yr) and Fe needs are small enough—0.1 percent of annual Fe ore production is 10⁶ t/yr, which could lead to 1 Gt C/yr sequestration (3.7 Gt CO₂) with Fe:C_seq efficiencies of only 1:1000—to warrant additional study. This biotic approach has relatively high scalability and low costs for deployment, though challenges would include verifiable C accounting and, as for most ocean CDR at scale, careful monitoring of intended and unexpected ecological effects up and down the food chain.

Even if the costs or impacts prove unacceptable for large-scale deployment globally, many companies are already suggesting OF as a way to enhance fisheries, and so having these studies in place could help to inform regulation of the scale and locations over which OF may be allowed or not. It is therefore important to conduct these studies as a basis of evidence for policy makers to contain entrepreneurs and other organizations that do not choose to follow international standards, or plan, organize, and report results in a transparent manner that upholds scientific standards and complies with international protocols. The relatively low cost of entry to initiate an Fe-induced bloom, $1M–$2M for hundreds of tons of iron and a small ship, make OIF an approach that does not require huge investments, making it prone to misbehavior by individuals or small organizations or companies. If done well, OF may be an imperfect action done for a good purpose, such as for fisheries enhancement or CDR, but if done poorly or outside of regulated and transparent studies, it has the potential to leave a legacy of unknown and possibly unacceptable impacts. Thus this investment in OF research is warranted whether one believes that it can work on large scales for CDR, or if one simply wants to regulate misuse of the global ocean commons.

TABLE 3.4 Research and Development Needs on Ocean Fertilization

NOTE: Bold type identifies priorities for taking the next step to advance understanding of ocean fertilization as an ocean CDR approach.

This page intentionally left blank.