4

Artificial Upwelling and Downwelling

4.1 OVERVIEW

The vertical movement of water in the ocean, termed upwelling and downwelling, acts to transfer heat, salt, nutrients, inorganic and organic carbon (C), and energy between the well-lit surface ocean and the dark, nutrient- and carbon dioxide (CO₂)-rich deep ocean. Since the 1950s, researchers have sought to artificially stimulate these physical transport processes to geoengineer localized regions of the ocean. For instance, wave-driven or density-driven artificial upwelling (AU) has been proposed as a means to supply growth-limiting nutrients to the upper ocean and generate increased primary production and net C sequestration. The latter outcome (increased C sequestration) would require that the biological production of carbon exceed the delivery of dissolved inorganic carbon (DIC) from the upwelled source water. Purposeful upwelling has also been proposed as a mechanism to sustain fisheries and aquaculture (Williamson and Turley, 2012), to generate energy (Isaacs et al., 1976), to provide a source of cold water for seawater-based air-conditioning (Hernández-Romero et al., 2019), and even to prevent the formation or severity of typhoons (Kirke, 2003). In contrast, the purposeful downward transfer of less-dense, oxygen-rich surface water, has been suggested as a mechanism to counteract eutrophication and hypoxia in coastal regions by ventilating oxygen-poor water masses (Stigebrandt et al., 2015; Feng et al., 2020). Both mechanisms may, under certain circumstances, act to enhance the oceanic sequestration of atmospheric CO₂.

According to criteria described in Chapter 1, the committee’s assessment of the potential for AU and artificial downwelling (AD), as a CO₂ removal (CDR) approach is discussed in Sections 4.2–4.5 and summarized in Section 4.6. The research needed to fill gaps in understanding of AU and AD, as an approach to durably removing atmospheric CO₂, is discussed and summarized in Section 4.7.

4.2 KNOWLEDGE BASE

Despite it being known for decades that AU could be achieved without an external C-based energy source (Stommel, 1956; Isaacs et al., 1976), to date all oceanic tests of AU detailed in the

peer-reviewed literature have been relatively small in scale with deployments typically less than a week and impacting an area no larger than tens of kilometers (Huppert and Turner, 1981; Liu and Jin, 1995; Ouchi et al., 2005; White et al., 2010; Maruyama et al., 2011; Pan et al., 2016; Fan et al., 2020). Enhanced oceanic C sequestration has never been documented in any sea trials to our knowledge. Pan et al. (2016) summarized the various types of AU mechanisms, spanning wave-driven pumps, electrical pumps, salt fountains, air-bubble pumps, and air-lift pumps (Table 4.1 and Figure 4.1). In all cases, if net C sequestration were to be achieved, the elemental composition of the source water would need to be carefully considered (e.g., Karl and Letelier, 2008) given that the proximate limiting nutrient varies across the global ocean as does the stoichiometry of potentially limiting elements and the concentration of DIC and alkalinity with depth. The energy utilization and pumping efficiency is also an important consideration and varies between pumping mechanisms (Table 4.1). In short, as opposed to ocean iron fertilization (OIF; Chapter 3), there exists no proof of concept that AU could act to sequester carbon below the ocean pycnocline.

In an open-ocean test in 1989–1990, Ouchi et al. (2005) report successful delivery of deep, nutrient-rich water from 220 meters, which promptly downwelled below the euphotic zone due to the high density of the upwelled seawater. In 2008, White et al. (2010) conducted sea trials of a commercially available wave pump with the aim of stimulating a two-phased phytoplankton bloom (e.g., north of the Hawaiian Islands in the North Pacific subtropical gyre). Sensors on the wave pump documented delivery of deep cold water to the surface for a period of ~17 hours after which catastrophic failure of the pump materials occurred. More successful tests of AU have been completed in semi-enclosed bodies of water such as Sagami Bay in Japan where an electrical pumping system moored to the seafloor was operated for ~2 years. This system delivered deep water from 200- to 20-meter depth and generated enhanced concentrations of picophytoplankton and nanophytoplankton (Masuda et al., 2010); C fluxes were not measured. While operational, the main limitations of this approach were reported to be high energy costs as well as construction and maintenance expenses (Pan et al., 2016). Additional successful trials were held in the western Norwegian Fjord in 2004–2005 using air-bubble pumps and uplift pumps (Aure et al., 2007; McClimans et al., 2010), generating flow of nutrient-rich deep water at a rate of 2 m³/s, which led

to a tripling of the concentration of the phytoplankton pigment chlorophyll-a in a 10 km² plume; again no measurements of C flux were made.

And so, in principle and at least in semi-enclosed water bodies where the impact of wind and waves are lessened, AU is a viable means of fertilizing the ocean with growth-limiting nutrients. In practice, there have been no proof-of-concept sea trials to confirm or deny the potential for net C sequestration, nor has any technical design proven to be sufficiently robust for long-term deployments in the open ocean. Were this technology to be utilized for CDR, pilot studies would first need to be conducted in the open ocean. Such studies would need to measure C fluxes as well as evaluate compensating flows that may lead to decreased thermocline stability and changes in surface temperature patterns that could lead to changes in atmospheric circulation and weather.

In contrast, there have been a wide range of modeling studies aiming to assess the efficacy of AU either as a means of ocean CDR or in support of aquaculture. By and large, these models suggest that AU would be an ineffective means for large-scale C sequestration (Dutreuil et al., 2009; Yool et al., 2009; Oschlies et al., 2010b; Keller et al., 2014; Pan et al., 2015) and would require a persistent and effective deployment of millions of functional pumps across the global ocean (Yool et al., 2009). For example, Oschlies et al. (2010b) use an Earth system climate model to simulate the impact of a network of “ocean pipes” of varying lengths and generating variable upwelling velocities. These authors conclude that even under “the most optimistic assumptions’’ AU would lead to sequestration of atmospheric CO₂ at a rate of about 0.9 petagrams (Pg) C/yr (most of this on land due to the slowdown of respiration in terrestrial soils under a cooler atmosphere), generally below the targets needed to mitigate anthropogenic emissions (>1 Pg C/yr).

Moreover, models indicate that upwelling of deep water may lead to cooling of the lower atmosphere, reduced precipitation, promotion of ocean acidification in certain regions, and even enhancement of terrestrial C storage (Keller et al., 2014). Although not yet considered in Earth system models, electrical pumps and air-bubble pumps have also shown promise as upwelling tools (Pan et al., 2016); however, these methods come with added energy costs and so are less likely to be effective in CDR.

Beyond direct ocean CDR, AU has been proposed as a means to support local aquaculture from fisheries to seaweed farms, which may have the cascading effect of enhancing ocean C sequestration (see Chapter 5). In a recent sea trial, Fan et al. (2020) demonstrated that a rigid AU system supplied sufficient nutrient-rich water to enhance seaweed growth in an enclosed bay. While this is but one trial, coupling of AU and large-scale seaweed farming could feasibly provide a means to offset upwelled CO₂. In contrast, models developed by Williamson et al. (2009) determined that nutrient levels injected via AU would be maintained at ~0.1 percent of their source concentrations, far below what would be needed to sustain ocean aquaculture. Hence, the nutrient supply rate and nutrient stoichiometry would need to be closely evaluated relative to aquaculture needs in order to determine potential efficacy.

AD, the engineered downward generation of vertical currents, has also been suggested as a means to enhance sequestration of dissolved and particulate organic carbon. Downwelling of surface water has never been tested in the field as a means of CDR; however, regional tests have been conducted to assess mitigation of hypoxia via downwelling (Stigebrandt and Liljebladh, 2011; Stigebrandt et al., 2015). Although the technology is promising, the efficacy and the biogeochemical consequences are less certain. AD could also be directly coupled with AU as a means to pump recalcitrant dissolved organic carbon to depth or prevent outgassing. As discussed above, upwelling of cold CO₂-rich water into the surface ocean comes with the risk of net outgassing if community production does not exceed the C inputs. Downwelling of recently upwelled water masses may reduce this risk, but again this is a wholly untested scenario. Modeling studies have indicated that modification of ocean downwelling to enhance the ocean’s solubility pump is “highly unlikely to

TABLE 4.1 Summary of Types of Artificial Upwelling and Their Advantages and Disadvantages

^a An air supply of 1 m³/min could uplift > 88 m³/min.

SOURCE: Pan et al., 2016.

ever be a competitive method of sequestering carbon in the deep ocean” due to impracticalities and costs (Zhou and Flynn, 2005).

Although not directly related to ocean CDR, AU has also been proposed as a means of providing energy and cooling and hence reducing terrestrial C emissions (e.g., ocean thermal energy conversion [OTEC]; Kim et al., 2021). Since the 1970s, it has been proposed that industrial-scale OTEC plants could pump high volumes of deep cold seawater to power turbines and generate electricity in an ecologically and economically sound manner to coastal and island communities.¹ Demonstration-scale plants (production on the order of hundreds of kilowatts) have even been established in the Hawaiian Islands² to provide power and desalinated water. Generation of power was one of the first proposed applications of AU (Isaacs et al., 1976), and certainly does have implications for C emissions, even if unrelated to oceanic CDR. Upwelling of deep cold water has also been proposed as a means of mitigating thermal stress in coral ecosystems (Sawall et al., 2020). Simulated AU experiments off Bermuda show early indications that controlled upwelling could abate coral bleaching during heat stress events potentially allowing corals to adapt to rising temperatures more steadily (Sawall et al., 2020). These examples are only meant to illustrate that while open-ocean applications of AU are unproven as a means of CDR, the principle of simulated upwelling has many applications.

In summary, AU and AD are proven means to transport water against concentration gradients without the need for C-based energy sources; small-scale pilot experiments conducted over the past few decades have shown that these technologies do work largely as expected, but the technology is not proven to be robust in the open ocean over timescales needed for CDR (longer than months). The larger challenge is not functionality, it is scaling, verification of CDR, and monitoring of ecological and biogeochemical responses to the magnitude, duration, and rate of perturbations. Undoubtedly, AU technologies are valuable as means to study ecosystem responses to nutrient disturbances (Masuda et al., 2010; Williamson et al., 2012); however, significant advances in technological readiness as well as durability and development of monitoring programs need to be made before this can be a reliable means of CDR. These issues are explored in more detail below. As done before, we also make comparisons to the other ocean CDR methods in this report and note when the same consequences would result from other less-well-studied ocean CDR methods.

4.3 EFFICACY

As noted above, there are no existing sea trials that have assessed C fluxes as a result of AU or AD. As a result, we simply do not have enough information about the long-term operation and efficacy of either AU or AD systems, and hence there has been a reliance on model simulations to assess potential efficacy. Nearly all model simulations suggest that large-scale deployment of ocean pipes would be a costly and ineffective means of ocean CDR with large uncertainty as to whether net C drawdown is attainable (Lenton and Vaughan, 2009; Yool et al., 2009; Keller et al., 2014). Model simulations of global-scale AU predict a range of perturbations to air–sea C flux, spanning <0 Pg C/yr (net outgassing, Dutreuil et al., 2009; Yool et al., 2009) to up to 3.6 Pg C/yr (Keller et al., 2014). In general, this is because most carbon exported from the surface is remineralized in the mesopelagic, and ocean upwelling would ultimately return this to the surface in excess of any enhancements in export production.

Yet, ocean physics and biogeochemistry are often more complex and nuanced than can be captured in even the most sophisticated models. There is a real need for a pilot-scale test of these technologies to assess the potential for AU and AD as either a component of CDR tools or as a

___________________

¹ See, for example, see https://www.makai.com/ocean-thermal-energy-conversion/.

² See http://nelha.hawaii.gov/.

means to sustain small-scale aquaculture or generate energy. Such pilot-scale tests would be where pumps can be operated autonomously for long periods of time with data transparency and international scientific cooperation, and the environmental risks and benefits can be more accurately weighted. If these were to be conducted, a determination of efficacy would be very similar to OIF in terms of cost and necessary components; however, the carbon potentially sequestered would need to be weighed relative to the upwelled carbon. Specifically, for AU to effectively sequester carbon from the atmosphere, we need to consider not only the C export ratios, that is, the ratio of inorganic carbon upwelled relative to the carbon sinking out of the upper ocean, but also the extent to which sinking organic carbon is attenuated with depth. As nutrients are delivered to presumably nutrient-limited surface waters, there is an expectation that there would be predictable changes in the phytoplankton community structure, with a shift to larger, faster-sinking cells (Karl and Letelier, 2008) that may increase export efficiency. Whether or not this would be the case is debatable because the nutrient stoichiometry of upwelled water as well as any concomitant changes in the mixed layer will also influence any shifts in community structure. The stoichiometry of limiting nutrient to CO₂ will also vary with depth and region. For this reason, the outcome of models upwelling water from a uniform depth across basins, as that discussed by Dutreuil et al. (2009), may not reflect the real potential of upwelling water masses from depth where we find the maximum deficit of CO₂ relative to other nutrients (see Karl and Letelier, 2008). The lack of spatial uniformity in the stoichiometry of growth-limiting elements relative to CO₂ is that the potential net C sequestration efficiency of a free-drifting upwelling pump may change over time. Moreover, the persistence or intermittency of upwelling will influence community structure and resultant changes in primary and export production.

Beyond the biogeochemistry, several additional research and development needs should be addressed before efficacy of these technologies for CDR can be determined. For example, siting analyses that evaluate potential wave energy needed for upwelling relative to the magnitude of nutrient injections and the physiological status of surface populations need to be conducted to determine ideal locations for pilot studies (Figure 4.2). This exercise is likely to reveal trade-offs between accessibility of coastal sites for maintenance and monitoring versus the relative permanence of C storage in other areas such as the highly stratified oligotrophic gyres. Lastly, it is important to note that the limited field studies have largely been constrained to coastal regimes with limited operational periods (shorter than weeks) and relatively shallow source waters and low upwelling rates (<0.1 m³/s), whereas model simulations have explored C sequestration potential using temporally and spatially extensive deployments with much deeper source waters and higher rates of upwelling (Figure 4.3). In effect, there is a gap between the technological readiness of AU and projected sequestration potential. Modeled flow rates far exceed those achieved in sea trials to date and so even the highly variable C export predictions that have been made should be considered unrealistic until AU technology could feasibly be deployed at demonstration scale.

Additionality

Downstream impacts of AU should also be further addressed because model simulations indicate the potential for ocean acidification, hypoxia, and changes in plankton successional patterns far afield from the site of upwelling. For example, Keller et al. (2014) estimated that AU would reduce ocean pH by up to 0.15 units below what is expected with the “business as usual” trajectory for ocean acidification as a result of anthropogenic CO₂ emissions. Additional potential consequences of large-scale AU are similar to OIF and could include production of greenhouse gases such as nitrous oxide, methane, or dimethyl sulfide, disturbance of benthic ecosystems, or mid-water oxygen depletion (Williamson and Turley, 2012). There is the potential for these ecosystem alterations to be felt far afield from the fertilization sites in both space and time, which would surely present

a challenge for verification of efficacy. Local depletion of nutrients could, for example, alter the productivity of regions downstream and shift global patterns of productivity, remineralization, and export (Aumont and Bopp, 2006).

The persistence and scale of AU also need further evaluation. Should pumping be intermittent or persistent? Baumann et al. (2021) indicates that the nutrient perturbation rates can impact the nutrient quality and sinking rates of the phytodetritus generated, which would affect permanence. Does AU need to be on the scale of millions of pumps deployed in the global ocean to be effective, as suggested by models? If so, this would have consequences for international shipping, fishing, and other unrelated activities at sea. Would there be any geochemical consequences if upwelling was stopped at some point? To this last point, prior models suggest termination of large-scale AU could result in a rapid net increase in global temperatures to levels even higher than in a world that had never engaged in AU (Oschlies et al., 2010b). Would AU lead to changes in global precipitation patterns or weather? Ricke et al. (2021) indeed suggest that variations in sea-surface temperature could lead to alteration of rainfall and drought. The full answers to all of these questions will surely require a carefully planned series of pilot or demonstration studies as well as incorporation of resultant data into Earth system models and evaluation of near-field and far-field biogeochemical effects.

Permanence

A commonly used time period to assess permanence of C sequestration in the ocean is ~100 years. For AU, this requires that exported organic matter exceeds the inorganic carbon delivered from source waters and that a significant fraction of particles are efficiently transferred below the winter mixing layer. If these criteria are not met, then permanence will not be achievable. Recent model results suggest that this ~100-year benchmark will not be widely achievable via ocean CDR; using an inverse ocean circulation model, Siegel et al. (2021) conclude that enhancement of ocean C export (e.g., via OIF or AU) will have a shorter-term influence on atmospheric CO₂ levels because ocean circulation and mixing will transport ~70 percent of the sequestered carbon back to the surface ocean within 50 years. This finding underscores the clear fact that the ocean’s response to the magnitude, duration, and rate of perturbations will vary widely across ocean basins and with time. It also suggests that ocean CDR approaches seeking to stimulate the biological pump may be a short-term solution to buy time but unlikely to support permanent (≥100 years) sequestration. This has implications for the representativeness and wider applicability of oceanic experiments, which are often necessarily limited in space and time.

For downwelling, it will be necessary to determine how to subduct buoyant organic matter and ensure that it does not return to the surface. Limited trials have proven the capability to upwell and downwell seawater against gradients in density, nutrients, and C resources. Yet again, with no “farm scale” tests of AU or AD, we cannot assess permanence of C storage. In lieu, we must currently rely on mechanistic model simulations of ocean biogeochemistry. Dutreuil et al. (2009) simulated deployment of 200-meter-deep pipes throughout the global ocean at a spatial resolution of 20° longitude and ~ 10° latitude operating at vertical velocities of 0.1 m/s and found that while biological productivity and export were enhanced as might be expected, particularly in the equatorial Pacific, the air–sea CO₂ flux declined significantly at the sites of AU as a result of the mixing of respired CO₂ into the surface mixed layer where it could exchange with the atmosphere. Dutreuil et al. (2009) conclude that “overall, our analyses demonstrate that the enhancement of biological productivity is never enough to compensate for the additional supply of DIC to surface waters” (see Figure 4.4). Using a different model structure, a biogeochemical model coupled to the Ocean Circulation and Climate Advanced Modeling (OCCAM) physical model with embedded ocean AU pipes of either 200 meters, 500 meters, or 1,000 meters, Yool et al. (2009) came to a similar conclusion. Over a 10-year simulation, strong regional heterogeneity was observed, and changes in the air–sea flux as a result of AU were found to be both positive and negative across the global ocean.

Reversibility is also an important consideration as the intentional upwelling of CO₂-rich deep water or downwelling of O₂-rich surface waters has the potential to significantly alter ocean ecosystems from the epipelagic to the benthic. It may be difficult or impossible to distinguish far-field changes in ecosystem structure and function from changes due to decadal shifts (e.g., El Niño-Southern or Pacific Decadal Oscillation–driven shifts in production patterns). More research is needed to determine the C sequestration potential of AU and AD and whether permanence is achievable, what ocean sites would best achieve permanence, and what the environmental risks and co-benefits would be.

Monitoring and Verification

An array of devices have been proposed to facilitate either AU or AD, including wave-powered systems, airlift and bubble pumps, and salt fountains (Pan et al., 2016). In a handful of limited sea trials (described above), these technologies have proven capable of vertically pumping seawater, even if the durability and longevity of the devices tested have not been sufficient for large-scale trials (e.g., White et al., 2010) and no tests to date have ever evaluated potential changes in C

sequestration at any temporal or spatial scale in situ. In lieu, prior tests have evaluated the “efficacy” of pumps as CDR tools based on whether surface chlorophyll was enhanced or whether water masses with the temperature and salinity signatures of the source water were detected. This is, of course, insufficient evidence of efficacy; verification cannot focus on evidence of enhanced growth of phytoplankton in the surface ocean because the fate of that material could be remineralized in the upper ocean. Rather changes in the sinking flux of particulate organic carbon into the deep ocean (below 1,000 meters) must be determined to assess efficacy and permanence of sequestered C pools. In addition to measuring efficacy for C sequestration, such pilot-scale studies would also need observations of ecosystem changes that might lead to co-benefits, such as enhanced fisheries, or negative impacts related to harmful algal blooms or alteration of food webs that are considered undesirable. Large-scale trials of AU or AD would need to employ proven technologies such as sediment traps, gliders, and instrumented profiling floats to verify C sequestration as a result of AU and AD. Verification cannot rely on proxies for production such as chlorophyll concentrations in the surface ocean.

4.4 SCALABILITY

Geographic and Temporal

The location and spacing of open-ocean pumps required for AU to serve as effective tools for CDR would first and foremost need to consider the deep-water nutrient stoichiometry (the ratio of

macronutrients to one another) and nutrient-to-metabolic DIC stoichiometry, both of which vary globally and with depth. Karl and Letelier (2008) predicted that in the North Pacific Subtropical Gyre (NPSG), C export could only exceed the upwelled DIC if a two-stage bloom were triggered, with the first stage characterized by growth of phytoplankton on the upwelled nitrate and the second stage following nitrate depletion and supported by nitrogen (N₂)-fixing cyanobacteria capable of consuming the residual phosphate (~0.05–0.5 mol/L depending on source-water horizon) in the absence of nitrate and leading to net C sequestration. Examination of the elemental stoichiometry of nitrate:phosphate:DIC in this region indicated that upwelling of water from 300 to 350 meters would be ideal to promote a two-stage bloom, optimally in summer months when diazotrophs were most abundant. White et al. (2010) sought to test this hypothesis using wave-powered pumps, and despite brief upwelling at rates of 45 m³/h, materials failures prevented a further assessment of the biological response. So although the Karl and Letelier (2008) hypothesis remains untested in the NPSG, it is still critical for open-ocean tests to evaluate the potential production and export relative to the magnitude of upwelled DIC and the nutrient stoichiometry needed for growth, which also varies globally with changes in community composition. Singh et al. (2021) show that N₂ fixation rates in long-term mesocosms (55 days) can respond positively to enhanced pulses of upwelling after a lag of several days and rapid depletion of nitrate, supporting the hypothesis of expected successional patterns in plankton communities that may be key to net CDR. In extrapolating bloom patterns resulting from AU, the rate of perturbation as well as the initial community structure in any impacted biome will also need to be considered, as suggested by Karl and Letelier (2008).

Using an ocean model framework, Dutreuil et al. (2009) found that AU would be particularly ineffective in iron (Fe)-limited regions such as the equatorial Pacific and Southern Ocean where the addition of limiting macronutrients leads to communities with an increased iron demand, resulting in a weak “fertilization effect.” While no regions were strong C sinks in response to simulated AU in that work, the subarctic Pacific was considered promising because it is a region where the upwelled water contains sufficient alkalinity to compensate for the additional supply of DIC to surface waters. In contrast, Yool et al. (2009) found the strongest effect of simulated AU in centers of subtropical gyres, albeit their model results also predict considerable spatiotemporal variability. If AU were confined to the tropics and C fixation efficiency was 2.2 percent relative to air–sea uptake, Yool et al. (2009) calculated that between 189 million and 776 million pumps would need to be deployed to increase ocean C sequestration by 1 Pg C/yr above current rates. Deployments on this scale would need to consider how a network of pipes would compete with (or complement) other needs such as transatlantic shipping routes and fishing activity. In summary, siting analyses will need to identify optimal nutrient, light, and wave conditions for growth and C sequestration potential as well as potential conflicts with other marine industries.

The temporal scaling of AU deployments also needs to be considered relative to the timescale of biological responses. Should pumping be continuous or pulsed? Are there regions where efficacy is seasonally dependent, for example, in the NPSG where a diazotrophic response may be central to net C export (Karl and Letelier, 2008)? What will be the potential lifetime or durability of a network of upwelling pumps? Can upwelling be conducted in a manner that prevents rapid downwelling and subduction of nutrient-rich water below the mixed layer or euphotic zone? Are there termination effects such as pressure effects or circulation compensations that might lead to rapid warming and maybe overheating? All of these questions should be considered in a research agenda and coordinated with high-resolution models that can account for complex interactions of ocean physics, chemistry, and biology as well as downstream impacts. These issues are explored further in the research agenda.

Impact Potential

As discussed in prior sections, model simulations of large-scale AU deployments diverge significantly in the sign, magnitude, and regionality of efficacy and potential permanence of CDR. In general, model simulations of the impact are consistent with the conclusion by Fennel (2008) that “controlled upwelling is unlikely to scale up and serve as a climate stabilization wedge as defined by Pacala and Socolow (2004), i.e., it would not sequester 1 Pg C yr^–1 over 30 yr.” Moreover, the sea trials needed to assess C export potential and downstream impacts have not been conducted, and further research and development are needed to achieve technical readiness. Large-scale deployments are expected to impact weather and climate, may have implications for ocean heat and oxygen content depending on the spatial footprint and upwelling rates or frequency, and have downstream impacts such as “nutrient robbing” and impacts to biodiversity and ecosystem function. On a more practical note, there are potential conflicts with global shipping routes and fishing efforts, again depending on the scale of deployment. None of these potential impacts has high certainty without “farm-scale” or “deployment scale” trials. Such trials should be coupled to feasibility studies to address upscaling potential of the technology as well as allow for adaptive governance of the research before further investments regarding CDR potential are made. The necessary components of a research agenda addressing CDR and AU/AD are described below in Section 4.7.

4.5 VIABILITY AND BARRIERS

Environmental Impact

Because of mass conservation, any up- or downwelling has to be balanced by down- or upwelling of waters elsewhere. The spatial patterns and controls of the balancing counterflow is still unclear, but a net effect of the induced vertical translocation of water parcels is a reduction in density stratification, similar to that of enhanced vertical mixing, essentially mixing heat downward against the mean stratification of the upper ocean. Changing the density structure of the ocean is expected to change ocean circulation on scales larger than the Rossby radius, that is, a few tens of kilometers. Theory and model studies predict an enhanced overturning circulation under enhanced vertical mixing and reduced stratification. Viewed globally, this will likely reduce the vertical gradient of DIC and, if nothing else changes, lead to a net outgassing of CO₂ from the deep ocean to the atmosphere. Not surprising, natural upwelling regions in the ocean are regions of elevated partial pressure of CO₂ in the surface waters and a general outgassing of CO₂ from the ocean to the atmosphere. Upwelling regions are also strongly influenced by the phase of natural climate oscillations such as El Niño/La Niña, which lead to decadal variability in C fluxes (Bonino et al., 2019).

If AU can stimulate upper-ocean biological production and subsequent export of organic carbon to the ocean interior, enhanced oxygen consumption and production of respiratory CO₂ as well as non-CO₂ greenhouse gases in regions underlying areas of enhanced productivity are to be expected. These have been identified in model simulations (Keller et al., 2014) and are similar to effects induced by other marine CDR methods that aim to enhance marine biological productivity. Even though biological production is not intended in the AD concept, it may well be stimulated by the fertilization effect of the compensating upward return flow.

Ecological impacts of AU or AD are a net cooling of surface waters and a net warming of subsurface waters. Model simulations of large-scale massive deployment (millions of pumps throughout the global ocean) suggest that subsurface waters may warm by a few degrees for a century-long deployment of AU (Oschlies et al., 2010b; Keller et al., 2014). In theory, subsurface warming would enhance microbial and geochemical remineralization rates and thus decrease net C flux to the deep

ocean (Cavan et al., 2019). This will affect metabolic rates of the mesopelagic region with likely shifts in remineralization profiles and associated vertical C fluxes.

While AU and AD will essentially lead to enhanced storage of heat in the subsurface ocean, part of this heat will be released back to the surface ocean, and hence to the atmosphere, upon termination of these CDR methods. In simulations with an Earth system model of intermediate complexity, this was found to generate higher global mean air temperatures after termination of AU than in a control experiment in which AU was never applied (Oschlies et al., 2010b).

Additional impacts that should be considered include pollution of the oceans if pump materials were to fail; depending on materials, this could introduce significant plastic, metal, and/or concrete pollution. If pumping were highly effective, there might also be practical impacts such as navigational hazards or ecological impacts such as biofouling and transport of invasive species, changes in light penetration, spectral quality of light penetration, and changes in surface heating.

Co-benefits

Several co-benefits may be realized with effective AU, including potential stimulation of the climate-cooling gas dimethyl sulfide (Taucher et al., 2017), localized lowering of surface ocean temperatures to prevent coral bleaching, and support of fisheries or aquaculture efforts (Kirke, 2003). The latter topic (see Figure 4.5) is an area of active research spanning mesoscale-based studies of food web changes to support fisheries (Taucher et al., 2017; also see https://ocean-artup.eu/vision) to offshore seaweed farms as part of a portfolio of solutions for nutrient remediation in, and C removal from, our oceans (ARPA-e, 2021b). These programs are nascent and not yet at the stage of evaluating technological readiness but should be critical to assessing the feasibility, effectiveness, associated risks, and possible side effects of AU as well as the potential co-benefits. Note also that if aquaculture yields are to be fairly considered a co-benefit, then the C budget of fisheries should also be considered. For example, Mariani et al. (2020) estimate that fossil fuel consumption and hence CO₂ emissions by fisheries are on the order of ~0.01 Gt CO₂/yr (~2.5 Mt C/yr) over the period of 1950–2014. This does not include blue carbon extractions (see Chapter 6). Also see Chapter 3 for additional discussion on the feasibility of marine aquaculture as a means for ocean CDR.

Cost

Model simulations of the efficacy of AU indicates that CDR with the potential of on the order of several Pg C per year would very likely require substantial expansion from small coastal pilot studies to sustained operations in deeper, offshore (>3 nautical miles from the coast) ocean habitats. To do so will require that scientists and engineers work together to develop upwelling prototypes capable of sustaining significant upwelling velocities (see Figure 4.3) with materials designed to resist biofouling and remain operational in a variety of sea states. Costs to do so will include materials for the pumps themselves, deployment costs, costs for development and maintenance of offshore monitoring and verification programs, any energy needed to power the pumps (e.g., ocean thermal energy conversion plants; Avery and Wu, 1994; Matsuda et al., 1998), as well as any costs for removal of pumps at the end of their life cycle or after sustained damage and any necessary maintenance. While there are a number of design specifications for AU devices with various dimensions and flow velocities (see Pan et al., 2016), Kirke (2003) estimated that a moored 500-m-deep, 12.9-m-diameter, wave-driven inertial pump made of carbon steel would cost $4.68 million in 2003 dollars plus the costs of floats, anchors, and mooring cables for an overall cost somewhere below $10 million; this value would be less if construction material were concrete or fabric, although flow rates and durability would be compromised from the author’s perspective. Johnson and Dicicco (1983) estimated the total cost for design of salt-fountain–style pumps (thousands to tens of thou-

sands of pumps) made of plastic and concrete suitable to support a 10-acre kelp farm in an optimal location (the Gulf of Mexico) and suboptimal location (the Pacific Ocean off Chile) to range from $24.2M to $139M in 1983 dollars with estimates scaling on pump dimensions and hence materials needs (Figure 4.6). These costs do not include installation or maintenance costs, or costs to measure C sequestration effectiveness or ecological impacts. Given the inflation rate between 1983 and 2020 of 167.02 percent (InflationTool, 2021), these costs would scale to ~$40M at a minimum, which might be considered as a lower bound for a demonstration-scale AU trial. On the low end of potential costs, a present-day estimate from a company called Ocean-Based Climate Solutions, Inc. suggests that fabrication, assembly, shipping, and ocean operations would cost ~$60,000 per 500-meter tube, which they estimate could sequester 250 t CO₂/yr; scaling to even 0.1 Gt CO₂/ yr and, neglecting costs of verification, would then require millions of pumps and hence tens of millions in costs. For comparison, the mesocosm-based research on ecological impacts of AU was funded at €2.5M in 2017 by the European Research Council (Ocean artUp, 2021), and the Advanced Research Projects Agency-Energy (ARPA-e) MARINER program in the United States has funded ~$50M of research “to develop the tools to enable the United States to become a global leader

in the production of marine biomass” which includes an assessment of AU technology (ARPA-e, 2021a). Given these limited cost comparisons, there would clearly need to be significant increases in research investments in AU to support even medium-scale deployments (hundreds of pumps).

Energy

The energy costs for AU would scale on the pump design being considered as well as any costs for deployment and recovery, monitoring, and production of the materials and supplies for the pumps as well as for verification and monitoring. There are no existing life-cycle analyses for a demonstration-scale project nor have any of the sea trials performed to date considered an energy budget for operations. These analyses would need to be a component of any proposed research and development programs aimed at CDR via AU.

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 AU and AD.

At the international level, the parties to the Convention on Biological Diversity (CBD) have adopted a series of decisions governing “geoengineering,” the definition of which is likely to encompass AU and AD.³ The decisions recommend that parties to the CBD and other governments avoid geoengineering activities that may affect biodiversity, except for “small scale scientific research studies . . . conducted in a controlled setting.”⁴ The decisions are not legally binding, however. The CBD itself arguably does not prevent countries from undertaking or authorizing AU and AD projects, provided that they comply with all applicable consultation and other requirements imposed by the Convention (Webb et al., 2021).

There is significant uncertainty as to whether AU or AD constitutes “pollution” of the marine environment under the United Nations Convention on the Law of the Sea (UNCLOS) or marine “dumping” under the Convention on the Prevention of Marine Pollution by Dumping of Wastes and Other Matter (London Convention) and associated Protocol (London Protocol). In Resolution LC/LP.1 (2008) the parties to the London Convention and Protocol agreed that “ocean fertilization activities” fall within the scope of those instruments. The Resolution defines “ocean fertilization” as “any activity undertaken by humans with the principal intention of stimulating primary productivity in the oceans,” which some have argued could encompass AU. However, the Resolution is nonbinding, and subsequent decisions by the parties have only applied to ocean fertilization activities that involve “the placement of matter into the sea from vessels, aircraft, platforms or other man-made structures.” Some scientists (e.g., Brent et al., 2019) have argued that AU and AD are not covered because they merely involve the transfer of materials from one part of the ocean to another and not the introduction of new materials.

The treatment of AU/AD under domestic U.S. law is similarly uncertain. Table 2.3, in Chapter 2, lists key U.S. federal environmental laws that could potentially apply to AU/AD. No detailed research has been conducted on the application of those and other U.S. laws. We understand

___________________

³ Report of the Conference of the Parties to the Convention on Biological Diversity on the Work of its Tenth Meeting, Decision X/33 on Biodiversity and Climate Change, Oct. 29, 2010 (hereinafter “Decision X/33”); Report of the Conference of the Parties to the Convention on Biological Diversity on the Work of its Eleventh Meeting, Decision XI/20 on Climate-related Geoengineering, Dec. 5, 2012 (hereinafter “Decision XI/20”); Report of the Conference of the Parties to the Convention on Biological Diversity on the Work of its Thirteenth Meeting, Decision XIII/4, Dec. 10, 2016 (hereinafter “Decision XIII/4”).

⁴ Para. 8(w), Decision X/33; Para. 1, Decision XI/20; Preamble, Decision XIII/4.

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.

4.6 SUMMARY OF CARBON DIOXIDE REMOVAL POTENTIAL

The criteria for assessing the potential for AU and AD as a feasible approach to ocean CDR, described in Sections 4.2–4.5, is summarized in Table 4.2.

4.7 RESEARCH AGENDA

Proof-of-concept field experiments are needed in open-ocean conditions to assess technological readiness to monitor biological responses to upwelling, determine C sequestration potential relative to upwelled macronutrients and inorganic C, and monitor or model local and downstream environmental impacts of AU and potential concomitant downwelling. There are several natural oceanic analogs that can inform our understanding of the CDR potential of upwelling, including eastern boundary currents and the Southern Ocean. Notably, these regions are characterized by large cell-sized phytoplankton communities capable of fast sinking rates, yet the upwelling of DIC tends to lead to net C outgassing (Takahashi et al., 1997; also see Chapter 5). As the technology for AU matures and field trials become practical, monitoring and verification plans need to be developed to assess the volumetric flux rate of upwelling, nutrient delivery rates, the biological response of the upper ocean, changes in the particle flux below the euphotic zone, and the air–sea CO₂ flux in both experimental and control stations. Necessary components of this flux include the following elements, summarized as a research agenda in Table 4.3.

Components

Necessary components of a complete research program include:

• Significant advances in marine engineering to develop durable pumping systems capable of sustaining long-term deployments (months to years) in a range of sea states at upwelling velocities sufficient for a sustained biological response. These systems also need to maximize mixing in the surface layer so that negatively buoyant, nutrient-rich, deep-ocean water does not immediately sink out of the euphotic zone.
• Parallel modeling efforts to estimate the feasibility of CDR potential globally and regionally given achievable upwelling velocities, elemental stoichiometry of deep-water sources, and potential biological responses. Feasibility studies would be based on model upscaling from results of technical trials and existing literature on biological response to nutrient perturbations. Mesocosm and laboratory experiments would help constrain biological responses. Efforts would help refine which regions of the global ocean would be optimal for a scaled-up research program.
• A robust monitoring plan focused on particulate and dissolved organic C export and air–sea gas exchange at the local site of upwelling as well as targeted downstream sites. This plan should span across water measurements, remote sensing, and high-resolution coupled physical/ecological models to assess large-scale and downstream impacts and the timescales and depth scales of sequestration.
• Monitoring that will assess the vertical extent of C export and any changes in remineralization length scales.

TABLE 4.2 CDR Potential of Artificial Upwelling and Downwelling

• Assessment of the CDR potential at a range of pumping frequencies including episodic versus continuous upwelling as well as any seasonal impacts on CDR potential and the optimal source-water horizons.
• A monitoring plan that would be able to estimate the additionality of sequestration—how much production and export would have occurred in natural phytoplankton communities in the absence of AU? The monitoring plan should be able to differentiate between the response to AU and the natural variability of the system, which will differ between regions (e.g., Fe-limited versus N-limited ecosystems).
• Siting studies that address potential conflicts with shipping lanes, fishing effort, and other ocean usage as well as regions where costs are minimized relative to source-water horizons (e.g., shallower pumps require reduced material costs).
• A data management plan with clear plans for data dissemination, accountability, and data transparency following the FAIR principles: findability, accessibility, interoperability, and reuse.⁵
• Complete life-cycle analyses of costs for materials, deployment, and local and downstream monitoring costs.
• Interactions with social scientists, legal experts, and economists to assess public perception, acceptability, governance, and cost feasibility as well as potential for coupling AU to macroalgae or fisheries production.

Cost and Time Frame of Research Agenda

Stable funding streams need to first be identified to build capacity and develop an active research program as well as begin the planning of complementary monitoring and modeling components necessary to evaluate environmental impacts, CDR, and additionality. Costs increase as we move from the technological readiness stages of pump development, deployment, and testing of operational needs (Can water be delivered at needed upwelling rates? Can pumps survive a range of sea states?) to large-scale deployment and monitoring of intended and unintended ecological effects. Based on prior cost estimates (i.e., Johnson and Dicicco, 1983), regional-scale networks of pumps could cost on the order of ~$40 million for technological development alone, which is

___________________

⁵ See https://www.go-fair.org/fair-principles/.

within the range of expected funding for CDR development funding in the United States and the European Union (see, e.g., Burns and Suarez, 2020). Given this cost magnitude and the current lack of technological readiness, model-based feasibility studies should lead the research agenda to identify optimal siting and scaling of pump networks and CDR potential. This should be followed by expansion of technological development and small-scale proof-of-concept studies intended to show durability of materials and achievement of necessary upwelling velocities coupled to life-cycle assessments of materials and deployment costs.

Environmental Impacts of Research Agenda

Upwelling of deep nutrient-rich seawater into the surface euphotic zone brings a source of DIC in addition to potentially growth-limiting elements. The ecological consequences may include stimulation of autotrophic growth and enhanced sinking of detrital carbon. Alternatively, if the photosynthetic drawdown of DIC does not exceed the carbon introduced in the source water, outgassing of carbon may occur, which is obviously contrary to sequestration. This latter potential is the most significant “Achilles heel” of AU as a CDR strategy. Otherwise, much like OIF (Chapter 3), nutrient additions via AU may lead to shifts in plankton community structure and, potentially, alternations of productivity of higher trophic levels including commercially harvested fisheries. Any monitoring plan should also consider the possibility of harmful algal blooms, other greenhouse gas production, and hypoxia.

4.8 SUMMARY

A range of studies have conclusively demonstrated that deep nutrient-rich seawater can be delivered to the surface ocean via a number of pumping mechanisms, each of which has different energy costs, deployment modes, and pumping capabilities. A recent summary by Pan et al. (2016) describes the current state of these technologies (see Table 4.1). These limited field trials have largely been conducted in short-term (less than weeks) experiments and primarily in coastal regimes, and none have yet verified enhanced C sequestration. A research agenda aimed at testing AU as a component of a CDR portfolio would principally need to address the long-term durability and efficacy of AU technology as well as the siting for sea trials. Since AU would deliver remineralized DIC as well as potentially limiting nutritional resources, it is key to understand the nutrient-use stoichiometry of the local plankton populations relative to the stoichiometry of deep-source waters (which varies widely across the global ocean); simply, C export flux would need to exceed the upwelled carbon. Additionally, the input of other elements, phosphorus or iron, for example, can also govern bloom dynamics and need to be considered. For example, in the oligotrophic North Pacific Subtropical Gyre, Karl and Letelier (2008) hypothesized that upwelling of water with excess phosphorus relative to nitrogen could trigger a two-stage bloom with net C sequestration being driven by the production of diazotrophic (N₂-fixing) microorganisms that may thrive after a primary pulse of nondiazotrophic plankton. While yet untested, this hypothesis underscores the importance of understanding C:N:P stoichiometry of source waters relative to plankton growth requirements as a driver of potential C sequestration exceeding that delivered by the upwelling process. And since nutrient stoichiometry varies across ocean basins and with depth, one would expect that the ecological consequences of AU would be site and depth specific and, perhaps, time dependent.

The current state of knowledge otherwise, via model simulations, indicates that even a persistent and effective deployment of millions of functional pumps across the global ocean would not meet CDR goals for sequestration or permanence. Moreover, natural analogs where upwelling

TABLE 4.3 Research and Development Needs: Artificial Upwelling and Downwelling

NOTE: Bold type identifies priorities for taking the next step to advance understanding of artificial upwelling and downwelling.

occurs are generally net sources of CO₂ to the atmosphere (Takahashi et al., 1997) versus net sinks. Even if these predictions are correct and AU proves too costly or impractical for large-scale ocean CDR, AU may prove to be a valuable tool to promote aquaculture or fisheries (assuming the extraction costs do not exceed the C sequestration potential) or simply as a research tool to better understand the biological responses of microbial communities to nutrient perturbations. Pilot studies are principally needed to address the CDR potential of AU and the ecological consequences of these activities. Only targeted, regulated, and transparent field studies can help to minimize current uncertainties and determine if this strategy could be an effective component of an ocean CDR portfolio.