Scientific Basis


Ocean Upwelling for mCDR

We welcome you to review the latest research on empirical data, scientific analysis, and conclusions that support the scientific basis for ocean upwelling (also known as artificial upwelling) for marine carbon dioxide removal (mCDR). Also, read our Frequently Asked Questions.


By Dr. Ian Walsh, Consulting Chief Scientist @ Ocean-based Climate Solutions, Inc

In the ocean we have challenges to measuring the impact of ocean upwelling and deriving time weighted sequestration fluxes.

But these are entirely tractable challenges: we have tools and experience through generations of oceanographers in measuring carbon flux through the ocean where we can and estimating where we don’t have data. We have local process studies and global models that provide bounds to carbon fluxes from diel to annual and greater time scales. Global models typically use well structured physical models (get the physics right first!) overlayed with biogeochemical models with more or less stocks and pathways to apportion energy and variability over significant time scales.

The challenge with ocean upwelling is that along with nutrients, the dissolved inorganic carbon (DIC) concentration also has a regeneration profile, with low concentrations at the surface relatively monotonically increasing with depth, with the gradient controlled mostly by water mass.

So we have to measure both the nutrients and the carbonate chemistry of the upwelled water. What we want with the upwelled water source is a source water for upwelling where there is an ‘excess’ of nutrients relative to the DIC such that there is a net uptake and fixing of atmoCO2 into the particle pool via phytoplankton rather than simply a quick shunting of upwelled CO2 as DIC into the fixed carbon pool. The presence of ‘excess’ nutrients, particularly phosphate, relative to DIC in the twilight zone is a steady state result of nutrients having higher first order remineralization rates than particulate organic carbon. This dynamic operates over long periods of time and space driving the linkage of atmoCO2 to ocean upwelling on a global scale through recent geological time has been long established (Dymond and Lyle, 1985)

There are many ways to define the ‘excess’ relative nutrient load relative to the carbon load. Karl and Letelier (2008) used the phosphate concentration relative to nitrate and DIC to define a sequestration potential for a given depth profile. The attraction for using this method is that temporal and spatial variability of dissolved constituents in the ocean decreases with depth and particularly in the central gyres a relative sparsity of data still yields volumetric/depth/time relationships that are reasonably stable, and hence the initial conditions of the setting of a particular pump deployment in a central gyre can be reasonably assumed to be the conditions over annual time scales.

The task then is to gauge the actual delivered seawater. For that we can use a mixing model over the volume addition:

Temperaturez,plume = A * Temperaturez,ambient + D * Temperatureintake

PO4ex plume = A * PO4z,ambient + D * PO4intake


A + D = 1

Since for the oligotrophic ocean we can assume that the available dissolved nutrient concentration in the mixed layer is non detectable, i.e.:

PO4ML,ambient = 0

Where z here is replaced by the mixed layer depth.

If we assume that the upwelled volume is dispersed into the plume, and that the plume is entirely within the mixed layer, and the mixed layer is equal to or less than the euphotic depth, then we can presume that all upwelled water that enters at the intake is dispersed within the mixed layer and is taken up over time into the biological system.

Since we know that the ambient excess phosphate is zero, then the additional phosphate added to the system will also devolve to zero and that this will occur on the order of days. Hence, the simplest measurement of the net sequestration effect of a given artificial upwelling pump can be based on the temperature and nutrient load relationship at the intake depth range and the measured temperature anomaly at the surface.


A leading research effort in artificial upwelling (AU) is the EU Horizon funded program “Art-Up” under the direction of Professor Dr. Ulf Riebesell, Geomar-Kiel. The Art-Up links to their papers follow.

Baumann, Taucher, Paul, et al. “Effect of Intensity and Mode of Artificial Upwelling on Particle Flux and Carbon Export.” October, 27, 2021.

Summary of study: Reducing CO2 emissions caused by people won’t be enough to stop global warming enough to meet the 1.5°C goal of the Paris agreement. To fight climate change effectively, we need to take steps to get rid of carbon dioxide from the air. One idea for getting rid of carbon dioxide is to create an artificial upwelling. By sending nutrient-rich deep water to the surface of the ocean to boost primary productivity, it might be possible to improve downward flows of particulate organic carbon (POC) and carbon sequestration. In this study, we looked at how different levels of artificial upwelling and two types of upwelling (repeated additions vs. a single addition) affect the export of POC, the stoichiometry of sinking matter, and the depth of remineralization. In the subtropical North Atlantic, we did a 39-day mesocosm experiment in which we added different amounts of deep water to oligotrophic surface water to make it grow. The total amount of nutrients went from 1.6 mol NO3– L–1 to 11.0 mol NO3– L–1. We found that, on the one hand, POC export under artificial upwelling more than doubled and the molar C:N ratios of sinking organic matter increased from values around Redfield (6.6) to 8–13, which is good for potential carbon dioxide removal. On the other hand, things that were sinking were remineralized faster and sank slower, which meant that they were remineralized at shallower depths. In the recurring upwelling mode, the properties of the particles were better for deep carbon export, while in the single mode, the C:N increase of sinking matter was stronger. In both ways that the water went up, about half of the organic carbon that was made stayed in the water column until the end of the experiment. This shows that the plankton communities were still getting used to each other. This could be because producers and consumers react at different rates. Because of this, there is a need for studies with longer experimental periods to measure the responses of communities that have fully adjusted. Lastly, our results showed that artificial upwelling changes many properties of sinking particles, and that the strength of the changes depends on how strong the upwelling is and how it is used.

Ortiz, Arístegui, Taucher, Riebesell. “Artificial Upwelling in Singular and Recurring Mode: Consequences for Net Community Production and Metabolic Balance.” January 12, 2022.

Summary: The artificial upwelling of nutrient-rich waters and the increase in primary productivity that comes with it have the potential to increase the amount of fish caught in the ocean and strengthen the biological pump that pulls CO2 out of the air. Climate change and overfishing are two of the biggest problems of the 21st century. Understanding this technology as a “ocean-based solution” to these problems is becoming more and more important. Yet, not much is known about how well artificial upwelling works or what side effects it might have. We did a large off-shore mesocosm study (about 44 m3) in the oligotrophic waters of the Canary Islands to find out how artificial upwelling affects a natural oligotrophic plankton community at the community level. For 37 days, two different upwelling modes (a single deep-water pulse or a steady supply every 4 days) were used to simulate four different intensities of upwelling: 1.5, 3, 5, and 10 mol L–1 of nitrate, phosphate, and silicate. Here, we show what happened to net community production (NCP), metabolic balance, and the make-up of the phytoplankton community (250 m). When there was more upwelling, the cumulative NCP was higher. After the upwelling started, diatoms took over the phytoplankton community in all treatments, but other taxa, such as Coccolithophores, grew as the experiment went on. The amount of deep water added changed the metabolic balance, leading to I a balanced to net-heterotrophic system in the singular and (ii) a net-autotrophic system in the treatments with repeated upwelling. So, how nutrients are brought into an oligotrophic system is a key factor in how the ecosystem responds, with repeated upwelling leading to a higher long-term net community production (NCP) than a single upwelling. These results show how important it is to measure local responses to upwelling, like the structure and metabolism of communities. This has big implications for the possibility of using artificial upwelling as a way to increase primary production in the ocean.

Kemper, Riebesell and Graf. Numerical Flow Modeling of Artificial Ocean Upwelling. January 3, 2022.

Summary: Artificial Upwelling (AU) of nutrient-rich Deep Ocean Water (DOW) to the sunlit surface layer of the ocean has recently been suggested as a way to increase the ocean’s ability to store CO2 and produce fish. Climate change and food security for a growing human population have been looked at in terms of AU and its possible benefits. But a lot more research needs to be done on the safety, effectiveness, risks, and side effects of AU before we can make better predictions about its potential. Fluid dynamic modeling of the AU process and the transport of inorganic nutrients can give us the information we need to figure out how much damage certain AU devices do to the environment and is a useful tool for improving them. Yet, getting all flow phenomena that are important to the AU process right is still hard, and only a few models are able to do it. In this paper, the results of simulations made with a new numerical solution method are shown. The method is based on OpenFOAM, which is a free modeling environment. It solves the Reynolds-Averaged Navier-Stokes (RANS) equations with additional transport equations for energy, salinity, and inorganic nutrients. The method is meant to be useful for a wide range of ocean flow problems, such as density stratification caused by temperature and salinity and passive scalar transport. The studies in this paper are mostly about how the AU process affects the spread and concentration of nutrients in the mixed surface layer of the ocean. The new method does a good job of capturing the expected flow phenomena. Even though it is known that cold DOW that rises to the surface tends to sink back down due to its high density, the simulations in this paper show that the upwelled DOW settles at the lower boundary of the ocean’s mixed surface layer, keeping a lot of the nutrients available for primary production. Comparative studies of different design options are also shown and analyzed. The goal is to keep as many nutrients as possible in the mixed surface layer.

Ortiz, Joaquin (2022) Oligotrophic Phytoplankton Community Effectively Adjusts to Artificial Upwelling Regardless of Intensity, but Differently Among Upwelling Modes. Front. Mar. Sci., 20 June 2022 Sec.Marine Biogeochemistry

Summary: Artificial upwelling has been suggested as a way to increase CO2 sequestration in the ocean and/or increase fishery yields by increasing primary production in parts of the ocean that don’t produce much. But there isn’t much proof that it works, can be used, or has negative effects. Here, we show some of the results of a 37-day mesocosm study done in oligotrophic waters off the coast of Gran Canaria. The goal was to see how artificial upwelling affected the pelagic community in real-time. Upwelling was simulated in two ways: I with a single deep-water pulse and ii) with a steady supply every 4 days. Each way had four different intensities, which were set by the total amount of nitrate added: about 1.5, 3, 5.7, and 11 mol L-1. In this study, we focus on how the phytoplankton responds by looking at the 14C primary production rates (PP), Chlorophyll a, and biomass. We saw linear increases in PP, PP accumulation, Chlorophyll a, and biomass as the upwelling got stronger. Upwelling helped larger phytoplankton the most, so pico- and nano- to nano- and microphytoplankton changed places. When the upwelling was stronger, adding deep water on a regular basis produced more biomass than adding deep water in a single pulse. It also accumulated a lot more PP per unit of nutrients added and showed a stronger decrease in the percentage of extracellular release as the intensity of the upwelling increased. These results show that oligotrophic phytoplankton communities can adapt well to artificial upwelling, no matter how strong it is, but in different ways depending on the mode of upwelling. When there was a steady supply of upwelled waters, primary production and biomass growth were more efficient than when there was a single pulse of the same volume and nutrient load.


Koweek, David A. “Expected Limits on the Potential for Carbon Dioxide Removal From Artificial Upwelling (June 29, 2022).

The Koweek study argues that field trials for ocean artificial upwelling are justified, pulling from a plethora of studies that have analyzed a variety of flow rates – which Ocean-based Climate Solutions’ ocean upwelling pump exceeds. Read Dr. Ian Walsh’s comments next regarding the Koweek study and its endorsement for further research.

Dr. Ian Walsh’s comments on David Koweek’s study on Expected Limits on the Potential for Carbon Dioxide Removal From Artificial Upwelling.

  1. The paper demonstrates that a flow rate approximately equal to that observed with published AU flow rates is unlikely to support a production and export response sufficient to reach an mCDR removal rate of 1 gigatonne of CO2atmo/year with pumps deployed uniformly over the world ocean. [N.B. I use tonne to refer to the SI measurement while the paper refers to the imperial ton at least in the text. The difference is immaterial to this review, but I’m presuming the text should use the SI unit spelling as SI units are used in the rest of the text.]
  2.  The model takes into account DIC and TA upwelling and potential outgassing
  3. The model uses a single flow rate (0.05 m3 s-1)
  4. The model uses a variety of C:N:P stoichiometries
    1. C:N:P fixed through primary productivity
    2. Micro and macro algae cases
  5. No explicit export function
    1. No regeneration scaling
      1. Time
      2. Depth
  6. Photoperiod modulation is included in the model
  7. The model uses a 1 AUP km-2 density over most of the world ocean

The nice thing about David Koweek’s study is that it couples gas exchange and DIC and TA upwelling kinetics with a reasonable range of nutrient stoichiometry and results in a positive net carbon sequestration potential. Along with Wu et al., 2022 currently under review (, the Koweek model demonstrates that artificial upwelling has the potential to impact CO2atmo.

This is important as I consider the primary unknown in terms of impact on AU sequestration potential is the scaling of nutrient stimulated production uptake of pCO2 relative to the carbonate system kinetics. There is still much to learn about this interaction, but at the least, we see that models that include carbonate system dynamics show positive sequestration potential.

Relevant to Ocean-Based:

  1. The flow rate for Ocean-Based’s AUP is estimated to be 5000 m3/hour
  2. This is 27.8 times the Koweek model flow rate
  3. Using the Ocean-Based flow rate to scale the Koweek model results generates a rate of mCDR of 1.39 gigatonnes/year
    1. This is above the 1 gigatonne/year
  4. From an enterprise perspective, the mCDR potential is not based on a basin or ocean scale uniform deployment but on maximizing mCDR income relative to enterprise costs.
    1. AU pump (AUP) density will be a function of maximizing pump efficiency on an mCDR basis per pump per expected pump lifetime. A uniform deployment scenario, while attractive to modeling, is not likely to be efficient from the standpoint of resource allocation or cash flow relative to enterprise costs (capital, operations, maintenance, and MRV costs) and income.
    2. The operative scaling for an enterprise is mCDR/AUP/ocean area/time: moles CO2atmo AUP-1 km-2 time-1
      1. Time: day (diel cycle), monthly (seasonal), annual (seasonal), decadal (life of the pump, ENSO)
    3. The operative scaling for the economic potential includes the cost of mCDR production (pump capital + operations + maintenance + MRV) over the time scale relative to mCDR payments.
  5. The upshot of taking the perspective of the enterprise is that the scaling of the enterprise will result in the selection of the most advantageous milieus for the placement of the marginal (next) pump. ASIDE: This is entirely different from the testing phase, in which experiential concerns and data gathering are the most important considerations. [running a Test, Fail, Advance loop]
  6. If we run the Koweek model as an optimization tool:
    1. What conditions need to be met to generate a global integrated value that is significant?
      1. Sensitivity analysis concerning volume transport
      2. Force depths with optimal PO4ex input values
      3. K&L 2008 & White et al. 2007 papers
    2. Use the HOT data set as input?
    3. Include the annual cycle relative to the wave energy and upwelling volume relative to the dispersion ratio
      1. There is probably an annual cycle in wave energy that begets a dispersion ratio that goes above and below the diatom bloom point (Ulf et al., mesocosm results)
      2. What does that look like on an annual basis?
      3. Determine the seasonal and annual export cycle using multiple modes based on dispersion ratio and volume pumped:
        1. If we look at three modes:
          1. Diatom bloom
          2. Top-down (1% export of primary productivity)
          3. K&L/White et al., diazotroph bloom
        2. Derive the modal export function
          1. Sum over the annual cycle
    4. Derive the future productivity decrease in the central gyres due to increased SST for future additionality
      1. Do a basin match up, e.g., find the warmer temperature data from BGC Argo and calculate areal differences based on projected future temperatures
      2. Look at Behrenfeld et al., 2006 and subsequent papers (where referenced) for a general relationship
  7. The model does not consider export and fractionation
    1. This is an upper limit model
    2. Use the tri-modal export model to refine export and sequestration
    3. Define other benefits from AU
      1. Temperature/stratification feedback
        1. Especially relative to marine heat waves
        2. Especially relative to trophic level restoration and or impact minimization
          1. Highest value might be as a way of creating pockets of ecological stabilization to retain significant populations through the warmer ocean

Striking coherence between maximum chlorophyll maps (e.g. Feldman maps) and the microalgae potential, with perhaps the exceptions of the eastern edges of the South Pacific and South Atlantic which look offset higher in the upwelling potential.

A more recent paper with similar mapping:

Behrenfeld., M., O’Malley, R., Siegel, D., McClain, C., Sarmiento, J., Feldman, G., Milligan, A., Falkowski, P., Letelier, R., and Boss, E. (2006). Climate-driven trends in contemporary ocean productivity. Nature, 444, 752-755.

Philip Kithil’s graph of pumping rate, and pump diameter (as a valid product) whereas Koweek uses pumping rate and depth)

As stated above, Ocean-based Climate Solutions, Inc’s ocean upwelling pump flow rate is significantly higher than the models analyzed in the Koweek paper. This is because flow rate is directly correlated to pipe diameter and less to pipe length (depth of the inlet) as shown in Koweeks’ Figure 1.

A critical consideration found in the Liu and Jin reference cited by Koweek in this Figure 1, and confirmed by us, is that the water column inside the pipe gains momentum, delivering far more volume than a simple wave height x period calculation.  Liu and Jin determined the simple calculated flow rate was 0.45 m/s, whereas the measured flow rate was 0.95 m/s – greater by a factor of 111%! Our own data found a 60% greater measured vs. calculated flow rate.


Siegel David A., et al. (2021) Assessing The Sequestration Time Scales Of Some Ocean-Based Carbon Dioxide Reduction Strategies. 2021 Environ. Res. Lett. (In Press)

Summary of the study: Carbon dioxide (CO2) removal (CDR) strategies that use the ocean are an important part of the set of steps needed to get negative greenhouse gas emissions. Many ocean-based CDR strategies depend on putting CO2 or organic carbon (which will eventually turn into CO2) into the deep ocean or making the biological pump of the ocean work better. These methods won’t lead to permanent sequestration because ocean currents will eventually bring the CO2 back to the surface, bringing it back into balance with the atmosphere. Here, a model of steady state global ocean circulation and mixing is used to figure out how long CO2 that has been put into the deep ocean stays out of the air. Because there are an infinite number of ways to get from a place at depth to the surface of the water, there will be a range of sequestration times for any single discharge location. The result is that the probability distribution is very skewed, with a long tail of very long transit times. This makes the mean sequestration times much longer than normal time scales. Deeper discharge sites will store CO2 for a much longer time than shallower ones. The average time to store CO2 is between a few decades and a few centuries, but in the deep North Pacific, it takes close to a thousand years. There are big differences in how long it takes for carbon to be stored within and between the major ocean basins. In general, the Pacific and Indian Oceans take longer to store carbon than the Atlantic and Southern Oceans. Assessments done over a 50-year period show that most of the carbon injected at depths greater than 1000 m will stay there, with a few places like the Western North Atlantic being an exception. Ocean CDR strategies that aim to increase the productivity of upper ocean ecosystems so that more carbon can be exported to depth will have a short-term effect on CO2 levels in the air because about 70 percent of the carbon will be brought back to the surface ocean within 50 years. The results shown here will help plan ocean CDR strategies that can help limit the damage to the climate caused by CO2 emissions from fossil fuels.

Dittmar Thorsten, et al. (2021) Enigmatic Persistence Of Dissolved Organic Matter In The Ocean. June 2021 Nature Reviews Earth & Environment.

Summary: Marine dissolved organic matter (DOM) has more carbon than all the Earth’s living things put together. Organisms in the ocean constantly release a wide variety of molecules that become food for microheterotrophs. However, for reasons that aren’t clear, some of these molecules stay in the ocean as DOM for thousands of years. In this Perspective, we look at two ideas that could explain this persistence and compare them. The “intrinsic recalcitrance” paradigm, which has been around for a long time, says that the stability of DOM comes from the way molecules are made. In the “emergent recalcitrance” theory, marine microheterotrophs change DOM all the time, and recalcitrance develops on an ecosystem level. Both ideas make sense based on what we know about the ocean today, but they predict very different responses of the DOM pool to changes in the climate. To learn more about how DOM stays around, we propose a new research strategy called “the ecology of molecules.” This strategy combines the ideas of “intrinsic” and “emergent” recalcitrance with the ecological and environmental context.

Baklouti M., Pagès R., Alekseenko E., Guyennon A., Grégori G. (2021). On the Benefits of Using Cell Quotas in Addition to Intracellular Elemental Ratios in Flexible-Stoichiometry Plankton Functional Type Models. Application to the Mediterranean Sea. Prog. Oceanogr. 197, 102634.

Summary: In the last few decades, there have been a growing number of marine Plankton Functional Type (PFT) models. Since the NPZD models that came before them, the general trend has been toward making models that are more and more complicated, either in terms of the number of variables they include or the level of detail they consider. In this way, flexible-stoichiometry models have been an important step in the history of this type of models. Since then, new developments have been written about in the literature, and this paper focuses on one of them: the addition of abundance to biomass to describe PFTs. With the growing number of studies and data sets from flow cytometry, this new feature makes it possible to test the model. It also makes it possible to get cell quotas (expressed as the amount of a given biogenic element per unit cell) for each PFT of the model. Cell quotas and intracellular ratios are used to control the speed of the different biogeochemical processes in the model. It is suggested here that their role may be crucial for accurately representing some processes, such as mineralization by heterotrophic bacteria. In this paper, we talk about some of the other benefits of cell quotas, such as how they help us understand how the trophic web works and how well organisms are eating. Lastly, the seasonal changes in the number of cells in the PFT that represent large phytoplankton can be used to explain why the NW Mediterranean spring bloom happens from an internal status point of view.


The following is by Ian D. Walsh;; @iandavidwalsh; adapted from Ian’s WHOI OCB Summer Workshop Presentation at Woods Hole, MA, 20-23 June 2022 for website form by Salvador Garcia. Acknowledgments: IDW consults for Ocean-Based Climate Solutions Inc. (Ocean-Based) which has developed and built artificial upwelling pumps and is testing the systems. Ocean-Based did not directly pay for or have editorial control of the research presented here.


The idea of artificial upwelling using vertical tubes in the ocean open has been around for a long time (Stommel et al., 1956). Similarly, the linkage of atmoCO2 to ocean upwelling on a global scale through recent geological time has been long established (Dymond and Lyle, 1985). Current considerations of the potential impact of artificial upwelling have outlined an uncertainty matrix and a research strategy for artificial upwelling within a broader review of ocean-based carbon dioxide removal (CDR) (NASEM, 2021). Meanwhile, the carbon offset ‘market’ is booming (your favorite news source), including ocean CDR.

Here I present some considerations and calculations on the scalability of artificial upwelling and verification pathways to hopefully advance the discussion and spark interest in the OCB community.


The oligotrophic ocean is an attractive deployment location or artificial upwelling at least through the testing phase because the generally low abundance of nutrients in the absence of upwelling provides an excellent background to track the impact of artificial upwelling and a source of excess phosphate at depth to test Karl and Letelier’s hypothesis that net sequestration CO2 is driven by excess phosphate.

Redfield ratio plot of the concentration of phosphate and nitrate at HOT in 2018 from bottle data taken between 480 and 520 meters depth. There is a consistent 0.3 uM/kg ‘excess’ phosphate concentration through the year except for the May 2018 data set.

Data obtained via the Hawaii Ocean Time-series HOT-DOGS application; University of Hawai’i at Mānoa. National Science Foundation Award # 1756517


Verification of sequestration should be independent from the deployment and operation of sequestration technologies, though these should be tightly linked to drive the system towards increase effectiveness.

Continuation of public strategies to measure and model ocean carbon dynamics will contribute to ocean CDR measurements and verification.

Local modeling for additionally and effectiveness of ocean CDR technologies should couple with wider scale public modeling efforts.

Costs for verification relative to ocean CDR will then be spread between the operators of ocean CDR and the wider public and reduce conflict of interest uncertainties.

Autnomous vehicles with physical and biogeochemical sensors such as the Bio-Argo float should be deployed in the vicinity of artificial upwelling to provide near-field data to verify additionality.
Schematic of the Ocean-based wave-powered Ocean Upwelling Pump technology, which uses flapper valves on the intake to drive deep water to the surface.
Deployment of an Ocean Upwelling Pump system surface package with the fabric tube spooled up on the bottom-weight one-way valve on deck.
A simple volumetric dispersion model suggests that a 5000 m3 hr1 pump rate would be detected by a backscattering sensor after seven days with a front width of about 60 meters, sixty kilometers ‘downstream’ of the pump.
Conservative potential sequestered carbon fixed per pump based on HOT nutrient data set. Karl and Letelier’s 2008 estimates are about a factor of 4 higher.

Stommel, H., Arons. A.B., and Blanchard, D., 1956: An oceanographical curiosity: the perpetual salt fountain. Deep Sea Research, Vol. 3, pp. 152-153.

Dymond, J. and Lyle, M. 1985. Flux comparisons between sediments and sediment traps in the Eastern Tropical Pacific: Implications for atmospheric CO2 variations during the Pleistocene. Limnology and Oceanography, 30(4), pp. 699-712.

Karl, David & Letelier, Ricardo. 2008. Nitrogen fixation-enhanced carbon sequestration in low nitrate, low chlorophyll seascapes. Mar. Ecol. Prog. Ser. 364, 257-268.

National Academies of Sciences, Engineering, and Medicine. 2021. A Research Strategy for Ocean-based Carbon Dioxide Removal and Sequestration. Washington, DC: The National Academies Press.

Kithil, P., Garcia, S., and Walsh, I., 2022. Ocean Carbon Dioxide Removal using Wave-powered Artificial Upwelling Pumps. The Journal of Ocean Technology, Vol. 17, No. 1, 2022.