Sustainable scale-up of negative emissions technologies and practices: where to focus

Most climate change mitigation scenarios restricting global warming to 1.5 °C rely heavily on negative emissions technologies and practices (NETPs). Here we updated previous literature reviews and conducted an analysis to identify the most appealing NETPs. We evaluated 36 NETPs configurations considering their technical maturity, economic feasibility, greenhouse gas removal potential, resource use, and environmental impacts. We found multiple trade-offs among these indicators, which suggests that a regionalised portfolio of NETPs exploiting their complementary strengths is the way forward. Although no single NETP is superior to the others in terms of all the indicators simultaneously, we identified 16 Pareto-efficient NETPs. Among them, six are deemed particularly promising: forestation, soil carbon sequestration (SCS), enhanced weathering with olivine and three modalities of direct air carbon capture and storage (DACCS). While the co-benefits, lower costs and higher maturity levels of forestation and SCS can propel their rapid deployment, these NETPs require continuous monitoring to reduce unintended side-effects—most notably the release of the stored carbon. Enhanced weathering also shows an overall good performance and substantial co-benefits, but its risks—especially those concerning human health—should be further investigated prior to deployment. DACCS presents significantly fewer side-effects, mainly its substantial energy demand; early investments in this NETP could reduce costs and accelerate its scale-up. Our insights can help guide future research and plan for the sustainable scale-up of NETPs, which we must set into motion within this decade.


Introduction
The current global levels of anthropogenic CO 2 emissions-approximately 40 Gt a −1 (UNEP 2021)must urgently decline to net zero to avoid transgressing the carbon budget that would limit global warming to 1.5 • C above pre-industrial levels, i.e. 500 Gt CO 2 at the beginning of 2020 (50% likelihood, IPCC 2021). In addition to implementing stringent emissions reductions measures, most climate change mitigation pathways restricting the temperature increase to 1.5 • C rely on carbon dioxide removal (CDR) to either neutralise residual emissions that are hard to prevent or offset a temporary temperature overshoot. The magnitude of the cumulative CDR needed throughout the 21st century is vast (median estimate in scenarios with no or limited overshoot: 584 Gt CO 2 , IPCC 2022) but it is still unclear whether we will be able to amass the technical expertise or mobilise the resources needed to tackle this challenge in a timely manner (Lawrence et al 2018, Nemet et al 2018.
Multiple CDR companies have recently emerged (CarbonPlan 2022), and initiatives like the XPRIZE Carbon Removal competition (XPRIZE Foundation 2022) or the ClimAccelerator program (Carbon Removal ClimAccelerator 2022) seek to promote the creation of novel start-ups and diversify the portfolio of available negative emissions technologies and practices (NETPs). Nonetheless, the effectiveness and sustainability implications of NETPs remain uncertain. Several studies point out that global net negative CO 2 emissions might be less effective at cooling than positive emissions are at warming (Vichi et al 2013, Jones et al 2016, Zickfeld et al 2016, Keller et al 2018, 2021, implying that the required amount of negative emissions might be higher than previously estimated. Moreover, the deployment of NETPs could raise other environmental concerns and lead to side-effects (Heck et al 2016Fuss et al 2018, Cobo et al 2022a, Smith et al 2019b, Qiu et al 2022. Here we build on earlier works (Fuss et al 2018, Minx et al 2018, Nemet et al 2018 to synthesise and critically discuss the growing body of knowledge on NETPs, including emerging NETPs that have previously been overlooked, e.g. marine NETPs. Based on five key performance indicators (KPIs), we assess the NETPs' performance level and conduct a Pareto analysis to identify those that could play a more important role in the future, and the areas where research and investments should focus. We screened the relevant academic papers in the Scopus database and used online resources-the Carbon-Plan database (CarbonPlan 2022) and events organised by the OpenAir (OpenAir 2022a), AirMiners (Air Miners 2022) and Ocean visions (Ocean visions 2021) communities-to track the most recent developments in the greenhouse gas removal (GGR) space.
We found that none of the assessed NETPs simultaneously outperformed all the others in terms of the five KPIs, supporting the thesis that a portfolio of NETPs will likely be needed. Our analysis indicates that terrestrial and chemical NETPs are the most promising. Forestation and soil carbon sequestration (SCS) practices show an overall good performance level and are currently ready for deployment, offering potential co-benefits that could incentivise their implementation. However, they also pose risks, chiefly the possible release of the stored carbon. By contrast, most NETPs based on chemical processes can reduce the unintended impacts of negative emissions. Among them, direct air carbon capture and storage (DACCS) presents particularly appealing KPI values; thus, investing in DACCS appears to be a good strategy to accelerate the sustainable scale-up of CDR, if developed in parallel to the renewable energy system. Terrestrial enhanced weathering deploying olivine also exhibits a promising performance, but still needs further research to guarantee that the associated impacts are tolerable. Concurrently advancing the development of NETPs involving the degradation of greenhouse gases (GHGs), still quite immature, could also contribute to sustainably meeting the demand for negative emissions.
Our results suggest that bioenergy with carbon capture and storage (BECCS) and marine NETPsthe latter showing incipient development levelsshould not be prioritised over more promising GGR strategies. Nevertheless, most climate change mitigation scenarios rely heavily on BECCS (IPCC 2022), which denotes the need to upgrade integrated assessment models to incorporate additional NETPs and better characterise their performance.
The paper is structured as follows: first, we revise the definition of NETPs and the state of the art. Then we assess their performance level and identify the best-i.e. Pareto-efficient-NETPs, to conclude with recommendations for their sustainable deployment.

NETPs definition
Although the terms CDR and negative emissions are often used interchangeably, the latter can refer to the removal of any GHG from the atmosphere. We define NETPs as the technologies and practices that attain negative emissions according to these three criteria: (a) The global warming impacts associated with the NETPs' life-cycle GHG emissions do not exceed the impacts prevented by withdrawing GHGs from the atmosphere. (b) The GHGs are either sequestered 'in a manner intended to be permanent' (Tanzer and Ramírez 2019)-i.e. in a sink that is not subject to foreseeable perturbations-, or permanently transformed into other compounds with lower global warming potentials. (c) NETPs achieve negative emissions relative to a baseline where they are not deployed, i.e. they must be additional to ongoing GGR. Hence, actions to preserve existing carbon sinks (e.g. forest conservation) are not classified as NETPs.
We can only verify that net negative emissions are achieved after quantifying the impacts of the GHGs emitted throughout the NETPs' entire life cycle (Terlouw et al 2021), a detailed analysis that is outside the scope of this study. Therefore, hereon we use the term NETPs more loosely, referring to those technologies and practices that could potentially generate negative emissions. On the other hand, some NETPs may have additional effects on the climate system, which could lead to a net warming effect despite achieving net negative emissions-e.g. forestation practices in high latitudes (Bala et al 2007, Bonan 2008. Figure 1 provides an overview of the reviewed NETPs. We differentiate between terrestrial, marine, BECCS and chemical NETPs. Terrestrial and marine NETPs enhance the CO 2 sequestration capacity of natural sinks, whereas BECCS NETPs capture and store the CO 2 generated in bioenergy production processes, which was previously taken up by biomass. NETPs relying on chemical processes either degrade GHGs into substances with lower global warming potentials or exploit the ability of CO 2 to react with specific compounds to separate it from the other air components. The CO 2 streams produced by BECCS and some chemical NETPs must be subsequently stored in the ocean or lithosphere, or transformed into carbonate products.

State-of-the-art NETPs
We do not consider the utilisation of atmospheric CO 2 for the production of chemicals as a NETP, given the uncertainties surrounding the carbon fate during the end-of-life treatment and its modest CDR potential (Hepburn et al 2019).
In this section, we review the NETPs within each of these four major categories. The CO 2 sequestration methods, integral to certain NETPs, are described in appendix A of the supplementary information.

Terrestrial NETPs
Terrestrial NETPs are the technologies and practices that increase organic carbon in the soil and landbased biological stocks. We distinguish between the terrestrial NETPs that predominantly sequester carbon in plants and forest products, soil organic matter and biochar.

Plants and forest products
The net carbon sequestered by plants is determined by the balance between the atmospheric carbon assimilated in the photosynthesis process, and the carbon losses related to plant and microbial respiration, organic matter decomposition, leaching, erosion, controlled burn or wildfire (Lal 2004, National Academies of Sciences Engineering and Medicine 2019). Although most of the captured carbon is stored in the plants and harvested plant products, the fraction of dead-root biomass and above-ground plant residues that is not subject to mineralisation or erosion losses becomes a source of carbon to the soil (Aalde et al 2006a(Aalde et al , 2006b). The net change in local soil organic carbon (SOC) depends on the environmental conditions (e.g. temperature and precipitation), edaphic factors (e.g. soil structure and C:N ratio), species characteristics and type of landuse change (LUC)-e.g. the conversion of grassland to short-rotation coppice such as willow or poplar plantations can result in net carbon emissions, i.e. the loss of SOC (Qin et al 2016).
Moreover, trees can produce trace GHGs, namely methane and nitrous oxide (Welch et al 2019), and forest ecosystems emit volatile organic compounds (VOCs) that control the formation of climate forcers responsible for warming (methane and tropospheric ozone) and cooling (aerosols). However, there is no consensus over the net climate impact of these VOCs (Unger 2014.
The effect of these NETPs on the climate is also governed by the surface energy fluxes and the hydrologic cycle (Bonan 2008). Increasing the forest cover reduces the share of the incident solar radiation that is reflected. The change in the surface albedo varies greatly across regions, but it is particularly relevant in the boreal ecological zones frequently covered by snow, where it can counteract the benefits of CO 2 sequestration, leading to further warming. On the contrary, tropical forests show a net cooling effect because the increased warming associated with the reduced albedo is offset by the high evapotranspiration rates. The net climate forcing of temperate forests is extremely uncertain and location-dependent, but in general their climate benefits are considered marginal (Bala et al 2007, Bonan 2008. Despite the enhanced precipitation that occurs in forested areas, their water demand can be substantial (Nolan et al 2021). Large-scale biomass plantations might also compete with the food system for land and nutrients, and they are chiefly constrained by land availability (Heck et al 2016, Boysen et al 2017, Ledo et al 2019. Moreover, the cultivation of non-native single species could have detrimental consequences for the local biodiversity if other ecosystems prevail in the area (Hulvey et al 2013, Liang et al 2016). Theoretically, trees could be genetically engineered to increase the photosynthesis rates (Jansson et al 2010) and improve nutrient and water use efficiencies (Jez et al 2016). It has also been suggested that the cultivation of plant varieties with light pigmentationby selective breeding or genetic modification-could maximise the surface albedo (Ridgwell et al 2009). However, the potential risks of introducing genetically modified species into natural ecosystems are still largely unexplored.
Forest conservation is a crucial strategy to simultaneously mitigate climate change and preserve biodiversity (Popp et al 2012). However, we do not classify the activities preventing forest degradation and deforestation, such as those supported by the REDD+ program, (FAO 2022) as NETPs because they do not involve the creation of new carbon sinks, as opposed to planting trees.
The global tree cover can be increased by integrating trees into agricultural systems and land with grazing livestock-i.e. through agroforestry and silvopasture practices-, or establishing new forests. Planted forests currently constitute 7% of global forests (FAO and UNEP 2020). Some planted forests comprise native species and their structure is not defined by the silvicultural practices; they are usually planted to restore and protect the ecosystems. By contrast, plantation forests are a type of planted forests typically aiming at timber production and composed of intensively managed indigenous or introduced tree species, established by planting or seeding one or two tree species of even age with equal spacing (FAO 2018).

Forestation
The large-scale implementation of afforestation and reforestation practices-currently spurred by global initiatives like the trillion tree campaign (Trillion tree campaign 2021)-could reverse the global trend of forest loss, with average deforestation rates of 10 Mha a −1 between 2015 and 2020 (FAO and UNEP 2020). Afforestation is the conversion of land that was previously not covered by forest to forestland, whereas reforestation takes place on land that has been recently deforested; i.e. afforestation implies a LUC, whereas reforestation occurs in an area classified as forest (FAO 2018).
Forests can maintain CDR for decades before it declines as the trees mature (Houghton et al 2015). Recent analyses cap the global CDR potential of forestation at 2.6 Gt a −1 over 2025-2055 (Austin et al 2020) and 5.6 Gt a −1 (average over 21st century, Favero et al 2020), with the largest potential being realised in the tropics (Doelman et al 2020). On the other hand, improved management practices in existing forests (e.g. reduced logging or extended rotations) could lead to the additional removal of up to 2 Gt a −1 CO 2 -eq (Smith et al 2020).
The advantage of forestation over other NETPs is that it is easy to deploy and cost-competitive (The Royal society 2009), with costs ranging between 5 and 53 US$ 2020 t −1 CO 2 removed (Fuss et al 2018). Nevertheless, the carbon sequestered in forests could return to the atmosphere due to human-induced LUC or unexpected perturbations such as pests, droughts or fires (Fuss et al 2018).

Building with wood and fibres
Agricultural residues and natural fibres like cork or hemp can be used as insulation materials, thereby sequestering the carbon contained in them (Kriegh et al 2021, Shen et al 2022a. Furthermore, the mechanical properties of certain engineered wood products-such as glued laminated timber (glulam) and cross-laminated timber-allow them to replace steel and concrete as structural materials (D'Amico et al 2021). Bamboo-a fast growing grass with properties comparable to those of steel and concreteis also an appealing material because of its ability to sequester carbon rapidly and thrive in degraded lands (Project Drawdown 2021a). The main climate benefits of these products stem from the substitution of construction materials (National Academies of Sciences Engineering and Medicine 2019, Shen et al 2022a); the GHGs emitted throughout the life cycle of wood and bamboo products are lower than those associated with the functionally equivalent amounts of steel and concrete (Upton et al 2008, Sathre and Gustavsson 2009, Sathre and O'Connor 2010, Zea Escamilla et al 2016. Nevertheless, the duration of the carbon sequestration in these materials is usually temporal, dependent on their lifetime and end-of-life treatment. It has been estimated that deploying timber as a construction material could sequester up to 0.04-2.5 Gt a −1 CO 2 by 2050 (Churkina et al 2020) at a low cost (McLaren 2012). Updating building codes to guarantee higher energy efficiencies in buildingswhich would favour wood over other materials with worse insulating properties-could help to reach the full potential of wood as a construction material (Wimmers 2017).

Wood burial or storage
Wood can be harvested and buried under anaerobic conditions or stored above-ground (Zeng 2008, Zeng et al 2013. It has been estimated that this CDR strategy could sequester 3.7-11 Gt a −1 CO 2 (Zeng et al 2013). The results of the field tests conducted to date (Carbon lockdown 2022) show that this NETP can significantly slow biomass decomposition, although it does not completely halt it (Adair et al 2010). Despite the low cost of this NETP (9-33 $ t −1 CO 2 ) and its easy implementation (Zeng 2008), some of its potential adverse side-effects-including nutrient lockup in the stored biomass, and disturbance of the local biodiversity-could hinder its large-scale deployment (Zeng 2008).

SCS
The NETPs in this category aim at increasing the amount of carbon sequestered in soil organic matter. The organic carbon content in soils is the difference between the added carbon-i.e. the carbon in roots, plant residues, and soil amendments such as compost or manure-and the carbon lost due to organic matter decomposition, microbial respiration, leaching and erosion (Paustian et al 2019).
Agricultural lands-which have lost 50-70% of their original SOC stocks (Zomer et al 2017)-show the greatest potential for SCS (Minasny et al 2017). Ideally, SCS does not compete with the food system for land (Nolan et al 2021); indeed, enhancing SCS in vulnerable soils with low carbon stocks can improve crop yields and the soil's water-holding capacity, which may reduce irrigation and fertiliser requirements . Additional co-benefits include improved habitats for the soil biota and reduced erosion (Paustian et al 2019, Bossio et al 2020. If implemented in bare soils, SCS could reduce the risk of nutrient leaching (INRAE 2021), although mismanagement could also lead to the loss of nitrogen and phosphorus-whose content in soil increases as the SOC pool is enhanced (van Groenigen et al 2017)-in the water runoff (Fuss et al 2018), causing eutrophication problems.
The SCS costs are highly variable and NETPspecific, typically ranging between 0 and 105 $ t −1 CO 2 removed (Fuss et al 2018). Currently, there are very few projects engaged in SCS practices, notably those operating under the Australian Carbon Farming Initiative, a voluntary carbon offsets scheme (von Unger and Emmer 2018). Nevertheless, recently launched programs such as the 4 per 1000 initiative (2021), which seeks to increase the SOC stocks in the first 30-40 cm of soil by 0.4% per year, or the carbon farming plan within the European Green Deal (European Comission 2022) could incentivise the adoption of SCS.
The technical CDR potential of SCS practices could reach 2-5 Gt a −1 CO 2 by 2050 (Fuss et al 2018). However, these CDR rates can only be sustained for 2-3 decades in mineral soils, until SOC levels reach a new equilibrium (West andSix 2007, Paustian et al 2019). The main risk associated with this carbon sink is its reversibility; the SOC management practices must be indefinitely maintained to avoid losing the sequestered carbon (Bossio et al 2020). Nonetheless, even if these practices are permanently implemented, the rise in temperatures due to global warming is likely to drive the net loss of SOC to the atmosphere (Crowther et al 2016, Melillo et al 2017).
Restoring the hydrology of organic soils and transitioning from conventional tillage to no-till agriculture may increase SOC relative to a scenario with no change in land management-although the efficiency of the latter is heavily debated (Ogle et al 2019). However, we do not classify them as NETPs because they do not involve the sequestration of atmospheric carbon, but rather the prevention of SOC losses (Paustian et al 2019). We analyse the main SCS methods next.

Plants with extensive roots
Roots are usually the main contributor to plant SCS; they are five times more likely than above-ground litter to be stabilised as soil organic matter (Jackson et al 2017). Plants with deeper and larger roots-e.g. perennial vegetation such as trees and grasses-could remove on the order of 1 Gt a −1 CO 2 -eq (Paustian et al 2016). However, these crops should be planted on low-carbon soils, otherwise the carbon loss due to the cultivation and land conversion processes could offset the carbon gains associated with the root system and the above-ground plant residues (Whitaker et al 2018).
While the roots and above-ground plant residues are left in the field to increase the SOC pool, biomass can be harvested and used for bioenergy generation or as a construction material (Shen et al 2022b), which provides the cultivation of these crops with an economic edge over other NETPs. Another advantage of extensive root systems is that they improve the plants' ability to retain nutrients and water, providing resistance to droughts and fertiliser runoff (Kell 2011).
Current research efforts focus on developing crops with larger and deeper roots through selective breeding or genetic engineering techniques (Kell 2011, Paustian et al 2019. Most notably, the Land Institute is working on the perennialisation of cereal grains and other annual crops, some of which are already commercialised (The Land Institute 2021).

Organic matter amendment
The carbon present in soil amendments (such as compost produced from municipal organic waste) contributes to increasing the SOC levels. Furthermore, organic amendments can improve the soil properties and nutrient availability, which stimulates plant productivity and additional carbon uptake (Paustian et al 2019). The CDR capacity of this NETP depends on the original fate of the materials-e.g. manure application does not sequester additional carbon compared to a baseline scenario where it is left in the field (Powlson et al 2011, Leifeld et al 2013-, and on the decomposition rate of the added carbon. The latter is typically slower than that of fresh plant residues and dependent on the soil properties and amendment characteristics (Diacono andMontemurro 2011, Paustian et al 2016).
Spreading light-coloured residues such as cereal straw on the soil could provide cooling benefits by increasing the surface albedo ). By contrast, dark soil amendments like compost could have the opposite effect (Meyer et al 2012). Furthermore, the application of organic matter to the soil may increase nitrous oxide emissions (Smith et al 2001), which could offset the climate benefits of carbon sequestration. Moreover, the leaching of the nutrients contained in the organic amendments could pollute water bodies, whereas the heavy metals and organic pollutants in the composted organic wastes could pose a risk for human health (Cobo et al 2018).

Managed grazing
The practices under this category seek to maximise the SOC inputs from plant roots and residues in grazing lands by maximising forage production or controlling grazing intensity (Paustian et al 2019).
Introducing more productive species with deeper roots, such as legumes, and adjusting the water and fertiliser inputs to the plants' demand can increase forage production (Smith et al 2008). On the other hand, techniques to manipulate the grazing intensity include decreasing the number of animals in a given area (to prevent overgrazing), and implementing rotational or multi-paddock grazing systems, which allow the land that has already been grazed to recover (Project Drawdown 2021b).
The main co-benefits of these practices are the improvements in biological diversity and soil properties (Bossio et al 2020). However, the carbon sequestration rates are highly location-dependent, i.e. subject to the climate, soil and vegetation characteristics (Conant et al 2017). It has been estimated that the global sequestration potential of grazing lands ranges between 0.3 and 1.4 Gt a −1 CO 2 -eq (Henderson et al 2015).

Improved cropping systems
Maximising the time during which soil is covered by vegetation can increase the SOC stocks and simultaneously enhance soil quality and fertility. Fallow frequency can be reduced by planting seasonal cover crops-whose global CDR potential is estimated to be 0.4 Gt a −1 CO 2 -eq (Bossio et al 2020)-and diversifying crop rotations, specifically by introducing perennials, legumes and species that produce large amounts of residues (National Academies of Sciences Engineering and Medicine 2019, Paustian et al 2019). The costs of these practices vary greatly across locations and methods; the available estimates in the literature range from 23 to 147 $ t −1 CO 2 -eq (Tang et al 2016).

Biochar
Biogenic carbon can be sequestered in the biochar produced as biomass is subjected to certain thermochemical processes, namely pyrolysis, gasification, hydrothermal carbonisation and flash carbonisation (Fawzy et al 2021). Slow pyrolysis is the main process used for biochar production, given the good product properties and high yields that it can achieve. In this process, biomass is heated in the absence of oxygen or in a low-oxygen environment, which allows around 50% of the carbon content in biomass to be transferred to the biochar (Schmidt et al 2019). The energy released in the combustion of the other pyrolysis products (bio-oil and pyrogas) is enough to satisfy the heat demand of the pyrolysis process and provide heat for other applications (Peters et al 2015a). The CO 2 generated in the combustion process could be separated from the flue gases and subsequently stored at the expense of substantially raising the energy penalty.
Biochar can be used as an aggregate in construction materials (Gupta and Kua 2017), but its main application is as a soil amendment. Although the biochar applied to soils may release the nutrients originally contained in the biomass feedstock, these are not sufficient to replace conventional fertilisers. This limitation could be overcome by co-composting the biochar-i.e. using it as a compost additive-, which has been demonstrated to improve the compost performance (Schmidt et al 2021).
The fertility and productivity of acidic soils typically improve after the application of biochar, whereas the crop yields of alkaline soils are more likely to decrease (Tisserant and Cherubini 2019). Given its high reactivity and specific surface area, biochar can adsorb nutrients and pollutants, contributing to soil remediation and water purification. Furthermore, it can increase the soil's water-holding capacity (Smith et al 2019a). However, its ability to immobilise chemicals may have detrimental consequences, such as a reduction in the efficiency of herbicides and pesticides. Biochar can also be a source of contaminants, such as heavy metals, organic pollutants, particulate matter, carbon black, etc (Tisserant and Cherubini 2019). Conversely, it reduces nitrous oxide and nitrogen oxide emissions from fertilised soils (Tisserant and Cherubini 2019). Ammonia and methane emissions rates are also affected by the soil application of biochar, but these can either decrease or increase, depending on the soil and biochar properties (Tisserant andCherubini 2019, Schmidt et al 2021).
Biochar application to soils is likely to reduce the surface albedo, which decreases the climate mitigation benefits of this NETP (Genesio et al 2012, Meyer et al 2012. Moreover, the biochar carbon content can degrade to CO 2 in a timeframe ranging from a few years to millennia, contingent on the soil and climate conditions, and the characteristics of the feedstock and pyrolysis process (Gurwick et al 2013); e.g. at a soil temperature of 15 • C, between 18 and 37% of the carbon will be mineralised after 100 years (Woolf et al 2021). Hence the importance of meeting biochar quality standards, such as those set by the European biochar certificate (European Biochar Certificate 2022).
It has been estimated that 0.44-2.62 Gt a −1 CO 2 could be sequestered as solid carbon in biochar without occupying additional land; the area needed to grow the biomass would equal the land reduction associated with the increase in crop yields due to the biochar application (Werner et al 2022). The costs of this NETP range between 32 and 127 $ t −1 CO 2 (Fuss et al 2018), and one of its main advantages is that it can be implemented at both small and large scales (Werner et al 2018). However, despite the high maturity level of this technology, pyrolysis plants at industrial scale are currently scarce, (Aines et al 2021) e.g. the one in Stockholm, which produces district heating and biochar from biomass residues (Bloomberg Philantropies 2021).

Marine NETPs
We define marine NETPs as the set of technologies and practices that seek to maximise the long-term storage of carbon in the ocean. We classify marine NETPs as physical, chemical or biological, based on their underlying mechanism. Figure 2 illustrates the CDR mechanisms of marine NETPs.
Most marine NETPs rely on the manipulation of ocean processes. The main drivers of the oceanic carbon cycle are the solubility pump-the transport of dissolved carbon from the surface to the deep ocean by downwelling currents-and the biological pump, which is based on the fixation of the carbon dissolved in the surface waters by photosynthetic organisms (Scott-Buechler and Greene 2019).
Chemical weathering also contributes to CO 2 sequestration in the ocean. The chemical weathering processes that occur over geological timescales break down rocks, releasing cations that are transported to the ocean and react with the dissolved CO 2 , producing carbonate minerals that are stored on the ocean floor (Berner 2003).
Enhancing the natural rate of CO 2 sequestration in the deep ocean would draw a shift in the equilibrium between the CO 2 concentrations in the atmosphere and the ocean surface water, leading to the transfer of atmospheric CO 2 to the ocean. However, this process is not instantaneous; it takes about one year for the CO 2 concentrations in the atmosphere and the seawater to reach an equilibrium, depending on the regional oceanographic properties (National Academies of Sciences Engineering and Medicine 2021).
The NETPs involving the placement of materials in the ocean are subject to the London Protocol (IMO 2006a), which aims to prevent marine pollution. This protocol was amended to allow the storage of CO 2 in sub-seabed geological formations (IMO 2006b) and prohibit ocean fertilisation (IMO 2013), although research activities focusing on the latter may be considered for a permit (IMO 2013). The protocol also establishes a framework to regulate additional marine geoengineering activities in the future (IMO 2013).

Physical marine NETPs
These marine NETPs only require unit operations involving physical changes, like the movement of fluids and heat exchange. Zhou and Flynn (2005) assessed the implications of enhancing the solubility pump by cooling the ocean surface water, which would increase downwelling currents. They concluded that it is unlikely that this technology will ever become a competitive carbon sequestration method; cooling 1 Mm 3 s −1 of seawater from 6 • C to 0 • C would require a heat flux of 25 TW and capture only 35 Mt a −1 CO 2 , while the estimated costs could range between 258 and 5826 $ t −1 CO 2 . Moreover, they emphasised the need for modelling research, suggesting that upwelling currents releasing more CO 2 than the amount captured could offset downwelling currents. To the best of the authors' knowledge, experimental downwelling has never been conducted before. Lovelock and Rapley (2007) proposed to use vertical pipes to pump up nutrient-rich deep waters to fertilise algae in the ocean surface and increase the transfer rate of organic carbon to the deep ocean via the biological pump. A 300 m long pipe was deployed by Maruyama et al (2011) to investigate the effects of artificial upwelling driven by the difference in salinity and temperature at both ends of the pipe. They found that the chlorophyll concentration at the pipe outlet was much higher than in the surrounding surface water, which suggests an increase in the CO 2 absorption rate. Nonetheless, a recent analysis shows that approximately 70% of the carbon exported to the deep ocean after increasing ecosystem productivity is transported back to the surface ocean within 50 years (Siegel et al 2021).

Upwelling
Earth system simulations reveal that this NETP could contribute to cooling the Earth's surface-due to the lower temperature of the ocean's bottom waters-and enhancing terrestrial carbon storage. However, discontinuing the upwelling could lead to rapid warming (Oschlies et al 2010, Keller et al 2014. Furthermore, the decrease in evaporation and evapotranspiration rates would lead to a decline in precipitations. Other side-effects include increased acidification, and a reduction in sea-ice loss (Keller et al 2014).
The global CDR potential of artificial upwelling is likely limited to 50-100 Mt a −1 CO 2 (Koweek 2022); it has been estimated that upwelling 1 Mm 3 s −1 of seawater could only sequester 59 Mt a −1 CO 2 (Lenton and Vaughan 2009). However, the results of the model developed by Yool et al (2009) showcased that pumping up water in some regions could lead to net CO 2 emissions to the atmosphere.

Chemical marine NETPs
These strategies aim at manipulating the pH of the seawater to either extract the dissolved carbon or transform it into other chemical species.

Ocean alkalinisation
This marine NETP relies on the reaction of alkaline substances with the CO 2 dissolved in the seawater. The enhanced weathering (EW) reactions that occur as a result of adding synthetic chemicals or minerals to the ocean are shown below (Harvey 2008, Renforth 2019). Here, Me represents a divalent cation, typically calcium or magnesium.
While the alkaline materials react with the dissolved CO 2 quite rapidly, the subsequent transfer of atmospheric CO 2 to the surface ocean is slower (National Academies of Sciences Engineering and Medicine 2021). Moreover, although the residence time of bicarbonate ions in the ocean is virtually permanent, an increase in the alkalinity levels would lead to mineral carbonation reactions that produce solid carbonate minerals and release part of the previously captured CO 2 (Renforth 2019): The reaction of acidic species (other than carbonic acid) with the produced bicarbonate ions could also lead to the emission of the captured CO 2 , in accordance with reaction (R6) (Zhang et al 2022): On the other hand, the nutrients released during the dissolution of the minerals could stimulate biological productivity, leading to additional carbon sequestration ( Lenton et al's (2018) simulations indicate that adding olivine ((Mg,Fe) 2 SiO 4 ) to the seawater could counteract ocean acidification and lead to the cumulative sequestration of 524-676 Gt CO 2 between 2020 and 2100, whereas Feng et al (2017) provided a conservative sequestration potential of 971 Gt CO 2 over that period.

Direct alkalinisation
The direct addition of synthetic chemicals to the ocean could also speed up the carbon sequestration process (The Royal society 2009, Renforth et al 2013). Renforth et al (2013) performed a techno-economic analysis of an ocean liming system based on the production of calcium and magnesium oxides via the calcination of limestone and dolomite. To ensure a net negative carbon balance, the study considered that the CO 2 generated in the calcination process was captured and stored in a geological reservoir. They concluded that an energy input of 0.7-6.8 GJ would be needed to achieve the net removal of 1 t CO 2 from the atmosphere, and the costs could range between 82 and 181 $ t −1 of atmospheric CO 2 removed.
Alternatively, alkaline wastes could be used (Renforth 2019). Davies (2015) proposed to decompose desalination reject brine using solar thermal energy to produce magnesium oxide, which can be subsequently added to the seawater. This process would require 13.8 GJ t −1 CO 2 removed from the atmosphere. By contrast, treating the desalination reject brine with electrolysis to produce magnesium hydroxide could reduce the energy consumption to 1.8 GJ t −1 CO 2 . However, 13.7 m 3 of water would be consumed per t CO 2 removed with this method (Davies et al 2018).

Coastal enhanced weathering
Coastal EW constitutes an alternative CDR strategy to directly releasing alkaline minerals in the seawater, whereby the minerals are spread on beach environments, and further comminuted and transferred to the ocean due to the action of the waves on the beach (Hangx and Spiers 2009, Montserrat et al 2017). Project Vesta (2022), a non-profit organisation aiming to advance the deployment of coastal EW, estimates that the cost of this NETP could range between 34 and 50 $ t −1 net CO 2 removed if implemented at scales above 100 Mt, with costs falling below 100 $ t −1 at the 1-10 Mt scale (Green 2022). Nonetheless, the CDR potential of coastal EW deploying olivine grains is limited by the toxicological effects associated with the nickel and chromium contained in the olivine; 0.51-37 Gt CO 2 could be sequestered until 2100 without putting benthic organisms at risk (Flipkens et al 2021).

Electrochemical enhanced weathering
Other authors have suggested the deployment of electrochemically mediated EW processes.
House et al (2007) estimated that, under an optimistic scenario, 3 GJ of renewable electricity would be required to sequester 1 net t CO 2 in the ocean by adding the NaOH produced via the electrolysis of an artificial brine. To manage the generated chlorine, they suggested producing hydrochloric acid and neutralising it with silicate rocks. The current CDR costs of Heimdal's pilot plant, based on a similar configuration (Heimdal 2022b), are below 500 $ t −1 (Heimdal 2022a).
Another approach that reduces chlorine generation as a byproduct is based on the integration of water electrolysis and the mineral weathering of carbonate (Rau 2008) and silicate minerals (Rau et al 2013). The protons generated at the anode dissolve the minerals. The metal cations move towards the cathode to form metal hydroxides, and the carbonate and silicate anions migrate to the anode to form carbonic acid, silicic acid, or silica. The hydroxides react with the dissolved CO 2 , enabling its capture as metal bicarbonate or carbonate. The energy consumption of this process (approximately 7 GJ t −1 CO 2 captured) could be reduced by oxidising the hydrogen generated in the cathode. The total costs of this technology, highly dependent on the energy source, could reach 614 $ t −1 CO 2 (Rau et al 2018). The company Planetary Technologies (2022) is currently working on the scale-up of an electrochemical EW process using mine tailings as a source of alkalinity (OpenAir 2022c).
A novel CDR strategy based on the electrochemical splitting of water into proton and hydroxide solutions has recently been proposed (Tyka et al 2022). Pumping the acidic water to the deep ocean would induce the dissolution of calcite depositsaccelerating reaction (R4)-, whereas the basic solution would be released at the surface water, speeding up the ocean's CO 2 uptake. The authors estimate that 3.7-11 t a −1 CO 2 could be sequestered with this method while limiting the pH decrease in the deep water by 0.2, at a cost of 89-285 $ t −1 CO 2 (Tyka et al 2022).

CO 2 extraction from seawater
Eisaman et al (2012) described an experimental set up to extract up to 60% of the carbon dissolved in seawater with bipolar electrodialysis membranes. This process is based on the premise that returning the CO 2 -depleted water to the ocean would draw the absorption of more atmospheric CO 2 into the seawater. The dilute acid and base products derived from the electrodialysis process are used to alter the pH equilibrium in the processed seawater and extract the dissolved carbon. Two process configurations were proposed: in the acid process, the produced acid is used to lower the pH of the treated water and convert the dissolved inorganic carbon to CO 2 gas, which should be sequestered or mineralised to attain carbon negative emissions. In the base configuration, the pH of the processed water is increased with the produced base, which causes the precipitation of the dissolved inorganic carbon as calcium carbonate. Before returning the water to the ocean, the pH is restored with the produced chemicals (Eisaman et al 2018). The associated energy consumption of the acid and base configurations is 11.3 and 15.8 GJ t −1 CO 2 extracted, respectively (Eisaman et al 2018). Capture costs between 393 and 637 $ t −1 of extracted CO 2 were estimated (Eisaman et al 2018).
This is an incipient research area with potential for novel technological developments; an experimental prototype of an electrolytic cation exchange process capable of extracting 92% of the CO 2 present in the seawater and simultaneously produce hydrogen has also been presented (Willauer et al 2014).

Biological marine NETPs
These NETPs sequester carbon in the ocean by biological means, either by boosting the CO 2 uptake rate of marine photosynthetic organisms-in the case of ocean fertilisation and blue carbon-, or through the accumulation of terrestrial biomass in the deep ocean.

Ocean fertilisation
Iron availability limits the photosynthesis rate in around a third of the open ocean (Emerson 2019), whereas nitrogen and phosphorus are the limiting nutrients in the rest of the ocean (Williamson et al 2012). Hence, fertilising the ocean with site-specific deficient nutrients could lift the constraint on the biological pump.
Several small-scale iron fertilisation experiments have been carried out, confirming that iron promotes biomass productivity and CO 2 drawdown from the atmosphere in many regions (Williamson et al 2012, Emerson 2019. Nonetheless, field experiments on phosphorus fertilisation registered a slight decrease in phytoplankton and biomass chlorophyll, suggesting additional limitations to nutrient availability (Williamson et al 2012).
The costs of ocean fertilisation are dependent on the regional conditions; it has been estimated that the CDR costs via macronutrient (nitrogen and phosphorus) and iron fertilisation are 24 and 519 $ t −1 CO 2 , respectively (Harrison 2013, Jones 2014). The cost difference between the two strategies can be mainly attributed to the low sequestration efficiency of the latter. Zahariev et al (2008) calculated that the maximal CDR rate that could be achieved by means of iron fertilisation would be below 3.6 Gt a −1 . In contrast, the maximum CO 2 sequestration potential of combined nitrogen and phosphorus fertilisation has been estimated to be 5.5 Gt a −1 (Harrison 2017). However, the permanence of the carbon sequestered in the deep ocean by enhancing the upper ocean biological productivity is very low: on average, only 32% of the carbon would remain sequestered after 50 years, and it would drop to 25% after 100 years (Siegel et al 2021).
According to Williamson et al (2012), the potential unintended impacts of large-scale ocean fertilisation may include: • The production of gases that may affect the climate, such as nitrous oxide, methane and dimethyl sulphide (DMS). The latter can increase cloud condensation and reflectivity (albedo), leading to a cooling effect (Wingenter et al 2007). • Far-field effects on primary productivity due to the depletion of non-limiting nutrients. • Decrease of oxygen levels.
• Biodiversity impacts. Large-scale ocean fertilisation will likely change the relative abundance of different species. • Increased acidification of deep waters.

Blue carbon
Blue carbon refers to the carbon sequestered in coastal and marine ecosystems.

Coastal blue carbon
The area occupied by vegetated coastal ecosystems (mangrove forests, seagrass beds, and salt marshes) is much smaller than that of terrestrial forests, but their current total contribution to long-term carbon sequestration (307-856 Mt a −1 CO 2 ) is comparable (McLeod et al 2011). It has been estimated that restoring coastal vegetation would lead to a cumulative CDR of 95.3 Gt by 2100 (Gattuso et al 2021). Nonetheless, these ecosystems can act as net methane sources, offsetting the global warming impacts prevented by CDR (Al-Haj and Fulweiler 2020). The costs of this NETP exceed 200 $ t −1 CO 2 and are highly dependent on the type of ecosystem (Gattuso et al 2021).

Macroalgae cultivation
Macroalgae are the most productive marine macrophytes; they currently promote the storage of about 634 Mt a −1 CO 2 in the deep ocean (Krause-Jensen and Duarte 2016). However, their net carbon sequestration capacity is uncertain; a recent study suggests that seaweed ecosystems may be a net carbon source (Gallagher et al 2022).
The cultivation of macroalgae forests (ocean afforestation or sea forestation) has been proposed as a CDR strategy where the natural drifting and sinking of a fraction of the macroalgal material allows carbon sequestration (Ocean visions 2022), with the costs of optimised farming systems at around 75-100 $ t −1 CO 2 (National Academies of Sciences Engineering and Medicine 2021).
Alternatively, macroalgae could be farmed and purposely sunk to the deep ocean. This CDR strategy could sequester up to 12.4 Gt a −1 CO 2 , and more than half of it could remain in the deep ocean for centuries (Wu et al 2022). Although recent economic evaluations indicate that the costs of this NETP could be above 460 $ t −1 CO 2 (Coleman et al 2022, DeAngelo et al 2022), the company Running Tide (2021) is selling CDR through macroalgae farming and sinking at a price of 250 $ t −1 (CarbonPlan 2022), with estimated costs of 150-200 $ t −1 CO 2 (Calacanis 2020).
The CDR potential and costs of NETPs based on macroalgae cultivation, as well as their potential (positive and negative) effects on the local biodiversity are still highly uncertain (Campbell et al 2019, Ricart et al 2022). The latter may include a reduction in phytoplankton primary productivity and oxygen concentrations, and increased subsurface CO 2 and nutrient levels, which can lead to acidification and eutrophication impacts (National Academies of Sciences Engineering and Medicine 2021). Moreover, while GHG emissions-linked to net primary productivity in the open ocean (Weber et al 2019)could counter the climate benefits, macroalgae could also increase the ocean albedo, reinforcing the cooling effect of this NETP (Bach et al 2021). Metzger and Benford (2001) proposed collecting crop residues and depositing them (ballasted with stone) on the ocean floor as a carbon sequestration method. In a later work, the carbon sequestration efficiency of this practice was estimated to be 92.5% (Strand and Benford 2009). Thus, assuming a global generation of crop residues of 4.98 Gt a −1 with a 40% carbon content, 6.75 Gt CO 2 could be sequestered annually, at the cost of 117 $ t −1 CO 2 (Strand and Benford 2009). The authors suggest that lignocellulosic materials are highly stable in the marine environment, yet it is hard to foresee the associated environmental impacts due to the lack of experimental studies. Biochar may be more suitable for this application than raw biomass, given that it is more resistant to microbial degradation and it would not require ballasting because of its higher density (Miller and Orton 2021).

BECCS
BECCS encompasses a wide set of technologies that derive various energy vectors-electricity, heat, methane, hydrogen, ethanol, oil and other biofuelsfrom biomass. The carbon released as CO 2 , which was removed from the atmosphere during photosynthesis, is subsequently captured and sequestered.
Biomass can be sourced from forest and agricultural residues, although their overall CDR potential is limited because of their important role in maintaining SOC stocks and soil fertility . Likewise, the energy valorisation of organic wastes coupled with CCS can provide additional opportunities for CDR (Pour et al 2018). Perennial energy crops including grasses such as switchgrass and Miscanthus, and short-rotation coppice like poplar are particularly promising feedstocks because of their high yields, ability to grown on marginal land and potential contribution to SCS (Canadell and Schulze 2014).
The cultivation of dedicated bioenergy crops in agricultural and forested lands could interfere with the global food supply and biodiversity conservation targets (Popp et al 2012. Thus, BECCS is primarily constrained by land availability and crop productivity (Boysen et al 2017, Jones and Albanito 2020). Starting the deployment of BECCS early would allow exploiting the biomass resources to a greater extent compared to a scenario where BECCS is implemented later, thereby releasing the pressure on land (Galán-Martín et al 2021, Xu et al 2022).
Currently, the largest BECCS facility worldwide is an ethanol fermentation plant located in Illinois, which captures 1 Mt a −1 CO 2 . Only four other BECCS plants-also producing ethanol-were operating in 2019; the total capacity of these five facilities is 1.5 Mt a −1 CO 2 (Global CCS Institute 2019). However, the global technical CDR potential of sustainable BECCS could reach 10.4 net Gt a −1 CO 2 -eq by 2050 (Koornneef et al 2012). The CDR potential of BECCS strongly depends on the technology and the biomass cultivation site (Hanssen et al 2020); greater CDR rates are achieved in subtropical and warm temperate areas, where biomass yields are usually high and initial LUC emissions, low (Hanssen et al 2020). Nevertheless, the soil carbon loss induced by LUC could even offset the sequestered CO 2 under suboptimal conditions Mac Dowell 2017, Harper et al 2018). Furthermore, tree plantations may emit other GHGs, such as methane, nitrous oxide and VOCs (Unger 2014, Welch et al 2019. Regarding the other impacts of BECCS on the climate system, it is commonly assumed that the decrease in the surface albedo linked to biomass plantations may further reduce the climate benefits, although a recent study indicates that bioenergy crops could induce a global cooling effect (Wang et al 2021).
By accounting for the LUC emissions of lignocellulosic crops over a 30 year period (typical plantation lifetime), a global spatially explicit analysis found that BECCS processes with a 90% carbon sequestration efficiency could remove 2.5 Gt a −1 net CO 2 -eq. By contrast, amortising the LUC emissions over 80 years (consistent with mitigation pathways towards 2100) could increase the CDR potential of BECCS to 40 Gt CO 2 -eq/a-at the expense of cultivating a considerable share of the crops on natural forests and grasslands (Hanssen et al 2020). If the substantial nutrient, freshwater and land demands of BECCS, and its impacts on the terrestrial biosphere integrity are factored in, the maximum net removal that would allow BECCS to remain within the Earth's safe operating space delimited by the planetary boundaries would be 0.2 net Gt a −1 CO 2 . CDR rates above 23 net Gt a −1 CO 2 would imply a high risk for the Earth's stability (Heck et al 2018). Indeed, sequestering 1 Gt CO 2 through BECCS could lead to the extinction of over ten vertebrate species if crops were grown on extremely biodiverse areas (Hanssen et al 2021).
Replacing terrestrial biomass with algae could minimise some of the side-effects of land-based BECCS at the expense of increasing costs (Beal et al 2018). Nonetheless, algal BECCS is still an immature technology (Global CCS Institute 2019), and the net impacts of the large-scale cultivation of algae on the climate and marine ecosystems are not well understood yet (Campbell et al 2019, Bach et al 2021). An optimistic gross estimate suggested that 19-53 Gt a −1 CO 2 could be sequestered by subjecting the macroalgae grown on 9% of the ocean's surface to anaerobic digestion (N'Yeurt et al 2012). However, transporting wet algae over long distances from the off-shore cultivation site to the BECCS plant could account for a great share of the energy penalty of algal BECCS, and even lead to net positive CO 2 emissions (Melara et al 2020).
On the other hand, the energy penalty linked to retrofitting existing bioenergy plants with CCS units would increase the energy production costs (IEA GHG 2009). The costs of BECCS are highly contingent on the biomass supply chain configuration, and region-and technology-specific (Fajardy and Mac Dowell 2020). Here we differentiate between two types of BECCS systems, namely those relying on biological and thermochemical processes. Figure 3 depicts the main mass and energy flows involved in different BECCS configurations.

Biological processes
BECCS systems based on biological processes couple fermentation or anaerobic digestion technologies with CCS to produce biofuels.

Ethanol fermentation
Fermentation is the process whereby microorganisms transform sugars into ethanol and CO 2 . As reaction (R7) shows, one third of the carbon in the glucose molecules subjected to fermentation is transformed into CO 2 that can be subsequently sequestered; the remaining carbon composes the ethanol and will be released as CO 2 when the ethanol is used as fuel. Certain substances with a high sugar content, such as the juice extracted from raw sugarcane or sugar beets, can be directly fermented (Laude et al 2011, Milão et al 2019. By contrast, lignocellulosic materials must previously undergo an enzymatic hydrolysis process, where polysaccharides are broken down to glucose and other fermentable sugars (Xiros et al 2013).
A purification unit is required to meet the commercial specifications of ethanol. Although several technologies such as membranes or molecular sieves could help reduce the energy consumption of the ethanol separation process, this step is typically carried out in a distillation unit, which accounts for 50-80% of the total energy input (Logsdon 2004). Therefore, the main CO 2 source in a conventional ethanol plant is the cogeneration unit, designed to supply heat and power to the system (Carminati et al 2019).
The CO 2 produced in the fermentation unit accounts for 11%-13% of the carbon emissions of an ethanol plant (Koornneef et al 2012); hence, sequestering this CO 2 stream alone does not make the overall process carbon negative (Laude et al 2011). However, most fermentation-BECCS plants-including the largest one to date, with a capacity of 1 Mt a −1 CO 2 -only capture the CO 2 released in the fermentation process (Global CCS Institute 2020). This CO 2 stream just needs to be dehydrated and compressed (Moreira et al 2016), whereas capturing the CO 2 produced in the cogeneration unit-which is diluted in the flue gas stream-is more energy-intensive and costly (Laude et al 2011, Bello et al 2020. The overall costs are likely to range between 21 and 185 $ t −1 CO 2 captured, with the upper bound corresponding to the capture of CO 2 from both the fermentation and cogeneration units (Fuss et al 2018). The CDR potential of fermentation-BECCS could reach 1 Gt a −1 by 2050 (Koornneef et al 2012).

Anaerobic digestion
Under anaerobic conditions, bacteria can degrade certain types of biomass to biogas, which is mainly composed of methane (50%-70%) and CO 2 (Thrän et al 2014). The latter must be captured-via CO 2 absorption in a liquid phase, pressure swing adsorption or membranes-prior to injecting the methane into the natural gas grid or using it as fuel in vehicles (Thrän et al 2014). Anaerobic digestion and biogas upgrading are well-established processes; in 2012 there were 277 biogas upgrading plants linked to anaerobic digesters, but none of them sequestered the separated CO 2 (Thrän et al 2014).
By 2050, 2.7 Gt a −1 CO 2 -eq could be removed from the atmosphere through the anaerobic digestion of energy crops, agricultural residues, municipal solid waste, sewage and manure (Koornneef et al 2013), although sequestering the CO 2 produced by burning the methane could provide additional negative emissions. A fraction of the biomass carbon contentcontingent on the feedstock-is lost in the digestate, i.e. the remaining material that cannot be further degraded. This material is rich in nutrients and can be used for soil amendment, but it may require further treatment because of its high water content (WRAP 2012).
The CDR potential of this NETP is modest relative to other BECCS processes because feedstocks with high lignin contents cannot be subjected to anaerobic digestion; this technology is particularly well-suited for biogenic wastes such as the organic fraction of municipal solid waste, sewage sludge or manure (Koornneef et al 2013). Since these biomass resources are decentralised, the size of anaerobic digesters is typically small, which can increase the costs of connecting these facilities to the natural gas and CO 2 pipelines (Koornneef et al 2013). The total costs of anaerobic digestion BECCS, which are highly dependent on the biomass source, could range between 235 and 557 $ t −1 CO 2 (IEA GHG 2013).

Thermochemical processes
In this type of BECCS, biomass is transformed into energy vectors through chemical reactions that occur at high temperatures, releasing CO 2 that is subsequently captured and sequestered.

Combustion
Theoretically, a global CDR potential of 10.4 Gt a −1 could be sustainably achieved by 2050 with combustion-BECCS (Koornneef et al 2012). As biomass is burnt to generate heat and power, all the carbon in the feedstock is transformed into CO 2 , whereas the non-combustible components of the biomass are transferred to the ashes. Several post-combustion capture processes (membrane separation, adsorption, etc) can separate the diluted CO 2 from the flue gas generated in the boiler. The chemical absorption of CO 2 into a monoethanolamine (MEA) solution followed by a desorption step is the most widespread of these processes (Bui et al 2018), and it can achieve CO 2 capture rates above 90% (Brandl et al 2021). Nevertheless, the associated energy penalty is substantial; e.g. the net energy efficiency of a 75 MW biomass power plant could drop from 36% to 23% when a CO 2 capture unit is added (IEA GHG 2009). The use of advanced commercial solvents based on amines blends could help reduce the energy penalty (Bui et al 2018).
Post-combustion capture BECCS is still in the pilot and demonstration phase, although several projects are expected to start sequestering on the order of Mt a −1 CO 2 from biomass combustion plants within this decade (Global CCS Institute 2019, 2020). Capture costs between 224 and 266 $ t −1 CO 2 have been estimated for BECCS based on post-combustion capture (IEA GHG 2009). However, retrofitting existing combustion and waste-to-energy plants with CCS could lower the CDR costs; sequestering the biogenic CO 2 generated in the energy recovery processes of kraft pulp mills could lead to 135 Mt a −1 of negative CO 2 emissions (Kuparinen et al 2019) at a cost of 62-79 $ t −1 (Onarheim et al 2017). Similarly, the incineration of the organic fraction of municipal solid waste produced worldwide coupled with CCS could produce negative emissions at the Gt scale (Pour et al 2018).
Oxy-combustion is an alternative technology to post-combustion capture where the feedstock comes into contact with pure oxygen instead of air, thereby releasing a stream of highly concentrated CO 2 , which only has to be cooled in order to remove the water vapour prior to compression and storage (Cabral In BECCS processes based on chemical looping combustion, the biomass feedstock reacts with the oxygen in a metal oxide, generating a gas stream consisting mostly of CO 2 and water with a low energy penalty. The reduced metal oxide is then transported to another reactor where it is regenerated with air (Adánez et al 2018). This technology has been tested at semi-industrial scale, but the complete conversion of the fuel is hard to achieve due to the formation and loss of char (Adánez et al 2018). Furthermore, a third reactor is needed to fully oxidise the gas, increasing the CDR costs, which could range between 171 and 200 $ t −1 CO 2 (Keller et al 2019). A novel variant of this technology, chemical looping with oxygen uncoupling-in which the fuel reacts with the oxygen released by the metal oxide-, could achieve carbon capture rates over 99%, but it is still in early development stages (Cormos 2017).

Gasification
Biomass can be partially oxidised with a gasifying agent (air, oxygen or steam) to generate high yields of synthesis gas (syngas), mainly composed of carbon monoxide and hydrogen. Part of the biomass feedstock is transformed into biochar, which is usually burnt to provide the energy required to sustain the endothermic gasification reactions. The gasification process also produces a liquid fraction of heavy hydrocarbons (tars), which must be removed to prevent equipment fouling and plugging (McKendry 2002).
Steam is used to increase the CO 2 and hydrogen concentration in the gas phase through the watergas-shift reaction (WGSR). The produced CO 2 is subsequently captured via pressure swing adsorption, absorption or membrane separation processes. Given the high concentration of CO 2 in the gas, this pre-combustion capture method consumes less energy than the alternative post-combustion capture (Shahbaz et al 2021).
After the CO 2 separation, hydrogen can be compressed and used as fuel off-site, or directly oxidised to generate heat and electricity. BECCS based on integrated gasification combined cycles (IGCCs)-which deploy both gas and steam turbines-can achieve higher energy efficiencies than combustion-BECCS at the expense of higher capture costs, between 175 and 364 $ t −1 CO 2 (Bhave et al 2017). Approximately half of the biomass carbon content can be recovered from the syngas, (Rhodes and Keith 2005) but higher CDR efficiencies could be achieved by capturing the CO 2 generated in the biochar combustion and gas turbine. Hence, the technical CDR potential of IGCC-BECCS could amount to 10.4 Gt a −1 by 2050 (Koornneef et al  2012). On the other hand, coupling biomass gasification with fuel cells could potentially double the power produced by IGCCs (Sadhukhan et al 2010), but this technology is still in its infancy (Bhave et al 2017).
A range of biofuels-including methane, methanol and dimethyl ether-can be synthesised after adjusting the carbon monoxide to hydrogen ratio in the syngas via the WGSR and removing the CO 2 (Gassner andMaréchal 2009, Sikarwar et al 2017). Notably, syngas can be converted into syncrude via the Fischer-Tropsch (FT) process, and further refined to produce synthetic gasoline, diesel or jet fuel, which would allow sequestering 57-74% of the biogenic carbon in the feedstock (Kreutz et al 2020), with costs ranging between 368 and 524 $ t −1 CO 2 captured (Larson et al 2020). The CDR potential of BECCS based on gasification and subsequent methanation could amount to 3.5 Gt a −1 CO 2 -eq (Koornneef et al 2013), whereas the FT-BECCS pathway could sequester up to 5.8 Gt a −1 CO 2 -eq by 2050 (Koornneef et al 2012). However, these figures may substantially drop (below 0.1 Gt a −1 CO 2 -eq) if LUC emissions are factored in (Hanssen et al 2020).
Although the feasibility of biomass gasification has already been tested at large scale (e.g. GoBiGas biomass-to-methane demonstration plant, Larsson et al 2018) and even reached commercialisation (Enerkem waste-to-methanol plant, Enerkem 2022) these facilities do not include CCS.

Fast pyrolysis
Heating biomass in oxygen-free or oxygen-restricted environments triggers the pyrolytic reactions that yield bio-oil, biochar and pyrogas. Fast pyrolysis systems, which require high temperatures (around 500 • C) and short residence times (below 2 s), can maximise the production of bio-oil (Perkins et al 2018). Although bio-oil is composed of highly oxygenated hydrocarbons, it can be directly combusted to generate heat and power for stationary applications (Czernik and Bridgwater 2004). However, using it as a drop-in transportation fuel requires upgrading via hydrotreatment and hydrocracking processes, which can be partially carried out with the hydrogen derived from the pyrogas. These processes produce a mixture of lighter hydrocarbons with a low oxygen content (Peters et al 2015b).
Under fast pyrolysis conditions, approximately 48% of the biomass carbon content is distributed between the biochar and pyrogas (Schmidt et al 2019). These products can be burnt to provide the heat required by the system, releasing the carbon as CO 2 that can be subsequently captured and stored (Meerman and Larson 2017). Hence, we estimate the maximum technical CDR potential of fast pyrolysis as 48% of the CDR potential of combustion-BECCS, i.e. 5 Gt a −1 (Koornneef et al 2012). It has been estimated that the costs of this BECCS configuration (including bio-oil upgrading) equipped with MEA post-combustion capture could amount to 404 $ t −1 CO 2 (Meerman and Larson 2017).
Several companies (ENSYN, btg bioliquids) have successfully commercialised the fast pyrolysis technology (Perkins et al 2018), but we are not aware of any bio-oil projects integrating the combustion of biochar and pyrogas with CCS. The company Charm Industrial has developed an alternative CDR pathway whereby bio-oil is not used to deliver energy, but injected into geological reservoirs. They offer this CDR service at a price of 600 $ t −1 CO 2 -eq (Charm Industrial 2021).

Hydrothermal liquefaction
Processing biomass in an aqueous medium at high temperatures and pressures (250 • C-374 • C and 40-220 bar) leads to the disintegration of the polymers in the biomass and the production of bio-crude (Elliott et al 2015). This bio-crude is more viscous, less dense and less oxygenated than pyrolysis bio-oil. It can directly substitute heavy fuel oil, but using the bio-crude as a transportation fuel requires a previous hydrotreatment process (Elliott et al 2015).
The hydrothermal liquefaction of biomass also generates biochar and gas-mainly composed of CO 2 and hydrogen (Lozano et al 2020). Part of the hydrogen required for the hydrotreatment can be sourced from this gas stream, whereas the remaining combustible gases can be burnt along with the biochar to provide heat to the system. Although the process is almost self-sufficient, an external supply of hydrogen and energy is needed (Lozano et al 2020).
Altogether, the gas and solid products contain 26-32% of the carbon in the biomass (Lozano et al 2020, SundarRajan et al 2020. Thus, capturing and sequestering this carbon could contribute to the removal of up to 3.3 Gt a −1 CO 2 -eq, estimated as a fraction of the CDR potential of combustion-BECCS in 2050 (Koornneef et al 2012). This is a conservative assumption because hydrothermal liquefaction does not require a drying pretreatment and therefore can process a wide range of biomass feedstocks with high moisture contents (Tekin et al 2014).
Currently, hydrothermal liquefaction is at pilot and demonstration scale, with companies such as Licella and Genifuel at the forefront (Lozano et al 2020). However, none of these facilities incorporate carbon capture units. The estimated cost of this technology ranges between 583 and 1365 $ t −1 CO 2 captured (Lozano et al 2020).

Chemical NETPs
In this type of NETPs, chemical reactions between GHGs and other compounds enable the GGR. Ocean alkalinisation and CO 2 extraction from seawater, classified as marine NETPs, also belong to this category.

Terrestrial enhanced weathering
Natural weathering processes can be accelerated by grinding silicate and carbonate rocks to small grain sizes and spreading them in warm and humid regions (Strefler et al 2018, Zhang et al 2022. The rock materials dissolve in the presence of water, reacting with CO 2 to produce bicarbonate anions and liberating base cations (e.g. Ca 2+ and Mg 2+ ), as shown in reactions (R2)-(R4). Runoff transports these ions to the oceans, where their residence time exceeds 100 000 years. Under certain soil conditions, some cations precipitate to form carbonate minerals or react with acidic species (reactions (R5) and (R6)), releasing part of the sequestered CO 2 (Beerling et al 2020, Zhang et al 2022). It has been estimated that between 7.1 and 21.3 Gt a −1 CO 2 could be sequestered globally through the enhanced weathering of basalt before reaction (R5) reduces the CDR efficiency of this NETP (Zhang et al 2022).
Although preliminary EW field experiments have been conducted, the long-term effects of this practice have not been quantitatively measured yet (Kantola et al 2017). The weathering rate is site-specific, but it can be accelerated by reducing the grain sizes. The CO 2 sequestration capacity of olivine-rich rocks is better than that of basalts (1.1 vs 0.3 t CO 2 per t rock, Strefler et al 2018), but olivine-rich rocks have higher concentrations of toxic heavy metalschromium and nickel-that could leach into the soil (Beerling et al 2018). Moreover, finely crushed respirable grains pose health risks, which are particularly relevant if olivine contains asbestos (Taylor et al 2016). On the other hand, basalt may contain larger amounts of other metals such as lead and zinc, which could also lead to toxicity impacts for humans (Cobo Costs between 81 and 211 $ t −1 CO 2 captured with basalt, and 63 $ t −1 CO 2 captured with dunite (mainly composed of olivine) have been reported (Strefler et al 2018, Beerling et al 2020. Deploying alkaline wastes as an alternative feedstock could help reduce the costs and improve the CDR potential of this NETP (Renforth 2019), (Bullock et al 2021,  2021). Climeworks operates the world's largest DACCS facility to date, with a CDR capacity of 4 kt a −1 , and they have announced the construction of a plant nine times larger (Climeworks 2021, 2022). On the other hand, Carbon Engineering is building a plant with a CDR capacity of 0.5-1 Mt a −1 (Carbon Engineering 2021), whereas CarbonCapture Inc. is planning for a 5 Mt a −1 project (CarbonCapture 2022). The technologies deployed by these commercial plants can be classified into two main groups (Fasihi et al 2019):

Direct Air Capture
• High temperature liquid sorbent DAC (HTLS-DAC). CO 2 is absorbed into a basic solution, the regeneration of which requires high temperature heat. Carbon Engineering commercialises HTLS-DAC. • Low temperature solid sorbent DAC (LTSS-DAC).
CO 2 is adsorbed onto a solid sorbent that can be regenerated by temperature swing adsorption (TSA) with low temperature heat (80 -100 • C), or by moisture swing adsorption (MSA). Climeworks, Global Thermostat and CarbonCapture Inc. are using this technology, and they rely on TSA.
Carbon Engineering's HTLS-DAC is based on two parallel cycles. The CO 2 entering the air contactor reacts with potassium hydroxide, forming potassium carbonate. In the regeneration cycle, potassium carbonate reacts with a calcium hydroxide solution, producing calcium carbonate, which is calcined at 900 • C to release a pure stream of CO 2 . The byproducts of these reactions are used as reactants in other stages of the process, thereby closing the loop. This process consumes 6.57-8.81 GJ t −1 CO 2 , a substantial energy demand compared to the thermodynamic minimum, approximately 0.5 GJ t −1 (APS 2011). The current capture cost of this technology is roughly 300 $ t −1 (Direct Air Capture Summit 2021), but it has been estimated that the levelised CO 2 capture cost of scaled-up systems could range between 99 and 245 $ t −1 (Keith et al 2018). A variant of the Carbon Engineering's process where the solvent regeneration is carried out through bipolar membrane electrodialysis would allow reducing the energy consumed in the regeneration phase by 30% at the expense of substantially increasing the total costs (Sabatino et al 2020).
In LTSS-DAC systems, as air passes through a filter, CO 2 chemically binds to the sorbent. Once the solid is saturated, CO 2 is desorbed with lowtemperature heat. Climeworks has patented sorbent materials based on amine-functionalised fibrillated cellulose (Maluszynska-Hoffman et al 2017) and potassium carbonate (Vargas et al 2019). Although the cost of LTSS-DACCS is roughly 600 $ t −1 (Fuss et al 2018), Climeworks expects this cost to drop to about 100 $ t −1 by 2030 (Beuttler et al 2019). The current total energy consumption-electricity and heat-of this DACCS configuration currently amounts to 14 GJ t −1 CO 2 , although scaled-up systems are expected to reduce the energy demand by half in the future (Deutz and Bardow 2021). Moreover, the low temperatures required to regenerate the sorbent allow the use of waste heat or heat derived from renewable geothermal energy. The carbon capture efficiencies of the two first commercial Climeworks plants are 85.4% and 93.1% (Deutz and Bardow 2021).
Another LTSS-DAC prototype deploys an anionic resin (amine ligands attached to polystyrene) that adsorbs CO 2 and releases it when it is exposed to moisture, following an MSA cycle (Lackner 2009). This device has been named 'artificial tree' because it does not require an external energy input to drive air through the filters or dry them in the regeneration stage. This feature results in a remarkably low energy consumption, between 1.36 and 2.25 GJ t −1 CO 2 (Goldberg et al 2013, van der Giesen et al 2017). However, its reliance on wind conditions makes this technology's performance highly dependent on the local weather (van der Giesen et al 2017). A capture cost of 110 $ t −1 has been estimated for this NETP (McGlashan et al 2012).
These HTLS-and LTSS-DAC systems deploy synthetic sorbents that improve the CO 2 uptake kinetics with respect to mineral sorbents. Nevertheless, slower DAC processes based on the successive calcination and carbonation cycles of mineral sorbents are gaining increasing attention. McQueen et al (2020) proposed to sequester the CO 2 generated in the calcination of mineral carbonates, and to subsequently spread the remaining metal oxides over land. In the presence of water, these are transformed into metal hydroxides, which subsequently react with atmospheric CO 2 to produce carbonates. After the carbonation process, the generated carbonates are collected and calcined again. Experimental results indicate that only 3-18% of the spread metal oxides would react in one year (Rausis et al 2022). The cost of this NETP could range between 47 and 161 $ t −1 CO 2 captured (McQueen et al 2020).
The company Heirloom (OpenAir 2022b) is developing a prototype passive air contactor (i.e. not reliant on forced air) that can carbonate 85% of the mineral sorbent-hydrated calcium oxide-in 2.5 days, with an energy demand (stemming mainly from the calcination process) of 5.4 GJ t −1 CO 2 . Calcite Carbon Removal is also developing a technology based on calcination/ambient carbonation cycles (8 Rivers 2022). The CO 2 capture costs of mineral sorbent DAC using passive calcium hydroxide structures has been estimated to be 140-340 $ t −1 (Abanades et al 2020).
The general agreement in the literature is that DAC costs will drop as the installed capacity increases, following technology learning curves similar to those of photovoltaic panels and batteries ( (Lackner and Azarabadi 2021). They also suggest that small modular units could be more effective than high-output plants at accelerating the learning rate of this technology (Lackner and Azarabadi 2021).
Furthermore, because of the ability of DAC plants to be ramped up within minutes, they could benefit from the excess generation of cheap renewable electricity (Wohland et al 2018). Nonetheless, Breyer et al (2019) pointed out that a constant supply of electricity is needed to reduce costs; i.e. it is unlikely that future DAC plants will run solely on excess renewable power. In fact, the results of integrated assessment models suggest that the scale-up rate of DAC systems could be constrained by their energy consumption, which could reach up to 25% of global energy demand by the end of the century (Realmonte et al 2019).
Other factors that could act as potential deployment barriers for DAC are the pollution associated with the large-scale production of the sorbents required for LTSS-DACCS (Realmonte et al 2019), and the high water demand of HTLS-DACCS (Fuhrman et al 2020), LTSS-DACCS based on MSA (van der Giesen et al 2017) and DACCS relying on mineral sorbents . By contrast, some of the adsorbents used in LTSS-DACCS relying on TSA can co-adsorb water from the air; Climeworks' DAC configuration can produce up to 2 t water per t CO 2 , depending on weather conditions and humidity (Fasihi et al 2019).

GHG degradation
The maturity level of technologies and practices aiming at the transformation of non-CO 2 GHGs into substances with lower global warming potentials is still low. Hence, further research is needed to establish the technical feasibility and effectiveness of these NETPs.

Methane oxidation
The atmospheric concentration of methane is low compared to CO 2 (1.9 versus 410 ppm in 2019, IPCC 2021), but its global warming potential is 34 and 86 times that of CO 2 for 100-and 20 year time horizons, respectively (Myhre et al 2013). Oxidising 1 Gt methane would prevent an increase in the global surface temperature of 0.21 ± 0.04 • C (Abernethy et al 2021). Furthermore, the averted formation of tropospheric ozone would entail benefits for human health and ecosystems.
The minimum thermodynamic energy required to separate methane from ambient air is almost five times that of CO 2 per unit mass (Boucher and Folberth 2010). However, expressed per t CO 2 equivalent, the minimum energy needed to separate atmospheric methane is 7 and 19 times lower than that of CO 2 , considering 100-and 20 year time horizons (Jackson et al 2021). Moreover, the combustion of the recovered methane could provide 28% of the energy required for the separation (Boucher and Folberth 2010). Some zeolites and porous polymeric networks have been preliminarily identified as promising materials for methane capture because of their selectivity and sorption capacities ( Bioreactors exploiting the natural capability of methanotrophs to metabolise methane at atmospheric concentrations could be an alternative pathway to oxidise atmospheric methane. However, this strategy has only been proven for methane concentrations of about 300 ppm in the animal husbandry sector (Stolaroff et al 2012); it has been estimated that the average atmospheric methane concentration is too low to enable the survival of the methanotrophs within the bioreactor (Yoon et al 2009).

Photocatalytic degradation
The application of photocatalytic coats on the walls of buildings with enclosed animals has been demonstrated to be an effective measure to oxidise methane , but relatively modest reduction rates of nitrous oxide have been achieved under similar conditions (Ming et al 2016). The costs of applying photocatalysis on surfaces could be as low as 166 $ t −1 CO 2 -eq by 2030 (Ming et al 2022). On the other hand, chlorofluorocarbons, and hydrochlorofluorocarbons have been successfully degraded to halogens, acid halides, and CO 2 in experimental photocatalytic reactors (de Richter et al 2016).
Another method proposed to photocatalytically degrade non-CO 2 GHGs is the injection of iron salt aerosols with a chloride content in the troposphere. Solar radiation would promote the formation of chlorine radicals, which could subsequently oxidise methane, VOCs, and carbon black (Dietrich Oeste et al 2017). Furthermore, the deposition of soluble iron over soils and oceans could lead to increased terrestrial and marine primary productivity (Ming et al 2021). Costs between 2 and 54 $ t −1 CO 2 -eq have been estimated for this NETP (Ming et al 2022).
Relative to a baseline without GGR, this NETP can avert damage to human health and ecosystems through two pathways: the decrease in tropospheric ozone formation (due to the oxidation of methane and VOCs), and the prevention of stratospheric ozone depletion, linked to chlorofluorocarbons and hydrochlorofluorocarbons (Huijbregts et al 2017).

Critical assessment of NETPs and sequestration methods
Identifying the most appealing NETPs is critical to design optimal GGR pathways. To this end, we assess the performance level of the reviewed NETPs and sequestration processes based on five KPIs estimated with literature data: (a) The technology readiness level (TRL)-a scale from 1 to 9 that rates the maturity of a given technology or practice. (b) The maximum annual GGR potential. (c) The cost of removing 1 t CO 2 -eq. (d) The number of negative side-effects. We identify the potential adverse consequences of NETPsincluding unexpected impacts on the climate system, substantial resource use and pollutants emissions-in table 1. (e) The number of co-benefits (i.e. positive sideeffects) compiled in table 1, e.g. decrease in ocean acidification, the generation of co-products or improved crop yields.
We discuss the limitations of the selected KPIs in appendix B of the supplementary information. While more detailed analyses including a wider range Table 1. Potential side-effects of the deployment of NETPs and CO2 sequestration methods in accordance with the reviewed literature (positive effects: green cells, negative effects: pink cells, and either positive or negative effects, depending on location and system characteristics: blue cells). Metals.
Water emissions.

Direct ocean injection.
In situ mineral carbonation. Seismicity.
EW of carbonates.  of KPIs would be possible, they would require large amounts of data currently unavailable for many NETPs and sequestration processes. Hence, the KPIs above are intended to represent a good compromise between data availability and performance coverage.

Reactions
To facilitate the interpretation of the KPIs, we define three intervals for each of them-denoting low-, medium-and high-performance levels-according to the cut-off criteria shown in figure 4. NETPs with low potential in terms of their TRLs (TRL ⩽ 3) have not been validated in the laboratory yet. In contrast, high TRLs (TRL ⩾ 7) indicate that a system prototype has been at least demonstrated in an operational environment (European Commission 2014). The selection of the GGR cut-off criteria is based on the median annual CDR rates across 1.5 • C scenarios with no or limited overshoot for years 2030 and 2050; i.e. 1 and 6 Gt a −1 , respectively (IPCC 2022).
Concerning the economic performance, we consider 100 $ t −1 CO 2 as the threshold below which NETPs show a high economic potential, in line with other works (National Academies of Sciences Engineering and Medicine 2019, Lackner and Azarabadi 2021), and deem costs doubling that value uneconomical. We set 20 $ t −1 -the highest cost reported for the most mature sequestration method (geological sequestration), according to the National Academies of Sciences Engineering and Medicine (2019)-as the lower bound below which the sequestration processes show a high economic potential, and consider that costs above twice that value relegate a sequestration method to the low potential category in economic terms.
Regarding the number of negative side-effectsranging from 0 to 7-, values lower than or equal to 2 denote low impact, whereas KPI values greater than or equal to 6 imply high impact. Finally, we define three bands for the co-benefits KPI: 0, 1 and 2 or more, indicating a low, medium and high number of positive side-effects, respectively. The number of negative and positive side-effects does not account for their magnitude or the probability of them occurring, and therefore, the implications of these KPIs are more uncertain. Following a more conservative approach, we applied stricter cut-off criteria to the side-effects KPIs in appendix C of the supplementary information. Table 2 shows the performance level of NETPs and CO 2 sequestration methods according to the five studied KPIs. We include NETPs for which at least three KPIs are available, neglecting some configurations with very incipient development levels, e.g. certain DACCS modalities. Here we compare the overall performance of the main NETPs groups (terrestrial, marine, BECCS and chemical) and storage processes based on the KPIs compiled in table 2, with a special emphasis on the discussion of the side-effects listed in table 1.
We first focus on the NETPs reliant on terrestrial biomass. While the high TRLs and low costs of terrestrial NETPs could facilitate their quick implementation, the BECCS KPIs predominantly Table 2. Performance level of NETPs and CO2 sequestration methods according to the selected KPIs (high performance: green cells, medium performance: yellow cells, low performance: pink cells). Note that the KPIs of NETPs and sequestration methods with low TRLs may not be as robust as those of the more consolidated ones, given their incipient development levels.
These NETPs can also generate other impacts on the climate system. LUC can lead to either GHG emissions or SCS-the extent of which depends on the soil, environment, species and LUC characteristics-, whereas the albedo shifts and GHGs emitted by trees may offset the climatic benefits of the NETPs relying on terrestrial biomass. Algal BECCS could avoid some of these side-effects, but more research is needed to thoroughly analyse its deployment potential and sustainability implications.
Marine NETPs are still quite immature, with those involving physical operations (downwelling and upwelling) performing poorly in terms of most KPIs. The marine NETPs seeking to stimulate the biological productivity of marine organisms (blue carbon and ocean fertilisation) could contribute to climate change mitigation by increasing the ocean albedo, although the emissions of other GHGs could counteract the prevented global warming. Furthermore, other environmental issues may arise, namely increased ocean acidification and eutrophication linked to the release of nutrients, which may impact marine ecosystems. One of the main risks of enhancing the ocean's carbon sink capacity is the potential release of the stored carbon.
Although the environmental impacts of the ocean alkalinisation practices are still quite uncertain, their ability to combat ocean acidification may pose an advantage over the other marine NETPs. Nevertheless, as with terrestrial EW, there is a lag between the time the alkaline materials are spread and the carbon sequestration. Accelerating the CO 2 sequestration rates of ocean alkalinisation and terrestrial EW requires small particle sizes, which entail a high energy consumption. However, the energy demand of coastal EW may be lower since the particle size can be reduced through natural processes. The rock extraction and grinding operations can be a source of particulate matter, which altogether with the potential leaching of heavy metals, can damage human health and biodiversity. On the other hand, the theoretical CDR capacity of terrestrial EW is substantial. Its potential co-benefits include improved crop productivity, the abatement of nitrous oxide emissions, and a reduction in soil and ocean acidification. The costs of this practice vary widely and depend on the type of rock used.
With the exception of terrestrial EW, the NETPs classified as chemical attain the lowest number of negative side-effects. Although DACCS cannot economically compete with other NETPs yet, it is much less constrained by biophysical limits; the main obstacle to its deployment is its high-energy demand-except for LTSS-DACCS based on MSA cycles, which uses water instead of energy for the desorption process. The water consumption could also hinder the scale-up of HTLS-DACCS and mineral sorbent DACCS deploying air contactors. Notably, LTSS-DACCS based on TSA can extract water from air-the only co-benefit across the DACCS configurations-, although its reported costs are the highest across the chemical NETPs group.
HTLS-and LTSS-DACCS (TSA) stand out as the most mature chemical NETPs. Mineral sorbent DACCS has recently appeared in the literature as a feasible CDR option, but it could avert the potential environmental problems associated with the production of synthetic sorbents at large scale. The degradation of GHGs is also an emergent area of research and therefore its impacts are not well characterised yet. However, it could simultaneously minimise the adverse side-effects of GGR and generate co-benefits for human health and biodiversity.
Finally, the CO 2 storage options show a wide range of KPI values. While none of them is limited by the storage capacity, in situ mineralisation and geological sequestration present high TRLs and low costs, indicating that they are ready for deployment.

Pareto-efficient NETPs and sequestration methods
The analysis above highlights that multiple tradeoffs arise between the considered NETPs and storage processes, making their relative assessment challenging. Ideally, a detailed optimisation model should be developed to identify the optimal region-specific portfolio of NETPs. This model should include a range of constraints limiting the NETPs´potential, which would require large amounts of data currently unavailable or highly uncertain. This endeavour is out of the scope of this article; we instead apply the Pareto dominance concept to identify those NETPs that could most likely integrate the portfolio that concurrently optimises all the KPIs. The Pareto concept is widely used in multi-criteria decision-making to identify the best solutions within a set of alternatives. In essence, an alternative A is Pareto efficient if there is not any other alternative B improving A simultaneously in all the relevant criteria. Further analyses can then be performed based on a more detailed inspection of the Pareto solutions once the non-Pareto are dismissed, for example with Pareto filters (Antipova et al 2015), using weights to aggregate the criteria into one indicator (Cortés-Borda et al 2013), or applying the eco-efficiency concept (Schmidheiney 1992). Figure 5 illustrates the dominance relationships between NETPs. Out of the 29 NETPs configurations whose five KPIs are available, we found 16 Paretoefficient NETPs that are not dominated by any others, i.e. their five KPIs cannot be concurrently improved by any other NETP (further details in appendix D of the supplementary information). These Pareto  In figure 5 we can see that SCS and combustion-BECCS (the modalities involving post-combustion capture, chemical looping combustion and oxycombustion) are the most dominant (each of them dominating six other NETPs), which denotes an overall good performance relative to the other NETPs in terms of all the KPIs. By contrast, forestation, wood burial or storage and mineral sorbent DACCS (land application) are also Pareto-efficient NETPs, but they do not dominate any others. The reason is that these NETPs present the worst value for the co-benefits KPI (according to our KPI definition, which does not account for potential co-benefits if these can turn into negative side-effects when the NETP is not implemented properly) and, in the case of the DACCS configuration, the lowest TRL as well.
Regarding the inefficient NETPs, downwelling and ocean fertilisation with iron are dominated the most (by 21 and 10 NETPs, respectively), reflecting a poor performance. The intragroup dominance relationships are particularly relevant within the BECCS cluster, where fast pyrolysis, hydrothermal liquefaction and gasification coupled with either combined cycles (IGCC) or the Fischer-Tropsch process are heavily dominated-by five, nine, three and six NETPs, respectively, primarily other BECCS configurations.
We found that all the analysed terrestrial and chemical NETPs are either Pareto-efficient or dominated by one Pareto-efficient NETP within their category. The latter include biochar amendment (dominated by SCS), mineral sorbent DACCS deploying an air contactor (dominated by HTLS-DACCS), and terrestrial EW with basalt (dominated by the same NETP using olivine). These results point towards the potential future role of terrestrial and chemical NETPs in optimal GGR portfolios.
The question remains as to which of the Paretoefficient NETPs and sequestration processes are most promising. We further screened the Pareto-efficient NETPs by defining two constraints. First, the number of KPIs denoting a low performance (lp) must not exceed one. Second, one of these conditions must be met: either the number of KPIs showing a high performance (hp) is greater than or equal to three, or the number of KPIs with the best values across the KPI ranges (hp * ) is one or more, while hp is greater than or equal to two. The mathematical formulation of these constraints is as follows: We graphically represent these constraints in figure 6, which illustrates how applying these filters reduces the pool of appealing Pareto-efficient NETPs MSA, and (f) terrestrial EW using olivine. Moreover, these criteria render geological storage the most appealing sequestration process.
Out of the selected Pareto alternatives, SCS is particularly well balanced, as it does not attain the worst performance level in terms of any of the KPIs. Furthermore, its TRL, costs and co-benefits denote a high performance. Forestation, on the other hand, presents the best TRL and appealing costs, which will facilitate a quick deployment. The negative-side-effects and co-benefits of these two terrestrial NETPs are largely dependent on the NETPs' location and characteristics, and can therefore be optimised.
While the GGR potential of these terrestrial NETPs is medium, that of the DACCS configurations is theoretically much greater, mainly limited by the scale-up rates and, possibly, resource availability. The major appeal of DACCS is that it attains the lowest number of unintended side-effects.Nonetheless, most DACCS modalities lack co-benefits to incentivise their deployment-except for LTSS-DACCS based on TSA, whose main bottleneck is its substantial cost. The preliminary cost estimates of LTSS-DACCS deploying MSA are much lower, despite its more incipient TRL. By contrast, the maturity level of HTLS-DACCS is within the upper end of the TRL scale (same as LTSS-DACCS using TSA), and shows intermediate costs.
Finally, terrestrial EW with olivine achieves the greatest number of co-benefits across the entire set of NETPs, in addition to appealing GGR potential and cost. Conversely, its potential number of adverse sideeffects is considerable. Further research could help improve its TRL and ascertain the extent of the sideeffects.
In general, research and progressive scale-up can improve poor TRLs and costs. Similarly, NETPs with low GGR potentials could still be integrated into a portfolio of NETPs. Nonetheless, the side-effects KPIs are more critical, since they could deter or incentivise the adoption of a given NETP. Hence, we investigated the sensitivity of our results to the sideeffects cut-off criteria. Figure 1 of the supplementary information reveals that only three chemical NETPs, namely HTLS-DACCS, LTSS-DACCS deploying TSA and terrestrial EW with olivine would be classified as most appealing if we considered stricter cut-off criteria for the side-effects KPIs. Furthermore, under a more conservative approach where the number of negative side-effects is limited, terrestrial EWwith the largest number of potential unintended consequences-would be excluded from this list.
Thus, DACCS appears to be the most resilient NETP to successive screens.
It is worth noting that the KPIs of several NETPs have not been reported yet, and therefore these were not included in the Pareto analysis. However, as we can visualise in figure 7 (which summarises the performance of all the assessed NETPs and sequestration processes), the NETPs involving the degradation of GHGs preliminary show some promising KPIs despite their low TRLs and incomplete profile, especially in terms of their side-effects. Figure 7 only provides a static picture of the status of NETPs and sequestration processes, since their performance will evolve as research and development intensify and scale-up continues. Hence, our results do not predict how the GGR space will progress; instead, they should be used as a guideline to help decision-makers prioritise investments and resources.

Conclusions and future work
Here we reviewed the most relevant NETPs, assessing their performance based on their technical maturity, economic feasibility, GGR potential and side-effects, including resource use and environmental impacts. Our results reflect the current state-of-the-art (or the best estimates based on current knowledge), but the costs and TRLs compiled here may improve with future research and development, while additional unexpected side-effects may also emerge. Despite these limitations, our Pareto analysis and subsequent screening of the optimal NETPs allow us to outline some of the strategies that could facilitate their sustainable scale-up.
Our findings point towards terrestrial and chemical NETPs as the most attractive GGR options. Forestation and SCS constitute easy-to-implement solutions that are ready for deployment. SCS practices are particularly appealing because of their potential co-benefits for crop productivity. However, these terrestrial NETPs could also generate adverse impacts; most notably, certain environmental conditions and human actions could induce the release of the stored carbon. Careful planning and continuous monitoring are needed to warrant that the undesired impacts of these NETPs are avoided or at least reduced.
The deployment of DACCS would minimise unintended side-effects, but removing CO 2 at the Gt scale with DACCS would require expanding the capacity of existing facilities by six orders of magnitude, a tremendous engineering challenge that would exert substantial pressure on the energy system. Early investments in DACCS could help drive costs down and enable its deployment at a pace consistent with 1.5 • C scenarios.
Our analysis also identifies terrestrial EW with olivine as one of the most promising NETPs. Despite its potential role in preventing additional environmental impacts and enhancing soil fertility, this NETP could also entail a considerable number of detrimental consequences, perhaps the most worrisome being the potential damage to human health. Further studies on EW are needed to better understand its risks and assess whether it can be safely deployed.
The NETPs relying on the degradation of GHGs are still quite immature and were not included in our Pareto analysis. Nevertheless, their preliminary KPIs are quite promising; investing in basic research on these NETPs could lead to negative emissions at a relevant scale with minimal trade-offs. By contrast, our results suggest that marine NETPs-also characterised by their overall low TRLs-should not be prioritised in the short term, although further research is still needed, especially to ascertain their environmental impacts.
Even though we identified several BECCS pathways as Pareto-optimal, they predominantly show intermediate performance levels, and none of them are included in the set of most attractive NETPs. Some of the impacts of the NETPs deploying terrestrial biomass are local-or region-specific and could therefore be minimised by carefully selecting the deployment locations. Nonetheless, because of the inability of large-scale biomass plantations to remain within biophysical limits, we do not foresee these NETPs representing the main share of sustainable GGR pathways.
Our findings suggest that climate change mitigation scenarios, which mostly rely on BECCS and forestation for CDR, should consider a wider range of NETPs. We recommend expanding the set of NETPs available within integrated assessment models to include those identified here as Pareto-efficient. This would improve our understanding of the interactions between NETPs and all economic sectors, and could help define detailed guidelines to prioritise future investments. However, drawing from the insights gained in this study, we can preliminarily propose two parallel investment plans focused on (a) companies and institutions commercialising the NETPs classified as most promising, and (b) basic research and development of the Pareto-efficient NETPs. This does not entail that investments in other NETPs should be discouraged, but rather that the Pareto-efficient NETPs should be prioritised.
This study represents a first step towards the design of optimal NETPs portfolios. Devising negative emissions pathways that capitalise on the synergies between NETPs and overcome their individual constraints while minimising overall collateral damage will require rigorous sustainability evaluations based on systematic life cycle assessment studies, multi-criteria decision-support tools and mathematical programming. Further research should build on spatially explicit data provided by Earth system models to develop regionalised NETPs portfolios adapted to location-dependent environmental and resource pressures.
The implications of this study are not only relevant for the climate modelling community investigating the environmental and socioeconomic consequences of GGR. Our results can also serve as a blueprint for scientists developing emergent NETPs or upgrading the more consolidated ones. Moreover, pinpointing the areas where further research and early investments would yield significant advances can help decision-makers draft climate policies and formulate sustainable business strategies to achieve carbon neutrality.
The decisions and investments made during this decade will determine the course of GGR throughout the 21st century; we should regard the 2020s as a training period to develop and upgrade NETPs, and ensure that they can be sustainably deployed at the needed scale. Thus, we echo other works calling for urgent action on negative emissions in addition to emissions reductions schemes.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).