1748-9326/20/11/111004

Export citation and abstractBibTeXRIS

The only way to reach net-zero greenhouse gas (GHG) emissions within decades is to implement rapid, deep and sustained reductions of global CO2 emissions and other GHG. The IPCC explains that, in addition, carbon dioxide removal (CDR) will be needed to offset residual CO2 emissions from sectors that are difficult and/or costly to decarbonize fully (Arias et al 2021). CDR also has a role to play in insuring against unexpected outcomes from warming overshoot trajectories (Schleussner et al 2024). Required CDR each year is estimated to range from 4.0 (i.e. reduced global energy demand scenario) to 5.9 Gt CO2 (rapid implementation of renewables scenario) in 2030, increasing to 5.3 (reduced global energy demand) and 10.4 (focus on CDR scenario)7 Gt CO2 annually by 2050 (Lamb et al 2024), equivalent to up to 25% of current annual anthropogenic emissions. This Perspective argues for the need to accelerate CDR research, within the new ‘One-Earth’ framework proposed below. It does not address the acceptability or not of deploying CDR as our objective is instead to propose a framework within which to conduct studies that will provide the society and decision-makers with scientific information about the potential positive and negative consequences of proposed CDR techniques. Research should address not only the characteristics of the different proposed CDR methods but also their potential biogeochemical, biophysical and other side effects (table 5.9 and figure 5.36 in Canadell et al 2021). The governance aspect of future CDR deployments is not examined below as it is treated in Boyd et al (2025).

The Perspective discusses successively the need to incorporate Earth System feedbacks (CDR tax8) into CDR studies, the current and projected CDR, the CDR tax in the dynamic Earth System, and CDR efficiency and the CDR tax under Earth system scenarios, to conclude with recent insights on the complexities that emerge when a portfolio of CDR approaches is modeled, including policy aspects. The early recognition of the need to incorporate Earth System feedbacks in CDR studies has its roots in the early concepts of atmosphere and ocean and atmosphere and land being systems exchanging CO2 in both directions (Keeling 1973), which subsequently found its way into coupled (ocean) models (Maier-Reimer and Hasselmann 1987). Climate and carbon cycle feedbacks have since been routinely assessed for land and ocean (Friedlingstein et al 2006, Canadell et al 2021). It has been known for over a decade (Cao and Caldeira 2010) and also recognized by the IPCC (Ciais et al 2013) that CDR-driven carbon influxes into one reservoir would lead to counterbalancing (feedback) effluxes from all Earth’s reservoirs. This negative feedback in the Earth System’s carbon cycle, which is recognized by the CDR modeling community (Jones et al 2016, Keller et al 2018), weakens the realized atmospheric CO2 reduction compared with the expected CDR influx. Because of this tight connection between land, ocean and atmosphere (Zickfeld et al 2023), the combined land- and ocean-based CDR required to reach the annual multi-billion ton target must be studied, financed, and managed as a ‘One-Earth CDR9’.

However, the dependence of feedback strength on the background emissions scenario is not generally considered in the discussion of CDR deployment, nor is the added complication of singling out feedbacks from CDR fluxes when land- and ocean-based CDR are applied at the same time. A holistic viewpoint is needed as current CDR approaches, whether conventional or novel (Geden et al 2024), are mostly discrete and hence unconnected, whereas the Earth System imposes a ‘tax’ on the cumulative effect of all CDR methods that we term ‘CDR tax’ and must be considered in both emerging CDR markets and national carbon accounting. The CDR tax is the fraction of the atmospheric carbon captured by a CDR deployment that is not reflected in the corresponding decrease in atmospheric CO2. For example, a CDR deployment capturing 1.0 Gt CO2 year−1—and causing effluxes of 0.2 Gt CO2 year−1 levies a 20% CDR tax on the Earth System.

Conventional land-based CDR are dominant in current and planned applications (Nemet et al 2024, Pongratz et al 2024), but studying a portfolio that also includes ocean-based CDR is essential for two reasons: first, it shares the heavy lifting imposed by CDR on sustainability limits between the land and the ocean; and second, it is useful to have a portfolio of CDR methods with different dynamics and durations. Indeed, for CDR to have a long-term effect on the climate, the CO2 storage period should be at least 1000 years (Brunner et al 2024). Deploying such a portfolio requires the One-Earth approach, which takes into account the CDR tax.

Conventional CDR methods include afforestation, reforestation, forest management, and soil carbon sequestration. Novel methods consist of evolving but often untested land- and ocean-based CDR approaches, which currently contribute <0.1% of total CDR but are growing more rapidly than conventional methods (Nemet et al 2024, Pongratz et al 2024). Currently, ∼2 Gt CO2 per year of CDR is taking place, dominated by conventional methods (Pongratz et al 2024). It is necessary to study all proven and sustainable land- and ocean-based approaches, the combination of which may be needed to attain the annual multi-billion ton target.

It is unknown whether CDR of 5–10 Gt CO2 year−1 could be achievable using combined deployments of all techniques. The projected growth of existing land-based conventional CDR by 2030 is ∼0.5 Gt CO2 per year (i.e. from 2.1 to 2.6 Gt CO2 year−1, Smith et al 2024), and is limited by factors ranging from land availability and the increasing threat by climate extremes to the permanence of CO2 storage in forests. Could the development of novel CDR help meet the ∼1.4 Gt CO2 year−1 shortfall relative to 4.0 Gt CO2 year−1 based on conventional CDR (Lamb et al 2024) by 2030? Several of the proposed novel techniques are the subject of modeling, laboratory or small- or medium-scale field studies. In the ocean, no CDR approach is ready to be deployed at scale (Doney et al 2025) and model projections are exploring their regional and global potential (Keller et al 2018) while small-scale lab and field trials on a range of methods are underway.

In the flurry of current economic activities into conventional and novel CDR approaches (Presty et al 2024), the central issue of the role of the Earth System is often being overlooked. If CDR is to compensate for residual emissions, it must operate annually at the gigatonne scale. This means that CDR at deployment scale must be embedded within an Earth System framework to take into account global climate and the associated carbon cycle feedbacks (Keller et al 2018), i.e. the One-Earth CDR including the CDR tax. The carbon reservoirs and fluxes of the Earth System are being perturbed by anthropogenic CO2 emissions (Ciais et al 2013), of which the joint removal of ∼50% of cumulative emissions by the land and ocean sinks since 1750 (Friedlingstein et al 2023) is an example of a negative feedback that has ‘rebalanced’ the Earth’s carbon cycle. The role played by carbon cycle CDR feedbacks will depend jointly upon the emissions reduction trajectory and the upscaling of CDR methods (Jones et al 2016, Babiker et al 2022).

Cao and Caldeira (2010) stated that ‘In response to a reduction in atmospheric CO2, there is an efflux of CO2 from both the terrestrial biosphere and ocean to the atmosphere’. In other words, any CDR is accompanied by a CO2 efflux from natural reservoirs to the atmosphere. Matthews and Weaver (2010) explored this efflux feedback explicitly for CDR, and discussed the interplay between (negative) feedbacks on atmospheric CO2 concentrations, changes in CO2 emissions and CDR. Modeling simulations are the only way to explore future Earth System responses to continuing anthropogenic perturbations, including feedbacks within the carbon cycle and biogeophysical responses (figure 1). Current models have limitations in both representing the multi-faceted pathways of the Earth System and the inherent complexity when multiple concurrent perturbations are overlaid on simulations. For example, simulations also reveal that emission reductions would decrease the ‘CO2 fertilization effect’ on terrestrial plants (Babiker et al 2022), but would also decrease detrimental impacts from droughts, heatwaves, and fires on forests (Anderegg et al 2022).

Figure 1. Overview of CO2 fluxes and feedbacks linked to the One-Earth CDR. The application of CDR results in an influx of atmospheric CO2 at the deployment sites, and also in compensating effluxes over both land and ocean. These illustrative examples showcase the effluxes in the interconnected Earth system relative to a reference site where CDR is not applied (panel (a), purple arrows denote the net influx of anthropogenic CO2 due to high atmospheric concentrations) for the application of: (b) land-based AR, (c) ocean-based OAE, and (d) combination of AR and OAE. The green arrows represent cumulative CDR sequestration (influxes) and the orange arrows compensating effluxes (despite ongoing influxes of anthropogenic CO2; see panel (a). Compensating effluxes do not suggest a sink-to-source transition of the respective carbon sink, but rather a weakening compared to the reference state. In cases where a CDR method is applied, the dominant driver of the compensating efflux is the reduction of atmospheric CO2, which lowers terrestrial photosynthetic rates and weakens the CO2 air-sea exchange, resulting in the 4%–45% CDR tax. In panel b, AR additionally triggers several complex biogeophysical effects (horizontal solid bi-directional arrows in panels b and d denote system connectivity), affecting even remote terrestrial regions, due to changes including moisture transport. When OAE is applied (panel (c)), alkalinity disperses to adjacent regions and therefore carbon is sequestered beyond the deployment region, which makes monitoring, reporting and verifying challenging. When both AR and OAE are applied (panel (d)), the newly planted forest responds to the reduction of atmospheric CO2 and temperature by OAE, and vice versa. This simplified scheme does not include effects resulting from altered water use in terrestrial ecosystems, such as changes in seaward freshwater flux. In a One-Earth approach, these complex interconnections should be represented, thus necessitating the utilization of Earth System Models.

Download figure:

Standard image High-resolution image

Keller et al (2018) graphically represented how the carbon reservoirs in the Earth System would respond to different CDR approaches: af(re)forestation (AR), ocean alkalinity enhancement (OAE), and enhanced-weathering. Each approach caused a CO2 efflux from the non-perturbed reservoir (e.g. marine CDR causing land-based efflux), except in the case of weathering which influenced all reservoirs. This conceptual study showed that successful reduction of atmospheric CO2 concentration by CDR may cause an efflux or a decreased CO2 influx (natural and CDR-related) from land and ocean reservoirs, i.e. the CDR tax. This has severe implications since, even assuming that the CDR-driven CO2 influx could be accurately monitored (e.g., additional carbon stored in a newly planted forest), this CO2 influx would overestimate the reduction in atmospheric CO2 concentration by the amount of the CDR tax. How high this tax is, however, depends both on the ‘background’ climate evolution and the scale of CDR deployed. Therefore, carbon-climate feedbacks need to be included in their full complexity, which is not the case in many current studies.

Jones et al (2016) considered how the efficiency of land-based CDR approaches could be decreased by altering the magnitude of sinks in the global carbon cycle. Zickfeld et al (2023) also explored the dependency of (land-based) CDR efficiency on the interplay of CDR magnitude with that of the background climate signature. In a wider context, an Earth System model study simulating global deployment of OAE (Jeltsch-Thömmes et al 2024) confirmed that, although this ocean-based CDR resulted in CDR, it was offset in two ways by the Earth System, i.e. by CO2 land efflux and less CO2 oceanic influx. The realized atmospheric CO2 removal was 27% less than the total increase in the ocean carbon sink (CDR tax = 100%—Removal Efficiency, RE, in table S1 of Moustakis et al 2025). It should be noted that the latter does not equal the isolated OAE CDR flux (i.e. additional CO2 oceanic influx over the regions of OAE), but incorporates the CDR tax. The isolation of the ‘true’ OAE CDR flux is a challenging task, which requires the development of more accurate and high throughput technologies for detection and attribution because of the diffusion of added alkalinity to adjacent model ocean grid cells (figure 1), OAE mixing to depth and natural variability.

CDR simulations for large-scale terrestrial AR by Moustakis et al (2024) also reveal negative Earth System feedbacks. Their study differs from that of Jones et al (2016) in that Moustakis et al represent land-based methods where the carbon removed remains within (and can interact with) the Earth System. They report that atmospheric CO2 removal was ∼26% lower than the total increase in the land carbon sink. As above, the land carbon increase does not represent the isolated CDR flux of additional carbon stored in the newly-planted forest, but includes negative feedbacks and effluxes within the terrestrial carbon sink (figure 1).

A comparison of six Earth System models reported a CDR tax range of 9%–25% (Wey et al 2025) which, together with the values of ∼26% (Moustakis et al 2024) and 27% (Jeltsch-Thömmes et al 2024) are within the range of 4%–45% in the summary of published values of Moustakis et al (2025; table 1, 100%—RE). These CDR tax values stress the importance of conducting studies within the One-Earth CDR framework (figure 1).

Jeltsch-Thömmes et al (2024) advocate for modeling studies that probe how Earth System feedbacks will respond to CDR both in ocean and on land. Moustakis et al (2025) explored the carbon cycle responses to combined land-based (AR) and ocean-based (OAE) methods. They found that combined CDR yields a similar decrease in atmospheric CO2 as would have been expected by summing the outcome of the single-CDR model runs with AR or OAE. The linearity of the global fluxes (but less so locally—i.e. grid-cell level) upon combining the methods suggests that these compensating fluxes likely persist when combining methods, even though to further support this, modeling setups that can isolate the CDR flux of each method excluding feedbacks are needed. This study shows that combining land- and ocean-based CDR methods does not compromise the efficiency of individual applications.

From now until 2030, conventional land-based methods will likely dominate CDR (Nemet et al 2024) leading to a projected annual CDR gap of ∼1 Gt CO2 year−1 in 2030 (Lamb et al 2024). Furthermore, projections centered on limiting warming to 1.5 °C by 2050 require rapid upscaling of novel CDR (Smith et al 2024). Superimposed on this dominance of terrestrial CDR are issues that include detrimental side-effects including mega-fires (Byrne et al 2024), and carbon mis-accounting linked to natural events like mega-fires (Luers et al 2022). Taken together with the effect of negative Earth System feedbacks in weakening atmospheric CO2 reduction compared to the CDR flux (Jones et al 2016, Moustakis et al 2024), i.e. the CDR tax, conventional land-based CDR is unlikely to be able to compensate globally for hard-to-abate residual emissions.

Based on the current evidence of the limits of conventional CDR, the only way to reach the annual multi-billion ton CDR target might be to extend CDR to novel methods both on land and in oceans. Extending CDR to novel methods would not be without issues, and the development and upscaling of methods, when proven, is estimated to require decades (Boyd et al 2023).

It is essential to study CDR within the One-Earth framework because deploying CDR without including the 4%–45% CDR tax would generate unrealistically high expected rates of atmospheric CO2 removal. In addition, the latter leads to the conclusion that CDR methods are mainly bound to the sovereign territory of nation states and do not therefore require transboundary governance (Bellamy and Geden 2019), which is incorrect because it neglects the Earth-System negative feedbacks to all CDR. Hence, effects and feedbacks must be clearly isolated, and their dependency on the climate trajectory and CDR portfolio must be appropriately considered. Otherwise, confidence in the science underpinning CDR deployments would be undermined.

The One-Earth CDR concept provides scientists with an initial framework required for modeling and monitoring the joint implementation of land- and ocean-based CDR. The simple climate models used in Integrated Assessment Models typically represent the terrestrial and ocean carbon sinks and should thus, in principle, include the CDR tax, but they lack the complexity necessary to robustly estimate it. With CMIP7 on the horizon, this framework can help to guide model development and to devise the simulation setups necessary for capturing the Earth System feedbacks to land- and ocean-based CDR, and their interplay. In that regard, emission-driven projections are a necessary tool (Sanderson et al 2024), but could also be complemented by concentration-driven ones, aimed at isolating feedbacks (figure 1). We advocate research within this framework to track the relative influences of concurrent activities (emissions reductions, CDR, feedbacks) to provide accurate and regular updates (Masson-Delmotte and Males 2024) on the status of global carbon cycle.

No new data were created or analysed in this study.