1. The overlooked ethics of negative emissions

With the signing of the Paris Agreement, the international community pledged to limit global temperature rise to ‘well below’ 2°C. This seems unlikely to be achieved by conventional mitigation alone. Around 87% of the Intergovernmental Panel on Climate Change (IPCC)’s 2°C scenarios include ‘negative emissions techniques’ (NETs) [Reference Fuss, Canadell, Peters, Tavoni, Andrew, Ciais, Jackson, Jones, Kraxner, Nakicenovic, Le Quéré, Raupach, Sharifi, Smith and Yamagata1], while the more stringent 1.5°C target pushes this number to 100% [Reference Rogelj, Luderer, Pietzcker, Kriegler, Schaeffer, Krey and Riahi.2]. By expanding the remaining carbon budget, NETs render current emissions ‘overshooting’ in the first half of the century compatible with stringent warming targets by the century's end [Reference Edenhofer, Pichs-Madruga, Sokona, Farahani, Kadner, Seyboth, Adler, Baum, Brunner, Eickemeier, Kriemann, Savolainen, Schlömer, von Stechow, Zwickel and Minx3]. As national emissions targets continue to fall well short of the Paris goals [4], NETs become increasingly indispensable.

However, NETs also raise serious ethical concerns. In a widely cited commentary, Anderson and Peters claim that NETs are “an unjust and high-stakes gamble” [Reference Anderson and Peters5] that might not work, and which would be ethically unacceptable if it did. Others also point to a ‘bet’ on NETs emerging, involving a gamble on successful implementation at extremely large scales to compensate for stalled mitigation [Reference Fuss, Canadell, Peters, Tavoni, Andrew, Ciais, Jackson, Jones, Kraxner, Nakicenovic, Le Quéré, Raupach, Sharifi, Smith and Yamagata1]. Despite this, NETs remain marginal components of a normative literature still dominated by discussion of solar radiation management (SRM).Footnote ⁱ

General references to ‘geoengineering’ can obscure the ethical differences between NETs and SRM. These stem from their different characteristics and aims. First, while SRM merely masks the radiative forcing of atmospheric emissions, NETs remove atmospheric carbon. Classifications sometimes locate NETs alongside conventional mitigation since both share the aim of reducing atmospheric emissions concentrations.Footnote ⁱⁱ Second, the most-discussed SRM technique, sulfate aerosol injection (SAI), is potentially fast acting and may be very inexpensive,Footnote ⁱⁱⁱ while NETs are medium-timescale options with costs comparable to mitigation. Third, unlike SRM, most NETs could be implemented within particular jurisdictions. There are also substantial differences between particular techniques. Much of the controversy about NETs to date is directed at bioenergy with carbon capture and storage (BECCS), a technique which removes carbon as biomass is grown and sequesters this underground when biomass is burnt. Substantially less tends to be said about other techniques, including afforestation and reforestation (AR); forms of enhanced weathering (EW) involving the dispersal of crushed silicates upon soils or of carbonate and silicate materials in the ocean; direct air capture (DAC), which extracts CO₂ from the atmosphere via chemical solvents, or via calcium or sodium-based ‘wet scrubbing’; ocean fertilization (OF) using iron particles to increase CO₂ absorption; and soil carbon sequestration (SCS), which increases the carbon stored in organic matter contained in soils, and restores previously degraded soils [Reference Edenhofer, Pichs-Madruga, Sokona, Farahani, Kadner, Seyboth, Adler, Baum, Brunner, Eickemeier, Kriemann, Savolainen, Schlömer, von Stechow, Zwickel and Minx3].Footnote ^iv Given these differences, it is reasonable to examine the ethical implications of individual techniques on their merits [Reference Preston13,Reference Baatz, Heyward and Stelzer14].

To date, however, this has not occurred. Instead, the ethics of NETs remains largely overlooked.Footnote ^v This may be due to a fourth difference. Grasping the potential implications of NETs requires reflecting upon scenarios created by integrated assessment models (IAMs). Unsurprisingly, ethicists tend not to be well-informed about IAMs, nor about their underlying assumptions. Unlike SRM, which has provoked significant debate despite the availability of relatively little modelling, the uncertainties and risks raised by NETs depend greatly upon assumptions about how NETs might be implemented. These assumptions are reflected in and illustrated by IAMs. Thus, while the Royal Society concluded that NETs pose ‘fewer uncertainties and risks’ than SRM [Reference Shepherd, Caldeira, Cox, Haigh, Keith, Launder, Mace, MacKerron, Pyle, Rayner, Redgwell, Watson, Garthwaite, Heap, Parker and Wilsdon22], IAMs reveal that this judgement may be true given some assumptions about the future, or false given others. Assessing the ethics of NETs requires reflecting upon these various possibilities, and the potential trade-offs these imply for climate mitigation, sustainable development and global justice.Footnote ^vi In order to do so, ethical reflection must be grounded in scenario evidence. At the same time, analysis must also go beyond this evidence. Not only are scenarios at best merely indicative of the future, they implicitly build upon assumptions that are by no means ethically neutral. These normative assumptions are seldom made explicit or adequately interrogated. There is thus a danger that some normative claims (e.g. historical responsibility) will be ‘framed out’ via scenario design itself [Reference McLaren and Preston24]. There is also a danger of reproducing expert visions of technological futures, while reducing questions of value to epistemic questions of quantified distribution or technological feasibility [Reference Flegal and Gupta25].Footnote ^vii

The aim of this paper is thus twofold: to clarify the ethical issues raised by NETs, and to attempt an assessment of their severity via the comparison of different climate futures. The comparative, interdisciplinary approach adopted here promises to be of greater relevance for climate policy than existing ethical analyses, which engage only fleetingly (if at all) with integrated assessment modelling [Reference Kowarsch, Edenhofer, Heyward and Roser26]. In the next section, I sketch three concerns that appear particularly urgent. First, there is the potential for NETs to displace near-term mitigation. Second, there is a potentially risky and dangerous ‘bet’ on NETs emerging. Third, it may be dangerously hubristic to assume that large-scale implementation of NETs can be adequately controlled or managed.

2. Three key concerns: mitigation obstruction, betting and hubris

3. Assessing NETs under three alternate climate futures

In order to choose between different futures, we must be able to imagine them first. Although we might identify possible harms or risks of different negative emissions techniques, this depends upon highly stylized assumptions about future implementation. Obviously, we do not know the circumstances under which NETs might be used. These circumstances include the climate targets pursued (i.e. 2 or 1.5°C), global emissions trajectories, development trends including population growth and poverty alleviation, technological innovation, international cooperation and so on. The wide variety of possible implementation scenarios imply markedly different risks or benefits from NETs.

One way to illustrate the ethical implications of such complex options is to reflect upon future scenarios. The latest development of pathway analysis features five climate futures, the ‘Shared Socioeconomic Pathways’ (SSPs) [Reference O'Neill, Kriegler, Ebi, Kemp-Benedict, Riahi, Rothman, van Ruijven, van Vuuren, Birkmann, Kok, Levy and Solecki66]. The SSPs are hypothetical worlds estimating alternative rates of economic development, technological change, population dynamics, greenhouse gas emissions, the state of geopolitics and so on. They will likely play a crucial role in future IPCC assessments [Reference Moss, Edmonds, Hibbard, Manning, Rose, van Vuuren, Carter, Emori, Kainuma, Kram, Meehl, Mitchell, Nakicenovic, Riahi, Smith, Stouffer, Thomson, Weyant and Wilbanks67]. The SSPs feature divergent policy narratives and are used to model and quantify developments over the century [Reference Bauer, Calvin, Emmerling, Fricko, Fujimori, Hilaire, Eom, Krey, Kriegler, Mouratiadou, Sytze de Boer, van den Berg, Carrara, Daioglou, Drouet, Edmonds, Gernaat, Havlik, Johnson, Klein, Kyle, Marangoni, Masui, Pietzcker, Strubegger, Wise, Riahi and van Vuuren68]. Each relies upon a baseline emissions trajectory in the absence of climate policies and envisages different degrees of reliance upon NETs.Footnote ^xx Given space constraints, I limit myself to comparison of three SSPs. However, two caveats are in order. First, much existing modelling (including the SSPs) features only two techniques, BECCS and AR. Nonetheless, I also consider the implications of other techniques on the basis of what has been identified above. Second, these reflections cannot be more than indicative since we do not know what sort of societies will exist in future and what priorities these societies will have. As we will see, assumptions of this sort determine the severity of concerns with NETs.Footnote ^xxi

Let us begin with an alarming example. Consider SSP5, which is a climate future envisioning a continuation of fossil-fuelled development for most of the century [Reference Kriegler, Bauer, Popp, Humpenöder, Leimbach, Strefler, Baumstark, Bodirsky, Hilaire, Klein, Mouratiadou, Weindl, Bertram, Dietrich, Luderer, Pehl, Pietzcker, Piontek, Lotze-Campen, Biewald, Bonsch, Giannousakis, Kreidenweis, Müller, Rolinski, Schultes, Schwanitz, Stevanovic, Calvin, Emmerling, Fujimori and Edenhofer69]. SSP5 implies the greatest overshooting of atmospheric greenhouse gas concentrations, and the greatest requirement for carbon removal. Although considered a worst-case climate scenario with a baseline warming of greater than 5°C, SSP5 is still ‘technically feasible’ with the 2°C target. The achievement of this goal would require implementing NETs at seemingly incredible scales. Assumptions such as these reflect an extremely dangerous bet on negative emissions. In the first place, the lack of effective climate policies for most of the century implies that upscaling would be very costly, and for this reason unlikely. Moreover, the size of the NETs gamble in an SSP5 world is staggering. Because of this, any side effects from NETs are also likely to be extreme. For instance, under SSP5 bioenergy becomes the dominant global driver of cropland expansion after 2050. At these scales, land-based NETs would create very severe conflicts between agriculture, bioenergy and AR, especially in key areas such as the tropics, greatly increasing extinction pressures and biodiversity loss. Such a scenario appears closest to what Shue has in mind when labelling BECCS as a seriously unjust climate gamble. Given biophysical limits to land-based techniques, DAC in conjunction with conventional CCS may be more likely. But this would raise the same problem of DAC functioning as a back-stop technology, the upscaling of which remained speculative (and extremely costly) in the absence of complementary climate policies. Perhaps the worst aspect of such a bet is that it seems to create a forced choice for future generations between NET implementation at extreme scales and abandoning the 2°C target.

SSP5 also raises extreme concerns with hubris. The sheer scale of carbon removal required may dramatically overestimate our abilities to understand and control natural systems. One key uncertainty are thresholds for biophysical ‘tipping points’ [Reference Lenton, Held, Kriegler, Hall, Lucht, Rahmstorf and Schellnhuber70]. The most well-known examples include the melting of Arctic permafrost, which would release enormous quantities of both carbon dioxide and methane, the melting of Greenland and Antarctic ice sheets, which would significantly raise sea levels, or the disruption of Atlantic meridional overturning circulation. However, modelled pathways currently ignore tipping points. Shue argues that this is a further reason to think that NETs at scale would be gravely unjust [Reference Shue16]. All the same, if the thresholds for tipping points are higher, NETs may instead buy time until they are eclipsed.

SSP5 represents an extreme case of mitigation obstruction. With no effective climate policies until late in the century, fossil fuel reliance would expand beyond business as usual assumptions. According to the SSP5 narrative, this development leads to substantial human development gains, lifting many out of extreme poverty. While this appears implausible given current global trends of worsening inequality [Reference Alvaredo, Chancel, Piketty, Saez and Zucman71], under such a scenario policy-makers would likely continue to see climate mitigation and development as mutually exclusive goals.Footnote ^xxii Perhaps more realistically, the continued presence of fossil fuel industries for much of the century implies severe obstacles to mitigation and a prioritization of currently powerful economic interests. The steepness of reductions required in the second half of the century, the lack of salience of climate change as a policy concern, and the path-dependencies of fossil fuel infrastructure point to an extreme mitigation obstruction generally. In such a case, NETs would be expected to further exacerbate this mitigation obstruction. Now, if significantly delaying mitigation is unjust despite the future availability of NETs [Reference Shue16,Reference Baatz, Ott and Preston21], SSP5 would present the clearest case of climate injustice. Although SSP5 may be technically feasible with the 2°C target according to narrow modelling assumptions, the upscaling this implies is extremely implausible, ignores climate damages including the possibility of exceeding tipping points and implies unbounded optimism about the potential of NETs. Of course, this judgement may be tempered if the hopeful assumption of extreme poverty reduction in SSP5 were borne out. But even if it were, there would instead be an intergenerational conflict between those made better off from fossil fuelled development and future generations who stand to be worse off because of climate change. Nonetheless, such a dynamic could not be continued indefinitely since a point would eventually be reached at which climate damages undermined potential poverty reductions.

Contrast this with SSP4, a climate future in which inequality increases steadily over the century, both between nations and within societies [Reference Calvin, Bond-Lamberty, Clarke, Edmonds, Eom, Hartin, Kim, Kyle, Link, Moss, McJeon, Patel, Smith, Waldhoff and Wise72]. SSP4 is a tale of two worlds, one which is high-technology, wealthy, well-educated and at stable population levels, and one which is low-technology, labour-intensive, uneducated, much poorer but much more populous. While less extreme than in the previous world, the risks of a bet on NETs may nonetheless be severe. Mitigation occurs very slowly until the mid-century mark, locking in a substantial requirement for NETs. How large this requirement turns out to be, and the distribution of risks that emerge as a result, depends greatly upon the priorities of the increasingly powerful elites controlling policy in SSP4. These elites are likely to be less accountable even to their own citizens as inequality increases. But the interests of those elsewhere, and especially the poorest among them, are likely to fare significantly worse. Consider that SSP4 features much regional and global variation in the effects of the land-based impact of NETs. Cropland devoted to bioenergy would greatly expand, especially in low- or middle-income countries. While forest coverage would increase in middle- and high-income countries, such gains entail greater deforestation in poor countries to meet the agricultural demands of the wealthy world. This will greatly increase global food prices, disproportionately affecting the poor. The carbon price is predicted to be above $2000/tCO₂ by 2100, a very large increase that would inflate food prices seven times over [Reference Calvin, Bond-Lamberty, Clarke, Edmonds, Eom, Hartin, Kim, Kyle, Link, Moss, McJeon, Patel, Smith, Waldhoff and Wise72]. In a very unequal world, this is likely to cause mass starvation [Reference Shue16]. Avoiding food shortages or sudden food price increases would require the establishment of agreements on sustainable biomass production, without which a kind of ‘energy colonialism’ may emerge incentivizing conversion of forests, existing cropland and marginal land in the global South [Reference Gomiero, Paoletti and Pimentel73]. But in a world like SSP4, this seems unlikely. Elites may be incentivized to gamble on larger NETs implementation if risks would be primarily borne by the increasingly powerless global poor.

SSP4 implies considerable, although mixed concerns with hubris. First, SSP4 implies a shift to a more technocratic global paradigm of planetary management. For instance, unlike in other worlds, nuclear power becomes available in SSP4, which would both decrease reliance on fossil energy and provide the power input requirements of DAC. As elites became less accountable, the choice of siting locations of carbon storage and nuclear waste are likely to be imposed from above rather than subject to public consultation or deliberation. A greater willingness by elites to engage in large-scale management of ecological systems implies few restrictions on future research, likely opening the door to more extensive experimentation with EW and OF, and perhaps also SRM. However, while implying an outlook of dominance or mastery over nature, technocratic responses do not necessarily entail worse sustainability outcomes. For instance, in order to preserve authoritarian political regimes, elites may choose to prioritize longer-term environmental sustainability.

Obstruction of near-term mitigation is implied by the SSP4 narrative itself, which states that the global elite will react ‘quickly and decisively’ to implement climate policy, but only in the second half of the century. Although this is not due to NETs alone, the availability of NETs would exacerbate such an effect. But as with hubris, assessing the role of NETs in such a situation requires knowing the priorities of elites. For instance, elites may decide to delay mitigation longer since NETs are assumed to be available and since there is little accountability for any potential side effects, especially if these could be displaced elsewhere.

Finally, let us consider an optimistic case. As we noted, NETs make possible the achievement of more stringent warming targets. Considered solely in terms of primary climate impacts, lower climate stabilization outcomes appear closer to the requirements of justice. While this conclusion can be undermined by the potential side effects of NETs, such side effects may be more benign. Consider SSP1, an optimistic future in which a variety of global development goals are achieved while economic growth is de-coupled from emissions [Reference van Vuuren, Stehfest, Gernaat, Doelman, van den Berg, Harmsen, de Boer, Bouwman, Daioglou, Edelenbosch, Girod, Kram, Lassaletta, Lucas, van Meijl, Müller, van Ruijven, van der Sluis and Tabeau74]. As a result, ethical concerns with NETs appear relatively minimal in such a world. Consider the potential bet upon NETs. In SSP1, in contrast to all other modelled worlds, effective mitigation policies are implemented in the first half of the century. The result is the smallest concentration of atmospheric emissions and the smallest overshooting. This renders both the 1.5 and 2°C targets much easier to achieve and implies the smallest requirements for negative emissions. Although NETs are still required for 1.5°C, and remain likely even for 2°C, there is much less overshooting and hence much less deferral of mitigation to the second half of the century. Under such optimistic assumptions, implementing even solely land-based NETs to achieve 2°C would enable a gain of 63 million hectares in forest land [Reference Fricko, Havlik, Rogelj, Klimont, Gusti, Johnson, Kolp, Strubegger, Valin, Amann, Ermolieva, Forsell, Herrero, Heyes, Kindermann, Krey, McCollum, Obersteiner, Pachauri, Rao, Schmid, Schoepp and Riahi75]. This implies that the sorts of alarming biophysical implications associated with BECCS do not apply to such a future. Moreover, the implementation of effective climate policies, especially carbon pricing, early in the century would lessen concerns with NETs upscaling. Instead, innovation driven by carbon pricing would lead to cheaper, more effective technologies being available earlier in the century. The optimism of SSP1 extends to geopolitics, where greater cooperation between nations decreases mitigation and NETs challenges even further. Given this, there is also the least likelihood that risks will be deferred onto the vulnerable, who have a more effective voice in political decisions than in other worlds.

Unsurprisingly for a ‘green growth’ scenario, the potential hubris of NETs also appears dramatically lower. SSP1 features the smallest global human footprint given the lowest population of all worlds, decreased materialism and status consumption, and decreased demand for animal products. Economic activity gradually shifts so as to prioritize wellbeing improvements and global consciousness about environmental problems improves greatly. These trends are diametrically opposed to SSP5, which envisions an expansion of current unsustainable consumption patterns of the rich world to most of the global population. Given the growth in environmental consciousness in SSP1, NETs with larger effects on natural systems such as BECCS may be perceived to be more problematic and hence may be less utilized. Techniques that enhance natural processes, such as AR and SCS, may be preferred, while techniques involving the dispersal of chemical or mineral compounds into open ecological systems, such as OF and ocean-based EW, may be discarded on principle in this more ecologically-minded future.

Finally, the potential for mitigation obstruction would be least worrisome in such a future world. SSP1 features the smallest absolute level of emissions and the fewest economic, political and social barriers to effective mitigation. Given such assumptions, it is also least likely that ambitious mitigation would be substantially deferred. At the political level, there would be relatively little incentive for policy-makers to defer mitigation in order to pursue other short-term goals. This is because a variety of sustainable development goals are envisioned to be achieved in SSP1, including universal education, the eradication of extreme poverty and sustainable economic growth decoupled from emissions. While this presents a comparatively rosy picture, it is worth noting reasons for scepticism. SSP1 envisions a future in which sustainable growth is de-coupled from emissions, while moderate population growth and improved living standards do not worsen existing burdens upon planetary resources. Leaving aside the economic debate about whether de-coupling of economic growth and emissions is possible, it is in any case clear that such a future is currently very far from reality.