Introduction

Fast dwindling carbon budgets are stimulating a lively debate on the role of negative emission technologies (NETsFootnote 1) for keeping warming below 1.5 °C and 2 °C, as illustrated by a growing number of scientific and policy discussions on the subject (Fuss et al. 2014; Geden 2015; Lomax et al. 2015; Gasser et al. 2015; Anderson 2015; Lewis 2015; Williamson 2016; Anderson and Peters 2016; Lackner 2016; Field and Mach 2017; Parson 2017; Peters and Geden 2017; Geden and Löschel 2017; van Vuuren et al. 2017; Obersteiner et al. 2018; Scott and Geden 2018). At the heart of these discussions lays new evidence from long-term climate change mitigation scenarios generated with integrated assessment models (IAMs) as summarized in recent assessments by the Intergovernmental Panel on Climate Change (Bruckner et al. 2014; Clarke et al. 2014; IPCC 2014; Kunreuther et al. 2014; Smith et al. 2014; Stavins et al. 2014).

The Fifth Assessment Report (AR5) by IPCC Working Group 3 (WG3) provided a good overview of the role of NETs for stringent climate stabilization targets. It highlighted that many 2 °C scenarios entail large-scale deployment of NETs after 2050 to compensate for residual CO2 emissions from sectors that are difficult to decarbonize, such as industry and aviation. It warned that these scenarios are mostly associated with a temporary overshoot of the climate goal and that delays in climate action and limitations in the availability of NETs can render the 2 °C goal infeasible. It also emphasized the challenges (e.g. societal concerns), risks (e.g. technological availability, biodiversity, water, food prices, inter-generational impacts) and uncertainties (e.g. geological storage, large bioenergy production) surrounding these technologies (see also Electronic Supplementary Material (ESM) for a complete review of NET statements in AR5).

Yet, the analysis of NETs in WG3 AR5 remained inaccessible because findings were scattered in various sections and sub-sections of the report (i.e. in Chaps. 2, 6, 7, 11, 13).

The recent IPCC Special Report on Global Warming of 1.5 °C (SR1.5) (IPCC 2018) filled this gap by drawing upon a set of recent reviews (Minx et al. 2018; Fuss et al. 2018; Nemet et al., 2018) that used formal methods of evidence synthesis. It further added a comprehensive analysis on the role of NETs in 1.5 °C scenarios based on newly emerging evidence. The report highlighted that all 1.5 °C scenarios with limited or no overshoot require NETs on the order of 100–1000 GtCO2 over the twenty-first century but that significant near-term emissions reductions (e.g. low energy demand, low material consumption, low GHG-intensive food consumption) can limit NET deployment to a few hundred GtCO2 without reliance on Bioenergy with Carbon Capture and Storage (BECCS). It also called attention to the lack of published pathways featuring NETs other than afforestation and reforestation (AR) and BECCS (see also Electronic Supplementary Material (ESM) for a complete review of NET statements in Fuss et al. 2018, IPCC 2018 and Rogelj et al. 2018b).

Despite all this progress in the understanding of NETs, the assessment practice on scenario evidence remains unsatisfactory. In particular, the imbalance in systematicity between the treatment of quantitative and qualitative findings is concerning. On the one hand, pathways are extracted from various studies, pulled together in a large database and analysed in a complete systematic fashion. On the other hand, the inclusion or exclusion of qualitative findings is left to the discretion of the expert review team. While analyses based on larger scenario ensembles are more robust and generate new synthetic insights, the lack of a systematic approach for qualitative evidence is problematic for at least three reasons. First, the absence of systematic methods overall in many circumstances leads to bias (Haddaway and Macura, 2018). Second, findings in reviews and scientific assessments cannot be traced back to all relevant studies. For instance, the sentence “CDR requirements are reduced under ‘sustainability’ related assumptions.” (Rogelj et al. 2018b, p. 149) only cites Strefler et al. (2018) while studies by Bertram et al. (2018) and van Vuuren et al. (2018) are equally relevant. In a similar vain, evidence resulting from the analysis of large databases—those created specifically for reviews and assessments—can be disconnected from original findings. These problems reduce the transparency of synthetic works, a crucial element of any systematic approach. Finally, we argue and will show in this article that the lack of systematicity also runs the risk of omitting important niche and newly emerging results.

To address these shortcomings with a view on the upcoming Sixth Assessment Report (AR6), we present here the first thematic review (Boyatzis 1998; Guest et al. 2012) of the available scenario literature on NETs; provide a comprehensive, traceable and easily updatable synthetic table of all statements on NETs; and re-analyse the most important ones quantitatively. The main novelty of our approach is that it links a systematic assessment of individual findings to the literature base and evidence from scenario data. As in any systematic approach, we describe the procedures to search and select the literature, to extract evidence and to evaluate the quality of included studies in a comprehensive and transparent manner (Petticrew and McCartney 2011; Minx et al. 2017a). Such a method provides a crucial synthetic account to inform upcoming climate change assessments. In particular, we contribute to the systematic assessment literature by developing and employing a computer-assisted method to review 138 studies on the subject in a comprehensive and transparent manner, allowing us to synthesize many qualitative findings. We complement this analysis by connecting it to a summary of the available quantitative evidence. Our review includes the rapidly emerging literature on the 1.5 °C limit (Luderer et al. 2013, 2018; Rogelj et al. 2015, 2018a, b; Manoussi et al. 2017; Marcucci et al. 2017; Mintenig et al. 2017; Bauer et al. 2018; Bertram et al. 2018; Holz et al. 2018; Kriegler et al. 2018a; Rickels et al. 2018; Strefler et al. 2018; Séférian et al. 2018; van Vuuren et al. 2018; Grubler et al. 2018) that include scenarios that emphasise the role of NETs even more than the 2 °C scenarios in AR5.

Importantly, our approach provides:

  • a ready-to-use method for systematically reviewing the scenario literature;

  • a comprehensive, traceable and easily updatable synthetic table that maps 1360 paragraphs from 138 studies related to NETs into 66 summary statements (see Excel spreadsheet in ESM);

  • a focused assessment of key evidence on the distinct role of NETs for limiting climate change to 1.5 °C and 2 °C, respectively (i.e. upscaling, maximum deployment, near-term mitigation trade-offs, role of socio-economic drivers, regional NET deployment);

  • an improved understanding of how deployment dynamics change as multiple NETs get deployed in small portfolios.

In the next section, we review some key findings on NETs that are important to understand scenario results and are particularly useful for the discussions of the 1.5 °C and 2 °C limits. We support this review with quantitative evidence on the dynamics of emission pathways, climate change mitigation costs, and regional CO2 sequestration whilst giving some special attention to small NET portfolios. Throughout the analysis, we discuss relevant key assumptions as appropriate. We close by providing a more comprehensive overview of the landscape of findings and discuss major open avenues for research.

Main findings from the scenario literature

The statements on NETs that we synthesized from the available literature and that are relevant to the focus of this review are organized in Table 1. We refer to them in the remainder of this article by using their unique identifier (e.g. O1, see column ID in Table 1). The methodology employed to find each statement is detailed in the ESM. It should be noted that in most mitigation scenarios, BECCS is the only explicit NET available.Footnote 2 Unless a statement is specific to a particular NET, we therefore consider BECCS as a proxy for NETs in the following discussion and as a result interpret its deployment as total NET deployment. Towards the end of this section, we explicitly consider portfolios with multiple NETs.

Table 1 Thirty-one of 66 qualitative findings retrieved from 138 selected studies and supporting the focus of this review, i.e. the role of NETs for international climate goals. Findings are clustered in 5 broad categories. Studies supporting each finding are provided and tagged in blue when they contain at least one 1.5 °C scenario and in red when they do not contain any 1.5 °C scenario but at least one 2 °C scenario. The 66 qualitative findings are available in the Electronic Supplementary Material (Table S3)

The literature highlights the importance of NETs for meeting stringent climate policy targets O1. This importance is illustrated by the large amounts of carbon withdrawn from the atmosphere in most 1.5 °C and 2 °C scenarios. In 85% of 2 °C scenarios (above 15th percentile), more than 300 Gt CO2 are removed over the twenty-first century as long as no drastic GHG emission abatement, radical changes towards sustainable lifestyles, nor constraints on technological availability and climate policy timing are imposed (see Fig. 1a1 vs Fig. 1a2–a4). This estimate increases to nearly 500 in 1.5 °C scenarios with similar assumptions (Fig. 1a5).

Fig. 1
figure 1

The role of NETs in meeting international climate targets under various technological and policy constraints and societal choices. Part B shows key statistics. The number of pathways, models and scenarios are provided in Figure 1a. The colour scheme used to differentiate scenario categories in Figures 1a and 1b are the same that is blue for 2 °C pathways with immediate action and full technology, purple for 2 °C pathways without CCS technology, orange for 2 °C pathways with delayed action to 2030, black for 2 °C pathways under SSP5 assumptions, red for 1.5 °C pathways with immediate action and full technology and green for 1.5 °C pathways with limited CDRavailability (< 5 GtCO2/year). The cyan pentagons refer to the qualitative findings in Table 1

Differences in cumulative carbon removal reflect qualitative differences in 1.5 °C and 2 °C scenarios. Global warming cannot be limited to 1.5 °C without NETs because the overshoot of the small carbon budget associated with this goalFootnote 3—resulting from socio-economic inertias—must be offset by negative emissions (Fig. 1a) (Luderer et al. 2013; Rogelj et al. 2015; Strefler et al. 2018; Rogelj et al. 2018a; Kriegler et al. 2018b). In other words, in the absence of large-scale negative emissions, scenarios without draconian near-term CO2 emission reductions are infeasible O11 (Fig. 1a6, see Table 2 for illustration). Even in recent 1.5 °C scenarios that explore goal achievement with minimum NETs, net negative emissions during the second half of the twenty-first century are required (Bertram et al. 2018; Holz et al. 2018; van Vuuren et al. 2018; Grubler et al. 2018) (Fig. 1a6). In contrast, a variety of 2 °C scenarios exist without any explicit NET deployment, although the near-term CO2 emission reductions envisaged in these scenarios are more severe than when NETs are deployed (Fig. 1a2) (Krey et al. 2014; Kriegler et al. 2014; Iyer et al. 2015; Eom et al. 2015). Moreover, the global economy is fully decarbonised by the second half of the twenty-first century in 1.5 °C scenarios D2 with a sustained period of net negative emissions thereafter. The scale-up of NETs between 2030 and 2050 occurs faster (0.3 [0.0–0.6] GtCO2/year versus 0.2 [0.0–0.4] GtCO2/year expansionFootnote 4) unless climate policy is delayed until 2030 in 2 °C scenarios (Fig. 1a3, b2) D1. In fact, once climate action is delayed until 2030, the negative emissions characteristics of 2 °C scenarios become increasingly similar to the 1.5 °C scenarios with immediate action. It is worthwhile to note that this difference in NET reliance between 1.5 and 2 °C scenarios is predicated on estimates of the climate response that remain similar to the value assessed in AR5, despite recent studies showing the potential for possible changes to climate response (Millar et al. 2017; Tokarska and Gillett 2018).

Table 2 The role of NETs in decreasing mitigation costs. Mitigation costs are the discounted sum of all costs associated with climate mitigation action over the period 2010–2100, calculated here for a 67% confidence level of limiting warming to below 2 °C (well below 2 °C case), or a 50% confidence level of limiting end of century warming to 1.5 °C by interpolating between respective scenarios. Red crosses indicate infeasible scenarios. Costs represent discounted and aggregated consumption losses (i.e. differences in consumption between a baseline and a climate policy scenario) as a share of discounted and aggregated GDP (in the climate policy scenario). A discount rate of 5% is used. Results adapted from Luderer et al. (2013)

The prominence of NETs in stringent climate mitigation scenarios is the product of two factors: geophysical requirements and economic incentives, which play very different roles in 1.5 °C and 2 °C scenarios (Minx et al. 2018). As indicated earlier, while carbon dioxide removal deployment in 1.5 °C scenarios is primarily driven by geophysical requirements —for compensating for the overshoot of the allowed carbon budget (Strefler et al. 2018; Kriegler et al. 2018b; Luderer et al. 2018)—this is not the case in 2 °C scenarios. Economic incentives however are an important feature of both 1.5 °C and 2 °C scenarios as depicted by mitigation costs in Table 2. Mitigation costs are a good indicator for understanding the overall mitigation challenge in IAMs: as mitigation costs increase as models move towards their infeasibility frontier (Azar et al. 2006; Luderer et al. 2013). Moving across the columns of Table 2 shows how mitigation costs rise: (a) as long-term ambition increases and (b) as short-term emission reductions get delayed until 2030. Looking across the columns shows the decrease in mitigation costs as NETs are introduced into the models C1 and how models are driven to infeasibility in their absence, as short-term emission reductions get further delayed and long-term ambition increases O11. No single IAM study documents pathways without NETs returning warming to below 1.5 °C by 2100, or limiting warming to well below 2 °C after a delaying action until 2030.

The lowering of mitigation costs resulting from the deployment of NETs has far reaching consequences on mitigation pathways D2. Indeed, regardless of any geophysical requirement, the introduction of NETs in IAM curb emission reductions in the short run and compensates for the carbon budget overshoot towards the end century when carbon prices are high (often reaching thousands of US dollars per ton of CO2) D2, C2. For example, Fig. 1a2 demonstrates that, until 2060, median net CO2 emissions in 2 °C scenarios with BECCS exceeds that of the median 2 °C scenario without BECCS. Consequently, this temporal reallocation of CO2 emissions mitigation by NETs allow for more fossil fuel consumption in the near-term, thus leaving more time for the energy system transition, and a longer utilisation of existing infrastructure in the power, industry and building sectors D5.

Temporal discounting is the main factor driving these dynamics. Indeed, by giving a lower weight to the costs of future mitigation actions, the use of a constant discount rate (generally 5%) during the twenty-first century makes the use of NETs in the second half of the century an economically attractive mitigation option that allow costly near-term emissions reductions to be postponed (van Vuuren et al. 2017; Obersteiner et al. 2018) D9. From these results, it is often concluded that NETs make the timing of climate mitigation policies more flexible D6. Aware of the potential misinterpretation of these findings, modellers described the risks, uncertainties and challenges in following such pathways O2, O3, O4, and have warned policy makers against postponing near-term abatement action O14.

By the end of the century, negative CO2 emissions from NETs compensate for past and current residual CO2 and non-CO2 emissions D4. Residual CO2 emissions can arise in the industry, buildings, transportation and land-use sectors and are a function of the ambition level of climate policy (Kriegler et al. 2018b; Luderer et al. 2018). They add to non-CO2 greenhouse gas (GHG) emissions like CH4, N2O and F-Gases, which can represent a large share of total residual GHG emissions depending on the mitigation portfolio represented in models (Kober et al. 2014; Gernaat et al. 2015). When BECCS is available, the power sector is often reported to be essential for decarbonisation and to compensate for these residual emissions (van Sluisveld et al. 2013; Kriegler et al. 2013a). However, other sectors can in principle also contribute since BECCS is a versatile technology which can also be used to produce synthetic fuels, heat and hydrogen T5, O15.

Little attention has been given to portfolios of NETs in IAMs even though there is a growing number of studies that feature two or more NET options—most prominently BECCS and afforestation (AR) (Rao and Riahi 2006; Wise et al. 2009; Calvin et al. 2009a; Edmonds et al. 2013; Humpenöder et al. 2014; Bertram et al. 2018; van Vuuren et al. 2018; Grubler et al. 2018), but also BECCS and direct air capture with carbon sequestration (DACCS) (Chen and Tavoni et al. 2013; Marcucci et al. 2017) or BECCS and enhanced weathering (EW) (Strefler et al. 2015). More recently, Holz et al. (2018) produced scenarios with 6 NETs (i.e. AR, BECCS, biochar, DACCS, EW, soil carbon management). There are a few noticeable findings from these studies O8. Compared to scenarios featuring only BECCS, the availability and use of one or more additional NETs makes emission reductions in the near term even less attractive and further reduces the political costs of achieving the same climate policy objective D9, C1. Indeed, any addition of negative emission potential in scenarios results in an increase in total cumulative CO2 removal over the twenty-first century and further exacerbates the intertemporal trade-off explained in the previous paragraphs D4, D11. This increase in carbon removal occurs at decreasing rates because of substitution effects. Competition between NETs for land and/or geological storage reduces the individual potential of these technologies. For example, Humpenöder et al. (2014) who examine two land-based NETs (i.e. competing for scarce land) find that the amount of CO2 sequestered by BECCS over the 1995–2095 period decreases from 600 to 400 GtCO2 when AR is added. Marcucci et al. (2017) find a lower effect of only a few gigatonnes between BECCS and DACCS, which compete for geological storage (see Fig. 2). Holz et al. (2018) also show that the contribution from AR is halved when 5 other NETs are added. These findings suggest that the availability of multiple NETs may help to better balance the various technology risks that are often scale dependent (Fuss et al. 2018). The degree to which this risk can be hedged depends however on technological competition.

Fig. 2
figure 2

The role of small NET portfolios in meeting the 1.5 °C and 2 °C climate goals. Each bar represents a scenario. NET availability is indicated by a small square below the bars. Data from Humpenöder et al. (2014) are cumulated over 1995–2095 while that of Marcucci et al. (2017) and Holz et al. (2018) are cumulated over 2010–2100

Finally, while several studies have eliminated doubts that geological CO2 storage is a major limiting factor for NET deployment in 2 °C scenarios T1, this is less clear when NET deployment with geological CO2 storage further increase—like in 1.5 °C scenarios and with the availability of additional NET options with geological storage like DACCS (Meinshausen et al. 2011; Dooley 2013; Edmonds et al. 2013; Tavoni and Socolow 2013; Vaughan and Gough 2016). Even though sequestration rate constraints can limit the amount of CO2 that can be stored annually, a meta-analysis concludes that 3900 GtCO2 are “practically”Footnote 5 available and 13,500 GtCO2 are “effectively”Footnote 6 available for storing carbon in onshore and offshore saline aquifers, depleted oil and gas fields, and deep “unmineable” coal basins (Dooley 2013). In a follow-up IAM analysis, Edmonds et al. (2013) made the assumption that 7000 GtCO2 of storage would be available globally. We compare the regional estimates of these two studies to the available 1.5 °C and 2 °C scenario data from the AMPERE (Kriegler et al. 2015b) and LIMITS (Kriegler et al. 2013b) projects (see Fig. 3). The geographical distribution of negative emissions reveals that BECCS is deployed globally (Calvin et al. 2009a; van Sluisveld et al. 2013; Edmonds et al. 2013) but particularly in regions with large biomass potential and/or large geological storage capacity such as USA, China, Former Soviet Union and Middle East D7. Interestingly, biomass trade allows regions without biomass but large geological storage like the Middle East to sequester carbon. This highlights the institutional challenges facing the large-scale deployment of BECCS. Technological change and financial requirements are also concerning particularly for developing countriesFootnote 7 T9. Nonetheless large discrepancies exist not only in quantities of geological storage but also across model results. For instance, sequestered carbon in 2 °C scenarios can differ by more than 100 GtCO2 in Africa, Europe, Former Soviet Union, India, Latin America, Middle East, USA and up to 200 GtCO2 in China. These differences are much smaller in 1.5 °C scenarios because the results come from a single model. Differences can be attributed to model structures (e.g. general equilibrium model vs partial equilibrium model, regional resolution, etc.) and (regional) modelling assumptions (e.g. geological storage potential, biomass potential, etc.) which can be more or less optimistic about implementable potentials given political, institutional and social acceptance-related constraints O4. It should be noted that IAMs use different estimates of geological storage. For instance, geological storage constraints in Marcucci et al. (2017) are based on Hendriks et al. (2004) whereas estimates used in GCAM are based on Dooley (2013) and Edmonds et al. (2014). In IMAGE, upper limits on geological storage are also based on Hendriks et al. (2004) with updates from Koelbl (2016).

Fig. 3
figure 3

Regional geological CO2 storage capacity and CO2 sequestered by BECCS. Geological CO2 storage data correspond to the “practical” category from Dooley (2013) which totals 3900 GtCO2 (left bar) and the modified version by Edmonds et al. (2013) which totals 7000 GtCO2 (right bar). Scenario data are from AMPERE (Kriegler et al. 2015a, b, c), LIMITS (Kriegler et al. 2013a, 2013b), RoSE (Kriegler et al. 2016) and Luderer et al. (2013). Scenario data for Canada and Southeast Asia were missing and so could not be included here. Similarly 1.5 °C scenarios for the Pacific OECD region are unavailable

Discussion and outlook

IAMs have played a central role in expanding the research frontiers of NETs and helping to discover policy-relevant findings about these technologies. Yet, qualitative evidence so far has not been adequately aggregated into a traceable and updatable body of knowledge and has been disconnected from reviews of quantitative evidence based on large scenario ensembles. In this study, we close this gap by systematically reviewing individual findings in a thematic synthesis of the underlying literature (Boyatzis 1998; Guest et al. 2012) connecting them explicitly to the quantitative ensemble evidence. We argue that such synthesis that systematically connects qualitative and quantitative evidence has been missing so far and impeded a comprehensive and transparent assessment of the role of NETs in climate change mitigation in previous IPCC reports. As such, this paper can help to establish a comprehensive and traceable account of evidence for upcoming IPCC AR6 that can be easily updated.

Table 1 provides an overview of the findings identified during our review. We emphasise that one should not interpret any statement omitted in the previous section as less significant. Instead we selected statements according to the focus of this review, i.e. the role of NETs in 1.5 °C and 2 °C scenarios, respectively. As such, it serves as a template for bridging qualitative and quantitative evidence systematically. For example, we did not discuss the role that BECCS can play in increasing food prices T3. Although the large-scale deployment of BECCS requires hundreds of millions of hectares to grow bioenergy T1, food prices would still be high in a world without BECCS because bioenergy would increase in importance as a mitigation technologyFootnote 8 D10. Similar choices will be required in upcoming climate change assessments that are inherently space constrained. The key point here is that comprehensive, traceable accounts of both quantitative and qualitative evidence are required to do make informed and transparent choices in this process (Huppmann et al. 2018).

Our approach could be extended in the future. First, it does not represent the entire universe of scenarios on climate change mitigation. A more comprehensive identification and selection of qualitative evidence based on a structured search in bibliographic databases such as Web of Science or Scopus could be a first step in addressing the longstanding concern of selection bias in IPCC assessments of scenario evidence (Haddaway and Macura, 2018). Second, although we are confident that our table of statements includes the major findings of the literature, our search could be extended to capture more peripheral topics (e.g. geological storage, food price). Third, our compiled scenario database contains important data gaps as all model output is not consistently made available publicly. This selection could introduce biases in our numerical results due to model and scenario design sampling. More comprehensive reporting and data provision requirements seem crucial. The qualitative findings presented here however are in good agreement with the list of major statements from the literature. Fourth, while our assessment can robustly identify and reflect insights from the available scenario literature on NETs, it is much harder to identify whether any pervasive scenario assumptions might bias these statements in a particular direction, and whether there are specific conditions under which these insights would apply.

Despite these shortcomings, synthetic research efforts within the scientific community like this study are increasingly valuable (Kowarsch et al. 2016, 2017), particularly in times of exponentially growing scientific publications on climate change (Grieneisen and Zhang 2011; Haunschild et al. 2016; Minx et al. 2017b). Not engaging in such efforts could exacerbate the difficulty for assessment bodies like the IPCC to meet their often ambitious mandates (Minx et al. 2017a). While such computer-assisted synthetic methods (Westgate et al. 2018; Nakagawa et al. 2018; Lamb et al. 2019) still require laborious inputs from and verification by humans, it allows for a more efficient and better evaluation of the veracity of findings by linking them directly to supporting studies. It may also reduce the risk of bias and misinterpretations by policymakers, who would be in a better position to confront arguments based on cherry-picking and infrequently-made claims.

Importantly, this review allowed us to identify important research gaps, some of which have been underscored in Fuss et al. (2018) and Rogelj et al. (2018b). First, it seems crucial to better understand large NET deployments frequently observed in the literature (Fuss et al. 2018; Rogelj et al. 2018b). Initial evidence suggests that this is directly related to the cost-effectiveness framework (and discount rate) underlying the models studied here, and that, even though CDR would also be large in a cost-risk analysis (Mintenig et al. 2017), a consideration of climate impacts in a cost-benefit mode would substantially decrease CDR deployments (Blanford 2013). Since climate impacts could also affect the potential of NETs, accounting for these effects in IAMs would allow identifying important trade-offs and provide additional insights on the optimal timing and levels of NETs (Rogelj et al. 2018b) E3, E4, E6, E7, E10. Second, as the wider discourse on NETs highlights that even some of the more modest deployment levels are unlikely to be reached with BECCS as a single NET, a deeper exploration of NET portfolios and their ability to hedge against some of the deployment risks are crucial O8. This would also require further exploring the adequacy of introducing NETs that have been neglectedFootnote 9 by the IAM community so far (e.g. blue carbon) as well as the synergies and trade-offs between NETs. Third, considering and understanding what benefits, risks and societal trade-offs these options could provide would be important. A number of recent studies have also highlighted the potential role of non-CO2 GHG removal technologies (Stolaroff et al. 2012; Lomax et al. 2015; Ming et al. 2016; de Richter et al. 2016, 2017). A large fraction of residual emissions at the end of the century come from non-CO2 GHG, some of which are stock pollutants that accumulate in the atmosphere. For these species, GHG removal appears a promising avenue for future research. Fourth, not only do large-scale deployments of NETs impact regional economies, energy systems and carbon budgets but they can also lead to food price increases (Wise et al. 2009; Calvin et al. 2009a; Reilly et al. 2012; Kreidenweis et al. 2016), local air and water pollution, water overuse and biodiversity losses (Williamson 2016) T3, E1. Consequently, NETs should be examined under a broader spectrum of societal and environmental goals. To that end, Planetary Boundaries and Sustainable Development Goals provide useful frameworks for future research E8, O7. Finally, net negative CO2 emissions, which could amount to a few tens of gigatonnes of CO2 by the end of century when carbon prices are high, would require substantive and sustained support from financing bodies (e.g. governments) O4. In an intertemporal optimisation framework, this can be financed by a banking and borrowing scheme. However, whether such large-scale financing product could see the light in reality is unclear at the moment. Therefore, it would be essential to investigate the role of NETs in an intergenerational modelling framework and from a political and business perspective.