The consequences of inaction on carbon dioxide removal

Carbon dioxide removal (CDR) will be essential to meet the climate targets, so enabling its deployment at the right time will be decisive. Here, we investigate the still poorly understood implications of delaying CDR actions, focusing on integrating direct air capture and bioenergy with carbon capture and storage (DACCS and BECCS) into the European Union power mix. Under an indicative target of − 50 Gt of net CO 2 by 2100, delayed CDR would cost an extra of 0.12 − 0.19 trillion EUR per year of inaction. Moreover, postponing CDR beyond mid-century would substantially reduce the removal potential to almost half ( − 35.60 Gt CO 2 ) due to the underused biomass and land resources and the maximum technology diffusion speed. The effective design of BECCS and DACCS systems calls for long-term planning starting from now and aligned with the evolving power systems. Our quantitative analysis of the consequences of inaction on CDR — with climate targets at risk and fair CDR contributions at stake — should help to break the current impasse and incentivize early actions worldwide. fossil power generation technologies and their retrofit with CCS systems (coal and natural gas); (ii) firm “clean” technologies, such as nuclear, biomass, hydropower, and geothermal; (iii) intermittent technologies, such as wind onshore, wind offshore, solar photovoltaic open-ground and solar photovoltaic flat roof installation; and (iv) CDR technologies, i.e., BECCS based on six different types of biomass resources and two DACCS technologies. RAPID models the entire supply chain of BECCS, i.e., from the cultivation, harvesting, and pelletizing of the biomass resources to road transportation and conversion in power plants with CCS systems. The model considers that biomass-based power technologies compete for the second-generation biomass resources available, including high-productivity energy crops grown on marginal (i.e., short-rotation woody willow and two perennial grass sources, miscanthus and switchgrass) and residues from agriculture and forestry Regarding DACCS, RAPID includes proven technology based aqueous


Introduction
Due to the growing carbon emissions and rising global temperatures, carbon dioxide removal (CDR) will become essential to combat climate change [1][2][3][4] . According to the most recent integrated assessment modeling scenarios (IAMs), limiting global warming to 1.5°C will require deploying CDR to remove 10-20 Gt/yr of CO 2 over the 21 st century, and cutting emissions sharply to reach carbon neutrality around mid-century and ultimately become carbon-negative 2,5,6 .
CDR technologies and practices deliver net negative emissions by removing and sequestering CO 2 from the atmosphere 7,8 . Nature-based strategies sequester the CO 2 in natural sinks (e.g., afforestation/reforestation, AR, and tailored agricultural practices), while engineered CDR stores the CO 2 either in geological sites or minerals (e.g., enhanced weathering, Bio-Energy with Carbon Capture and Storage, BECCS, and Direct Air Carbon Capture and Storage, DACCS) [8][9][10] . To date, most IAMs already include AR and BECCS 11 , while other CDR options with low technology readiness levels and limited removal potential are often omitted 1,5,8 . BECCS is particularly appealing because it removes CO 2 while delivering renewable and reliable energy 12 ; however, it can lead to large impacts on ecosystems and biodiversity, which could be alleviated by resorting to DACCS [13][14][15][16] .
DACCS shows a large removal potential limited mainly by the storage capacity 5,17 , yet its large heat and power requirements hamper its large-scale adoption 16,18,19 , suggesting that an optimal regionalized portfolio of CDR strategies should be sought.
The CDR deployment to date has been minimal 20,21 with only 1.5 million t CO 2 /yr removed via BECCS 9,22 and around 0.01 million t CO 2 /yr 18 with DAC technologies, often deployed without long term CO 2 storage. The main barriers for CDR deployment include the lack of consensus on the need to start CDR today -as it is often perceived as "a problem for later" -, the absence of market incentives and strong political drivers, and governance challenges. Moreover, the uncertainties surrounding CDR have led to the so-called "moral hazard" and "betting" concerns about negative emissions 1,23-26 , which have delayed both emissions cuts and CDR deployment 27,28 under the assumption that CDR could be adopted to the extent and with the motion needed to meet the climate goals. In practice, however, future technological, social, and environmental barriers remain largely unexplored [29][30][31] , which may hinder the implementation of CDR and the attainment of the longterm temperature targets 26,[32][33][34][35] .
Here we fill this gap by exploring the economic, environmental, and technical implications of CDR inaction, focusing on BECCS and DACCS deployment in the European Union (EU) as key engineered CDR strategies intrinsically linked to the evolving energy sector 51,52 . The EU is an appealing regional case study, given its historical leadership in climate change mitigation and expected role in future CDR actions [53][54][55] . The EU has proposed a legally binding climate-neutrality target by 2050 56-58 , yet its continued and residual emissions from hard-to-abate sectors (e.g., aviation and agriculture) call for urgent CDR measures in the countries' agendas yet to be defined 55 . To carry out our analysis, we develop a bottom-up multi-period energy systems optimization model, named RAPID (as an acronym of RemovAl oPtImization moDel), which captures the interplay between CDR technologies (BECCS and DACCS) and the power sector. In essence, RAPID identifies optimal roadmaps -to achieve a net CO 2 emissions target at minimum cost or to remove a maximum amount of CO 2 -considering a prolonged delay of the CDR actions entailing shorter time frames for climate action (details in Methods). Our integrated model performs foresight planning assuming various starting points for BECCS and DACCS deployment and considering realistic technology diffusion rates and resource potentials to shed light on the implications of delaying CDR actions.
Our results highlight the urgent need to deploy CDR early to meet the climate goals on time and in the most cost-effective and fair manner. In the EU, each year of inaction could increase the removal costs quite substantially (e.g., 0.12-0.19 trillion EUR2015 per year of inaction to deliver −50 Gt of net CO 2 emissions).
Moreover, postponing CDR actions would drastically reduce our removal potential due to the underuse of resources and slow capacity expansions of the CDR technologies, putting the EU at risk of missing targets (e.g., from −73.73 to −35.60 Gt CO 2 by delaying CDR action from 2020 to 2050, respectively). Early integration of BECCS and DACCS in the power system would optimize the use of biomass, land, and geological storage resources to maximize the removal potential, while underpinning the penetration of intermittent renewables sources in the power mix. Our results aim to raise awareness of the economic and environmental consequences of delaying CDR actions to promote public incentives and private investment to enable CDR deployment in an effective, fairly, and timely manner.

Results and discussion
Consequences of delayed CDR actions: extra-costs and under-exploitation of resources We start by analyzing the implications on costs and emissions of delaying the large-scale deployment of BECCS and DACCS. To this end, we use RAPID to identify the minimum cost (Fig. 1a) and maximum net negative emissions potential (Fig. 1b) roadmaps, starting the CDR deployment at different points in time in 2020-2100 (optimal solutions every five years depicted with markers in Fig. 1). RAPID identifies the optimal portfolio of power technologies, BECCS and DACCS, including their location and installed capacities, which may vary over time to meet a given energy demand pattern and CO 2 removal target (details in Methods).
We first investigate the economic implications of delaying CDR for various net CO 2 emissions targets, finding that postponing CDR increases the removal cost, first smoothly and then sharply as we further postpone We next focus on the feasibility of the removal roadmaps, assuming an emergency plan to maximize CDR, starting actions in different years. We find that delaying CDR constrains the total removal potential substantially (Fig. 1b), again, first smoothly and then sharply. Notably, postponing the deployment of BECCS and DACCS beyond 2050 might prevent the EU from removing −50 Gt CO 2 before 2100 (i.e., EU emissions emitted in the last decade 55,59 ), while delays beyond 2080 might even impede reaching carbon neutrality. The EU would maximize its CDR potential by deploying BECCS and DACCS from today (i.e., net −73.73 Gt CO 2 by 2100 starting from 2020, Fig. 1b). This maximum CDR potential would be constrained by the geological storage capacity in the EU, estimated at 90.53 Gt CO 2 for hydrocarbon reservoirs and aquifers 60 , but subject to social acceptance issues, regulatory limitations, and economic feasibility challenges. Our results only consider domestic biomass resources and onshore geological sites, although trade of biomass and CO 2 beyond the EU borders and offshore storage sites may substantially increase the potential despite facing international governance issues 61,62 . Moreover, other CDR strategies beyond the energy sector (e.g., planting trees or alternative agricultural practices improving soil carbon sequestration) and other CO 2 storage options (e.g., mineral trapping) could further increase the EU's CDR ambition.
The gradual temporal decline in CDR potential is due to the underused resources (i.e., biomass residues and unexploited marginal land to grow energy crops) and the maximum diffusion rates of the CDR technologies.
Lacking sufficient BECCS capacity to process the biomass residues, they would degrade and eventually release the biogenic carbon to the air in the form of CO 2 , while the unused marginal land would represent an opportunity lost. Furthermore, the technology diffusion factor constraining the deployment speed (e.g., 20% of annual capacity up-scaling 63 ) critically limits the maximum attainable DACCS capacity. The potential reduction varies linearly over time for BECCS (green curve in Fig. 1b) and follows a sigmoidal shape for DACCS, with a critical point around 2045. This behavior is due to the unused biomass and land resources, the main factors constraining BECCS, which accumulate almost linearly over time. In contrast, the maximum attainable DACCS capacity, only limited by the diffusion rate, increases exponentially over time; consequently, further delays result in much fewer DACCS plants ready to be deployed on time.
Overall, in both the cost-optimal and the maximum removal roadmap, BECCS emerges as predominant regardless of the starting year, providing double benefits by contributing to CDR while delivering reliable power to meet the demand (CDR contribution breakdown in pie charts in Fig. 1a,b). When maximizing the CDR potential, the DACCS contribution increases with later starting years, while the BECCS contribution declines due to the loss of biomass potential (e.g., DACCS from 23% starting in 2020 to 46% in 2045, Fig. 1b for the yellow profile). However, the initial capacity for DACCS of 1 Mt of CO 2 captured per year -reflecting the current scale ambition-and the maximum growth rate of 20%/yr observed in the historical deployment of power technologies 19,50,63 would strongly limit the maximum DACCS capacity when actions are further delayed.
Hence, deploying DACCS from 2050 would result in a contribution of −30.57 Gt of CO 2 by 2100 (i.e., 42% of the total gross −72.94 Gt of CO 2 removed), while the DACCS share starting after 2080 would become negligible (i.e., <2%). Notably, the synergetic integration of BECCS with DACCS into the energy system emerges as an appealing option to enhance the CO 2 removal capacity (yellow curve showing always higher net CO 2 removal than green and blue profiles). Therefore, deploying DACCS appears as a complementary option under a rapid emergency deployment, lacking biomass resources or if BECCS deployment is constrained by the diffusion speed, being environmentally benefited from the carbon-negative electricity supplied by BECCS at the expense of increasing costs 19,51,64 .

Emissions pathways through 2100 for delayed CDR actions
We next analyze the emission pathways, considering three representative scenarios, namely, "Now", "Slow", and "Late", which maximize the removal potential starting CDR actions in years 2020, 2055, and 2085, respectively (Methods section). The net CO 2 emissions balance accounts for the: i) gross removal, i.e., total CO 2 removed from the air, either via photosynthesis (BECCS) or through physicochemical processes (DACCS); ii) anthropogenic life cycle emissions embodied in the power technologies, BECCS and DACCS supply chains; and, iii) CO 2 emissions during biomass/natural gas combustion in BECCS/DACCS facilities, respectively, due to capture efficiencies below 100%. These emissions contributions vary across pathways, which differ in the availability of resources and maximum BECCS and DACCS capacities (constrained by the technology diffusion rates).
In the "Now" scenario ( Fig. 2a), starting CDR actions in 2020, the gross negative emissions would amount to −94.05 Gt CO 2 by 2100, mostly provided by BECCS (77% in absolute terms) and a small contribution by DACCS (23%), yielding −73.73 Gt of net CO 2 removed from the atmosphere. This scenario shows the largest positive residual emissions (+20.31 Gt CO 2 ), 63% attributed to the uncaptured biogenic CO 2 and the life cycle emissions linked to BECCS supply chains (+7.11 and +5.74 Gt CO 2 , respectively), 30% due to the life cycle emissions of the other power technologies (excluding BECCS), 5% associated to the life cycle emissions from DACCS and, finally, a marginal contribution from the CO 2 transportation and storage infrastructure (<+1%). This scenario would fully exhaust the domestic storage capacity in the EU, which would act as a bottleneck for CDR. Furthermore, almost all the biomass resources available would be exploited (i.e., 95% of the biomass residues and 94% of marginal land), with a small amount lost due to the limited rate at which BECCS could be scaled up during the first years ( Supplementary Fig. 2). The overall storage efficiency -i.e., total net CO 2 removed per kg of CO 2 stored-would reach 81%, where most geological sites would store the biogenic CO 2 captured via BECCS (71%), a smaller amount of atmospheric CO 2 captured with DACCS (24%), and finally the captured emissions linked to the heating needs of DACCS (5%). Notably, when starting CDR actions today, the geological capacity needed to store the CO 2 captured in fossil power plants with CCS would be negligible, thus delivering the maximum CDR constrained by the domestic storage capacity (Fig. 3). DACCS would play a role complementing BECCS and ultimately helping to remove CO 2 at the pace required, benefitting from the carbon-negative electricity delivered in the system, and exploiting its flexibility to be located closer to the geological sites in countries with scarce biomass resources 51,64 .
In the "Slow" scenario starting CDR actions in 2055, the maximum gross and net removal potential would drop considerably (−49.61 Gt and −35.60 Gt, respectively, vs. −94.05 Gt and −73.73 Gt in the "Now" scenario), while the storage capacity would not be fully exhausted (only 57% of the capacity utilized). Notably, net negative CO 2 emissions would not be achieved until 2070 due to the need to offset the residual emissions taking place until that year. Before 2080, the deployment rate would limit the BECCS removal capacity, while beyond 2080, biomass resources would become the bottleneck (residues and land). 86% of the residues and 90% of the marginal land available from 2055 to 2100 would be exploited, representing only 63% and 57% of their respective total potentials (if actions were started in 2020 and continued until 2100). In contrast, the maximum deployment rate would constrain the DACCS capacity ( Supplementary Fig. 2). Furthermore, the storage efficiency would be reduced to 69%, with 88% of the storage devoted to atmospheric carbon and the remaining part storing fossil carbon (i.e., from natural gas combustion powering DACCS and power plants with CCS). The global EU geological storage capacity would not be fully depleted 5 , yet competition between fossil and atmospheric captured CO 2 for the sites available at the regional level could become an issue, particularly given the increasing policy support for CCS in fossil-fired power plants in many countries 65 .
Finally, in the "Late" scenario, the maximum gross negative emissions would be substantially reduced to −7.09 Gt CO 2 . The removal potential would be limited by the maximum diffusion rates of BECCS and DACCS, which would even impede reaching CO 2 neutrality in the EU power sector (+1.54 Gt of net CO 2 emissions by 2100) and constrain the use of residues and land to 40% and 20% of their maximum availability from 2020 to 2100, respectively. Most of the biomass resources would be consumed by biomass power without CCS during the inaction periods (only 8% of the total residues and 14% of the land available in the period 2020-2100 would be exploited with BECCS after 2085). Here, DACCS would play a minor role, removing only −0.05 Gt of CO 2 by 2100 due to the slow speed at which it could be scaled up. The storage efficiency would substantially decrease, with 25% of the total capacity available devoted to fossil CO 2 emissions.
Overall, delaying the CDR deployment would lead to the underuse of biomass and land resources, tighter bounds on the BECCS and DACCS facilities, and domestic storage sites depleted with fossil carbon, which altogether would reduce the future ability of individual countries on CDR. However, transboundary agreements enabling import/exports of biomass and CO 2 , new estimates of suitable geological sites, less conservative biomass potentials, and a broader CDR portfolio beyond the energy sector could enhance the EU's ability to deliver net negative emissions 37 . Regional implications for the energy systems The carbon-negative electricity supplied by BECCS and the large electricity and heating requirements of DACCS create strong links between them and the power system. Therefore, their deployment would require longterm planning, ensuring their effective integration into the evolving portfolio of power technologies starting at the earliest. In the "Now" scenario (shown in Fig. 3, "Slow" and "Late" scenarios in Supplementary Figs. 3 and   4, respectively), the total electricity generated in the EU by 2100 would be produced mostly by wind onshore (41%), followed by nuclear (23%), BECCS (8%), hydropower run-of-river and wind offshore (both with 7%), concentrated solar power (6%), hydropower from reservoirs (6%), and marginal contributions from natural gas, solar photovoltaic open ground, biomass w/o CCS, and geothermal (<1%) (Fig. 3a).  Fig. 3b). Both forestry and agricultural residues would be fully exploited in all countries starting BECCS deployment from today. Forestry residues would contribute the most to the CO 2 removal (i.e., 45% of the total gross CO 2 removed by 2100), while miscanthus production would occupy all the marginal land available due to its overall superior carbon sequestration potential, removing −15.87 Gt CO 2 by 2100 (17% of the total gross removed) and becoming the main carbon sink in some countries (i.e., −6.16 Gt CO 2 removed in Spain). Switchgrass would also be cultivated in some regions, and the storage capacity would no longer be the bottleneck if actions were further delayed ("Slow" and "Late" scenarios in Supplementary Figs. 3b and 4b,   respectively). This is due to its relatively higher carbon removal capacity (i.e., CO 2 uptake per kg of pellets), which provides more CDR in the remaining time but at the expense of reducing the electricity delivered owing to its lower energy density.
Regarding DACCS, the configuration relying on electricity and heating would be the only one installed, benefitting from the decarbonized electricity mix. In the "Now" scenario, DACCS would be established in eleven countries, with France, Spain, the United Kingdom, Italy, and Romania providing 97% of the gross removal from DACCS (i.e., −18,72 out of the −21.46 Gt CO 2 by 2100), all of them with enough geological sites for storing the captured CO 2 domestically (Fig. 3b). For example, in France, −6.65 Gt CO 2 by 2100 would be removed via DACCS, taking advantage of its abundant saline aquifers and decarbonized mix dominated by wind onshore and nuclear (Fig. 3a). Similarly, in the United Kingdom, which lacks enough biomass resources to exploit its storage capacity only with BECCS, Delaying actions to 2055 ("Slow" scenario) would imply that less biomass is available, so the BECCS capacity would diminish accordingly, and additionial DACCS facilities would be deployed in Czech Republic, Denmark, and Slovakia to maximize CDR, taking advantage of the geological storage available (Supplementary Fig.3).
Wind capacity would be massively deployed in most countries, becoming the dominant source in France, Spain, Italy, Germany, United Kingdom, Sweden, and Finland ("Now" scenario in Fig. 3a). Notably, offshore wind would become predominant in some countries, with 82% of the EU capacity located in the United The optimal roadmap, assuming full cooperation among EU countries, would entail an intensive trade of biomass, CO 2, and electricity ( Fig. 4 for the "Now" scenario and "Slow" and "Late" scenarios in Supplementary   Figs. 5 and 6). However, to minimize the transport flows, the vast majority of the biomass demand would still be supplied domestically, the captured CO 2 stored locally, and the electricity generated consumed on-site (domestic consumption in Fig. 4 depicted with the chords leaving and entering the same country). Some regions would be net exporters of biomass (e.g., France or Sweden) and some net importers (e.g., Netherland, Germany, or Denmark). The same would apply to CO 2 , with, for example, Sweden, the United Kingdom, and Romania acting as net importers and the Netherlands, Poland, Finland, or Portugal becoming net exporters.
Regarding electricity trade, countries such as France, Spain, and Sweden would emerge as pivotal in the power system, acting as net exporters of electricity to exploit their abundant low-carbon intensity resources (e.g., electricity trades from France to Germany, Italy, Netherlands, Belgium, and the United Kingdom, Fig. 4c).
The largest exchanges of biomass and CO 2 would occur between France-the Netherlands, and the Netherlandsthe United Kingdom, respectively (Fig. 4a, b). Notably, Sweden would export biomass resources to Germany and Denmark to fully exploit its abundant forestry residues (i.e., 1.34 and 1.15 Gt of pellets on a dry basis, respectively, Fig. 4a). Other countries would export CO 2 , e.g., Finland would send 2.21 Gt CO 2 via pipeline to the abundant deep saline aquifers and hydrocarbon fields in Sweden (CO 2 trades in Fig. 4b). Some countries would be almost self-sufficient in terms of biomass resources, like Portugal, which would transport CO 2 to the Spanish geological sites due to its low geological capacity. Overall, the transport of electricity (and CO 2 ) would be prioritized over the transport of biomass due to the larger emissions of the latter considering a given electricity demand (e.g., 0.01 vs. 0.11 kg CO 2 to satisfy one kWh in an importing country, respectively, considering miscanthus as biomass source and a distance of 800 km in both cases). Hence, BECCS plants would be mostly installed near the biomass sources, leading to decentralized supply chains spread across the EU territory. However, some countries would still import biomass pellets to reduce their reliance on energy from outside the country's borders and to support the high penetration of intermittent wind and solar with dispatchable carbon-negative electricity from BECCS (e.g., the Netherlands importing pellets from France or Denmark from Sweden, Fig. 4a).
The uneven distribution of domestic capacities (i.e., biomass resources, storage sites, and renewable resources) would make national and transnational collaboration essential to exploit bio-geophysical endowments 62,66 and remove CO 2 to the extent required. Hence, new agreements and regulatory frameworks will be needed, and shaping them may further delay CDR deployment.

Conclusions
Here we studied the implications of delaying the roll-out of CDR to raise concerns on the need to set effective plans to promote its large-scale deployment at the right time to avoid extra costs and miss climate targets.
To shed light on the economic, environmental, and technical implications of the prolonged delay of CDR actions, we focused on the deployment of BECCS and DACCS in the EU as prominent strategies intrinsically linked to the power system. We found that postponing CDR could substantially increase the total cost of the power system, with each year of inaction translating into 0.12-0.19 trillion EUR2015 of extra cost to meet the -50 Gt of net CO 2 target. Notably, this extra cost, which could be avoided in the EU by starting actions from today, compares to the estimates of the additional annual investments required globally by mid-century in the energy systems to limit global warming to 1.5°C (i.e., between 0.15-1.70 trillion USD2010 2 ). Moreover, delaying CDR actions would critically limit the removal potential (e.g., −73.73 Gt CO 2 starting in 2020 and −35.60 Gt CO 2 after 2055) due to the underuse of biomass residues and marginal land and the maximum diffusion rates of BECCS and DACCS. Hence, postponing CDR deployment beyond mid-century might prevent the EU from delivering a CDR level aligned with its fair responsibility and its expected leading role 37,55 . After 2050, the maximum CDR potential would be reduced to −56.35 Gt CO 2 , representing around the CO 2 emitted by the EU only during the last decade and less than 10% of the global CDR required in pathways showing limited overshoot of the 1.5 °C target 2 . With the EU contributing short and its leadership efforts questioned, there might be a row of dominoes with other countries remaining impasse on CDR, which ultimately may impede a fair and rapid transition for effective climate change mitigation. Furthermore, we found that delaying BECCS and DACCS may result in the inefficient use of the available domestic geological storage. Retrofitting the existing coal and gas-fired power plants with CCS to decarbonize the power sector would limit the storage capacity available, potentially raising competition issues with atmospheric CO 2 sequestration. Although the global geological capacity is deemed sufficient to meet the climate targets, it could become a bottleneck at the regional level due to the asymmetric distribution of storage sites. With policies promoting CCS in fossil fuel power plants, some countries might exhaust the most suitable geological sites with fossil CO 2 , which ultimately will compromise their future CDR commitments. This reinforces the need to phase out coal and natural gas for a full transition towards renewable energy systems.
Nevertheless, these policies could offer a testbed to support CCS projects, including the development of the CO 2 transportation and storage infrastructure needed to support CDR schemes. Without a supportive policy framework, CDR will never take off in the short term despite being imperative to offset hard to avoid energy and non-energy emissions. In this context, governance issues should also be considered to ensure that BECCS and DACCS will remove CO 2 at the scale and pace required. BECCS and DACCS supply chains involving multiple countries and stakeholders will encompass a wide range of activities (e.g., biomass growth, storage, transportation and processing, and CO 2 transportation and storage). In the absence of well-designed policy mechanisms rewarding all CDR actors or direct incentives to engage with BECCS and DACCS, their large-scale deployment is unlikely to occur. Hence, integrating BECCS and DACCS into the evolving power systems must be done at the earliest, ensuring the best use of the existing resources through effective long-term strategic decisions and planning.
CDR is gaining increasing attention in the political agendas of governments, stakeholders, businesses, and industry. The recent EU climate neutrality target by mid-century puts CDR backs on the table of reticent countries 58 . At the regional level, the least-cost distribution of efforts may differ greatly from equitable shares and the willingness to act in different countries 37,68 . In practice, key countries might still be reluctant concerning CDR. These nations might not fully participate and fail to shoulder their fair share of the EU CDR burden, which will result in some countries having to do more and faster and others contributing less and going slowly towards the CDR required. Nevertheless, our findings demonstrate that failure to start CDR at the right time would increase the total costs and make climate targets slipping out of reach.
Overall, our work underscores the importance of taking early actions on CDR. The potential economic and environmental benefits that would be missed if we delayed action may act as an incentive to spur CDR efforts worldwide. We focused on BECCS and DACCS and the EU, yet other technologies should also be considered and evaluated at the regional and global levels to further motivate early CDR deployment. Regardless of the technological and spatial scope of the analysis, the underuse of resources and the diffusion rates will limit our ability to remove atmospheric carbon and increase the associated costs. We hope that this quantitative analysis will help to break the current climate impasse, accelerate the take-off of CDR and get CDR deployed at the right time to avoid missing the climate targets.

The RAPID model
We developed a bottom-up multi-period linear programming model to explore the consequences of delaying CDR actions, referred to as RAPID model (RemovAl oPtImization moDel). RAPID is an energy systems model that identifies the most cost-effective emissions and technology pathways by jointly optimizing the power mix and the deployment of BECCS and DACCS, where CDR is assumed to start from a particular year within the time horizon. RAPID can be solved in two alternative ways, i.e., by minimizing the power system's costs to meet a net CDR target or by maximizing the net negative emissions balance. RAPID includes a set of technical, cost, and emissions-related constraints that the solution sought should satisfy. In each run, the model can retrofit the power mix from 2020. However, it can only deploy BECCS and DACCS from a given year defined beforehand and varied iteratively across runs. RAPID is implemented assuming perfect foresight for the modeling timeframe, i.e., the model optimizes decisions for every period with full visibility of the entire time horizon. For simplicity, the model considers five-year intervals between 2020 to 2100.
RAPID is mathematically formulated in compact form as follows.
Where continuous variables denote technical decisions, e.g., the area of land devoted to a dedicated energy crop, the amount of residues exploited, the technologies' capacities, the transportation flows between

Scenario definition and solution approach
We define three different scenarios, labelled as "Now", "Slow" and "Late", differing in the starting year for the CDR deployment. The "Now" scenario considers immediate CDR action starting in 2020. The "Slow" scenario assumes that CDR action starts in 2055, while the "Late" scenario delays CDR until 2085.
We solve RAPID considering the time horizon 2020-2100, divided into 16 periods t of five years length each, assuming that BECCS and DACCS can only be deployed from a particular period t' onwards. As an example, when RAPID is solved for 2050, we assume that CDR can only be deployed from 2050 to 2100, but not before. balance to quantify precisely the CO 2 removed from the atmosphere, whose storage is ensured in the long term. This adjustment applies as well to the DACCS, which also captures atmospheric CO 2 .
Bearing the above in mind, the biogenic carbon and the CO 2 captured with DACCS were tracked manually to adjust the CO 2 balance. Hence, we first excluded all the biogenic CO 2 from the LCI of all the supply chain activities, thereby considering only the non-biogenic emissions to air (list of elementary flows in Supplementary Information section 2.2.4). Note that omitting the biogenic carbon is a common practice in most LCIA methods focused on climate change under the assumption that the CO 2 uptake by biomass via photosynthesis will be eventually released again into the air. Second, the CO 2 uptake from the atmosphere via photosynthesis or chemical reactions is modeled as a negative flow of CO 2 entering the system. For the biomass resources (i.e., energy crops and residues from agriculture and forestry activities), the CO 2 uptake is estimated from the carbon and water content (Supplementary Table 25). Finally, these negative CO 2 flows are tracked along the supply chains by accounting for the flows leaving the system as positive flows (e.g., biomass losses, uncaptured CO 2, or other leakages), to precisely establish the CO 2 emissions balance. More details on the emissions data and associated references are presented in the Supplementary Information in section 2.2.4.

Technology deployment
Diffusion rate constraints preclude solutions that are not consistent with the deployment rates often found in similar technologies. BECCS is constrained by both the availability of biomass resources and the maximum deployment rate (as well as by the storage capacity). DACCS, however, is only limited by the maximum diffusion rate (and the storage capacity). All power technologies are constrained by the resource availability and diffusion rate, while the model limits, in turn, the penetration of intermittent renewables to ensure the grid's reliability. In essence, the diffusion constraints (Supplementary Equations 24-29) impose limits to the rate at which technologies can scale up under the assumption that external factors constrain the speed of deployment, e.g., market forces, competition issues, infrastructure adaptation, learning rates, or social acceptance 50,63 . The diffusion constraint, therefore, relates the capacity in period t to the capacity already installed in the previous period t-1, i.e., ≤ −1 (1 + ) where is the maximum annual growth rate ( ∈ [0,1]) and is the duration of the period. In RAPID, we assumed an annual growth rate of 20% ( equals 0.2 and equals 5 years) based on the historical diffusion rates of energy-related technologies 63 .
According to the diffusion constraint, technologies with larger installed capacities can diffuse faster. The maximum deployment rates of technologies over time and their initial capacity are shown in Supplementary   Fig. 2.

Uncertainty analysis
We carry out an uncertainty analysis to quantify how uncertainties in the economic and emission parameters affect the outcome of RAPID. In particular, we perform an a posteriori analysis of these uncertainties to establish confidence intervals for the costs and CO 2 emissions in each solution. Hence, besides the base case scenario considering nominal values, we define best-case and worst-case scenarios assuming optimistic and pessimistic values for the parameters, respectively.
Uncertainties in the inventory data are modeled based on Ecoinvent 75 . We used Simapro v9.0 76 to generate 1,000 samples from the probability functions of the uncertain emissions via Monte Carlo sampling; from these samples, we defined the optimistic and pessimistic scenarios, which consider the mean value of the emissions parameters minus and plus two times the standard deviation, respectively (Supplementary Table 27). For the cost uncertainty, we defined the optimistic and pessimistic scenarios considering low and high estimates, respectively, for the CAPEX and OPEX parameters sourced from the literature (Supplementary Tables 10 and   11). The uncertainty analysis results for the costs are provided in Supplementary Tables 43-50, while for the emissions, the results are depicted in Fig. 1b with a shaded area covering the best-and worst-case scenarios.

Code availability
The computer code supporting our analysis is available from the corresponding author upon request.