Scenario set up
We focus first on the case of a stringent climate policy, with a global target on GHG concentrations set at 490 ppm-eq by the end of the century. This is a rather ambitious objective which is roughly in line with maintaining global temperature increase below 2C above the pre-industrial levels with some even chances. We assume an idealized policy setting in which global cooperation is in place starting from 2015 onwards, mitigation is allocated efficiently across countries by means of a frictionless global carbon permits trading schemeFootnote 4, and emissions can be borrowed and banked freely thus ensuring perfect temporal flexibility. In Section 4 we depart from this assumption and look into a more fragmented policy architecture. We compare a base case in which DAC is assumed to be not available, to a DAC case in which DAC is modelled as described in Section 2.
The global role of DAC
We begin by assessing the potential deployment of DAC for the climate stabilization target considered. The global amount of DAC in Fig. 1 ranges both ‘realistic’ and ‘optimistic’ cases. Figure 1 indicates that, for either case, DAC is not a viable strategy until the second half of the century. DAC is deployed between 2065 and 2070 but quickly develops into a massive programme, capturing as much as 37 GtCO2/year in 2100Footnote 5. The ‘optimistic’ cost estimation implies an earlier as well as higher level deployment, although the difference is not big. From now on, we stick to the ‘realistic’ case for further analysis. We have also tested with more lenient climate objectives, and we found that for climate stabilization target equal or larger than 550 ppm-eq DAC is never deployed over the century. Therefore, DAC is a mitigation strategy only in the case of ambitious climate policies, and only late in the century after other main mitigation options have been put in place and when the cost of removing the final 17 % of global carbon emission is comparatively expensiveFootnote 6.
To meet the 490 ppm target the scale of deployment of DAC, also as a consequence of the assumed temporal flexibility, is very significant, with about 482 GtCO2 captured via DAC cumulatively over the century. As a result, DAC has a significant impact on the climate mitigation strategy. The possibility of large negative emission substantially affects the optimal mitigation pathway, as shown in Fig. 2a: mitigation is reduced for several decades with respect to the base case, and this additional carbon budget is compensated late in the century by achieving globally net negative emissions. This result is consistent with the analysis of IAMs which show a considerable impact of negative emission technologies on short and mid-term emissions pathway (Clarke et al. 2009). Figure 2a also shows how much mitigation the conventional abatement technologies could do, under the DAC emission pathway (the dashed line). The distance from the base case emission projection indicates 7.2 GtCO2/year less of mitigation by conventional abatement with respect to DAC. Equivalently, following the optimal abatement pathway assuming DAC will be available in the future, only to find out that this will not be the case, would result in an increase of carbon concentrations at the end of the century of 540 ppm.
The availability of DAC brings down the marginal abatement costs and the total abatement costs (measured by the percentage change of global GDP with respect to the BAU projection), as Fig. 2b and c show. In particular, DAC reduces the total abatement cost in the first decades due to more lenient mitigation effort, and from 2065 it results from the deployment of DAC. The difference of policy cost reaches its peak around 2080, when DAC is available and the same mitigation is foreseen in the two scenarios. This distance shrinks towards the end of the century when (much) more mitigation is undertaken in the DAC case with respect to the base case.
We conduct series of sensitivity analysis on the deployment of DAC related to model parameters, available competing technologies and certain climatologic consideration.
First, we check the sensitivity of DAC to other competing abatement technologies. These are conventional options, which are featured in the current version of WITCH but with certain limits. They are: (1) a more efficient capture rate of CCSFootnote 7 (2) carbon capture of emissions from coal use not only in power plants but also in the industrial sector, at a cost of $81.8/tCO2 (3) a more progressive decarbonisation of mobile emission sources via larger availability of next-generation bio-fuelsFootnote 8 (4) extra abatement of the residual non-CO2 GHG emissionFootnote 9.
We include these technologies first one at a time and then altogether on top of the standard modelFootnote 10, see the first part of Table 2. According to this sensitivity analysis, the role of DAC is significantly reduced if these extra technologies are put in place. The peak deployment does not change much, but the average rate of deployment decreases by 38 % from 16 to 10 GtCO2 per year. When these options are all available, the average deployment further shrinks by 81 % to 3 GtCO2 per year.
The second sensitivity analysis is on the penetration rate of DAC. When changing the penetration rate from 50 % to 80 %, we do not observe significant differences among the results in terms of the scale of deployment. The variation of the average number is about 2 GtCO2.
Our third sensitivity analysis relates to the inter-generational discount rate. As WITCH assumes perfect foresight, policy makers look forward to negative emissions (Fig. 2b), and the result is more emissions in the near term and fewer in the long run. The preference for the present versus the future is captured in the model through the social rate of time preference (SRTP). By default, WITCH sets SRTP to 3 % initially (in 2005) and it decreases by 0.25 % at each 5-year time stepFootnote 11. Intuitively if the SRTP is set to zero, which means that the future welfare will be equally important as it is now, the emission would be more equally distributed across the time-span. The comparison in Table 2 suggests that when time preference is zero and no discount is applied to future mitigations costs vis à vis with current ones, we can observe significantly lower deployment of DAC.
Finally, since DAC changes the CO2 concentrations in the atmosphere quite rapidly, it might change the global carbon balance. Climatologists look into the carbon cycle to consider the oceanic feedback as a response to large-scale carbon dioxide removal (CDR) from the atmosphere. Given the carbon exchange between the atmosphere and the ocean, the sudden removal of carbon from the atmosphere would cause the outgassing from the ocean to restore the natural carbon balance (Gruber et al. 2009; Vichi et al. this volume). Our DAC scenario simulates a continual CDR action during the last 30 years of the century, resulting in 61 ppm negative emission. According to Vichi et al. (this volume), who quantify the outgassing effect caused by CDR actions, 65 ppm constant CDR within 30 years induces 8.5 ppm (18 GtC) of outgassing. If this effect is taken into consideration, we expect that about 15 % of the carbon removed by DAC would go back to the atmosphere. Taking the outgassing effect into consideration, our simulation result shows less deployment of DAC, by roughly 30 % (141 GtCO2) less in total over the 30 years.
A regional assessment of DAC
One of the most appealing features of DAC is its decoupling from the specific emission sources: it can be implemented where it is more convenient to do so, accounting for the cost of carbon storage and the energy requirements. As already discussed, in WITCH the cost of carbon storage is differentiated among the regions according to the availability of storage reservoirs. In principle, DAC could allow for more underground storage of carbon, while at the same time keeping storage and energy costs in check by choosing the most appropriate sites. This can be illustrated by Fig. 3a: compared to the base case where 815 GtCO2 are stored cumulatively to 2100, DAC allows the storage of an additional 788 GtCO2 at a similar storage cost (around $100/tCO2 captured).
However, the CO2 storage flexibility provided by DAC could also be achieved by transporting CO2 (from any CCS facility) across macro-regions. In the standard version of the model this feature is not allowed and the CO2 captured is assumed to be stored regionally, an assumption justified by the coarse geographical detail of the model. In order to address this issue, we have run an additional model experiment, assuming that (1) DAC is not available (2) CO2 storage sites are perfectly fungible, thus aggregating the regional cost curves into a single, global one. We find that 1240 GtCO2 is stored over the century, which can be compared with 815 GtCO2 when no storage flexibility is allowed. This storage value is slightly lower than the case where DAC is allowed but the storage is regional (1603 GtCO2). The total abatement cost would also be lowered in the case with storage flexibility compared to the base case, but storage flexibility without DAC results in higher abatement cost than regional storage with DAC after the mid-century. This analysis suggests that DAC would be economically attractive even when compared to a situation in which CO2 can be freely transported and stored in any region where it is economical to do so.
What is the relationship among DAC and other types of CCS? DAC may crowd out others given the limited CO2 storage space, though the flexibility of DAC location might alleviate this substitution effect. On the other hand, as shown in Fig. 2a DAC allows more emissions headroom till 2080. Therefore, alternative CCS options which are not totally carbon free, i.e. coal with CCS, could benefit from the resulting lower carbon prices vis-à-vis with virtually zero-carbon technology like nuclear power or negative-emission technology like biomass CCS. The overall effect is shown in Fig. 3b, where we plot the regional differences of the cumulative sequestrated carbon for the DAC case with respect to the base case. In most regions DAC is shown to crowd out some biomass with CCS but to induce some additional coal with CCS. The overall crowd-out effect is around 111 GtCO2 over the century, suggesting that DAC is mostly additional to the other types of CCS.
Figure 3b also provides indication about the regional distribution of DAC, with Transition Economies (TE), and Middle East and North Africa (MENA) being the two biggest DAC players. These energy exporting countries (EEX) have a comparative advantage in carrying out DAC because of the large CO2 storage availability and abundant energy resources that can be used for power and high-temperature heat at the DAC facilities, the cost of which accounts for around 30 % of the total cost of DAC in 2100 (see Fig.A1 in appendix for a breakdown of DAC costs). Compared with the base case, in 2100 an additional 65 EJ of power and 298 EJ of high-temperature heat will be needed to fuel the DAC plants, resulting in an increase of 84 % in primary energy supply. For DAC plants only, apart from the increased demand of gas, which provides all the heating, the additional electrical demand is mainly met by nuclear (36 % in 2100) and renewables (wind and solar, 57 % in 2100).
These regional impacts are important when evaluating the economic impacts of DAC. Figure 4, which compares the net present value of the climate policy cost with and without DAC, shows that DAC lowers the global costs of reaching the climate target since it provides social planners with an additional mitigation lever. This economic benefit is mostly captured by EEX, as these are the ones where more DAC is implemented in the first place. This cost reduction is particularly relevant since EEX is the region with the highest policy costs in the model, mostly as a result of having an energy-intensive economy which relies heavily on international energy sales.
In other words, EEX is the region able to benefit the most from this technology. The source of benefit for EEX is two-fold. First, DAC allows the preservation of the value of oil reserves, as noted by Nemet and Brandt (2012). The size of the oil market in 2100 (measured by the market value) is 5-time larger in the DAC case than the base case, and the overall effect is that the market value of international oil trading does not fall as dramatically as in the base case, since more oil is used and traded and its price is higher (See Fig.A2-a in appendix). Second, EEX countries also gain from the carbon market. As they implement DAC, they are able to balance their carbon account by importing fewer permits. In fact they turn from buyers to sellers of carbon permits in the international markets (See Fig.A2-b for the breakdown of the benefit).