Marginal cost curves of leakage risk
Figure 2a plots the marginal leakage cost for each incremental tonne of CO2 injected, referred to as the MLR curves. The MLR increases because less ideal sites are used. The convex shape of the curves suggests that most of the basin can be used for GCS at a low cost of leakage risk, and only a small portion of the injection sites would have very large leakage costs. For the two largest modeled leakage pathway permeabilities (10−10 and 10−12 m2), 75 and 96% of the basin has MLR below $5/tCO2, respectively (Fig. 2b).
Because the MLR curves for leakage pathway permeabilities of 10−12 to 10−16 m2 are essentially the same (differ by < $0.5/tCO2, Fig. 2c), we only implemented in GCAM the MLR curves for two leakage cases: “high” and “extremely high,” which correspond to leakage pathway permeability of 10−12 and 10−10 m2, respectively. These cases are worst-case scenarios because (a) the permeability values are very high compared to reported field measurements and (b) these values are assigned to all of the 45,000 leakage pathways within the basin (Bielicki et al. 2015). The results are compared with those for a “no leakage” case in which all leakage pathways are assumed impervious.
The majority of the leakage costs result from leakage that does not reach groundwater or the land surface (Fig. 2d). On average, 74% of the MLR in the extremely high leakage case and 90% in the high leakage case are attributed to cost that arises from the benign leakage that undergoes secondary trapping. The remaining costs are incurred when leakage interferes with other subsurface activities—mostly with oil and gas production; the costs due to groundwater contamination and surface seepage are negligible. Even in the extremely high leakage case, only 0.005% of the total amount of injected CO2 migrates into the groundwater aquifer after 50 years of simulated injection over all of the 42 injection sites.
Impact on CCS deployment and the global energy system
The GCS costs in GCAM, and the increased costs due to leakage risk, do not vary across the fuels and sectors that deploy CCS (see Supplementary Material). Here, we focus on the electricity sector, which has more CO2 stored than others. We quantify the projected changes in CCS deployment in terms of the amount of electricity that is generated by CCS technologies: coal, natural gas, oil, and biomass facilities with CCS.
Figure 3 shows the projected mix of the electricity generation technologies for the “no,” “high,” and “extremely high” leakage cases with Ctax2. Leakage risk reduces the deployment of CCS technologies because of the increase in electricity generation costs. In the “no leakage” case, the 5482 EJ of cumulative electricity generation by CCS technologies between 2020 and 2100 is equivalent to 67 years of the ~82 EJ global electricity generation in 2012 (IEA 2014). Relative to the no leakage case, there is a decrease of 143 EJ (3%) and 658 EJ (12%) in the high and extremely high leakage cases, respectively. The results for the Ctax1 and the two RCPs scenarios show similar reductions (see Supplementary Material): The cumulative electricity generation from CCS technologies decreases 10–14% in the extremely high leakage case and 2–4% in the high leakage case. The reduction is smaller with a lower CO2 emissions mitigation target (e.g., lower CO2 tax, RCP4.5), but the percentage changes are higher because of less reliance on electricity as a final energy carrier.
The decrease in the electricity generated by CCS technologies does not produce an equivalent decrease in the total electricity generation, partly because of the increase in the use of fossil fuels and biomass without CCS. In the extremely high leakage case with Ctax2, the cumulative electricity production using fossil fuels and biomass decreases by only 106 EJ. There are smaller increases in the use of fossil fuels and biomass without CCS under the more stringent climate goal (e.g., RCP2.6).
The shift to energy technologies that do not emit CO2, mostly nuclear, further reduces the impact of leakage risk on electricity generation. For all of the leakage cases and mitigation scenarios that we investigated, the reduction in cumulative electricity generation is within 1% of the no leakage case.
Overall, total primary energy production is essentially unaffected across the scenarios (<1% difference, see Supplementary Material). Accordingly, the energy end-use sectors are not noticeably affected by leakage risk. For example, the use of different types of fuels (e.g., delivered gas, electricity, refined liquids) in the transportation sector for the leakage cases changes <0.1% from the no leakage case under all CO2 emissions mitigation scenarios (see Supplementary Material).
Impact on atmospheric CO2 concentrations
The indirect effects of leakage risk have two opposing effects on CO2 emissions and thus the atmospheric CO2 concentration under carbon taxes: (1) The amount of CO2
produced decreases because of shifts away from fossil fuel and biomass in primary energy production, and (2) the amount of CO2
stored decreases because of the reduced CCS deployment (see Supplementary Material). Everything else held equal, the first effect decreases CO2 emissions, whereas the second effect increases CO2 emissions. Figure 4 shows that the second effect is larger. The atmospheric CO2 concentration in 2100 for the extremely high leakage case is approximately 4 and 5 ppm higher than the no leakage case with Ctax2 and Ctax1, respectively. For the high leakage case, the corresponding differences are about 1 and 1.5 ppm for the respective taxes. These differences increase faster with Ctax2 in both leakage cases because CO2 storage is consistently higher, and riskier and costlier storage reservoirs are used sooner. However, this difference plateaus around 2085, while the differences under Ctax1 continue to increase because CO2 emissions mitigation is more elastic under the less stringent climate mitigation targets.
For the atmospheric CO2 concentration, the indirect effects of the increased cost of GCS with leakage are greater than the direct effects of CO2 seepage. The surface seepage of CO2 that would increase the atmospheric CO2 concentration by 1 ppm (~2.2 GtC) amounts to 0.7 and 1.2% of the CO2 that is stored under Ctax2 (1104 GtCO2, no leakage) and Ctax1 (649 GtCO2, no leakage), respectively. These values are orders of magnitude greater than the seepage of CO2 to the atmosphere for the worst-case scenarios considered here. Even if secondary trapping is ineffective and all of the leaked CO2 eventually reaches the surface, which is ~0.003 and ~0.2% for the high and extremely high leakage cases, respectively, the resulting increase in the atmospheric CO2 concentration would be much less than 1 ppm.
For the RCP scenarios, the goal is to achieve a desired trajectory of radiative forcing, and thus, the atmospheric CO2 concentration is not affected by the leakage risk. Instead, CO2 prices for the leakage cases increase in addition to some shifts in the energy system (see the Supplementary Material).
Uncertainties and implications
Our primary finding is that leakage risk is likely to have a negligible effect on CCS deployment in the global energy system and the effectiveness of climate change mitigation policies—especially for more stringent climate mitigation targets. Here, we discuss this finding in the context of uncertainties generated by the model framework’s assumptions, approaches, and abstractions.
The GCAM projections of the effects of leakage are expected to be greater than from other IAMs because inter-model comparison studies have shown that GCAM tends to deploy CCS more than other IAMs (van der Zwaan et al. 2013; Wilkerson et al. 2015). The increase in CO2 storage costs from the worst-case leakage cases considered here leads to at most a small reduction in the deployment of CCS, which is negligible compared with the inter-model discrepancies. Further, the projected atmospheric CO2 concentrations under the two CO2 tax scenarios are higher than those for the no leakage case by only a few ppm. Such differences are within the uncertainties in climate modeling (IPCC 2014b) and are negligible when compared to the increase in atmospheric CO2 concentration that would result from not using CCS (Fig. 4a). In addition, since the impact of the worst-case scenario assumption on the overall leakage costs is far greater than that may arise from other geophysical modeling options (e.g., the choice of target formation, see Supplementary Materials), we feel confident that the primary finding of our study would hold for a wide range of sedimentary basin conditions.
This is not to say that leakage risk is not important in CCS deployment and operational decision-making. It will always be important to carefully site GCS reservoirs, monitor CO2 movement in the subsurface, and track the potential for environmental damage (Herzog 2011, Pawar et al. 2015). Furthermore, the costs of leakage may still be substantial for some stakeholders at individual storage reservoirs (Bielicki et al. 2016).
Our second major finding is that the indirect effects of leakage risk are greater than the direct effects of surface seepage on CO2 emissions mitigation. As such, it is important to ensure a twofold assessment, i.e., for potential GCS reservoirs to be assessed on the probability for leaked CO2 to reach the atmosphere and on the consequences of leakage that remains in the subsurface. The twofold assessment can guide GCS siting to minimize leakage costs from interferences with subsurface activities. The benign CO2 leakage that is contained in the subsurface by secondary trapping may account for the majority of the expected cost of leakage risk. If secondary trapping was to be considered a reliable backup trapping mechanism, then relaxing the requirements to address leakage could further reduce costs and the effect of leakage on CCS deployment and on the atmospheric CO2 concentration could be even smaller than what was predicted by our GCAM simulations.