For the purpose of answering these questions we constructed an optimal growth model in which prices in the global economy yield general equilibrium between supply and demand of a growing but stabilizing world population (UN 2009). A representative planner optimizes welfare, which is expressed by the discounted sum of utility per time period. Welfare optimization leads to the Ramsey rule for the intertemporal rate of exchange. The OCEAN model thus fits in the tradition of top-down integrated assessment models, as initiated by Nordhaus (1994) with the DICE model. Like DICE, our model accounts for damages incurred as a result of climate change and proffers abatement options with which, at a certain cost, emissions of greenhouse gases can be avoided (see the appendix in the online Support Information for the full details of our model). As in Gerlagh and van der Zwaan (2002); Hoel and Sterner (2007), and Sterner and Persson (2008), OCEAN also includes ‘intangible’ damages (i.e. costs that are hard to quantify or express in monetary terms) associated with climate change (see also Gerlagh 2014).
Our new model has certain similarities with the DEMETER model, which we previously used for related analysis and has been instrumental for studying several climate-related policy questions (see e.g. van der Zwaan et al. 2002, and Gerlagh and van der Zwaan 2006). Our new model, like DEMETER, simulates the global use of fossil fuels, the CO2 emissions resulting from their combustion, as well as the means that allow for avoiding these emissions. It includes generic production and consumption behavior and a basic climate module, and is in several ways an extension of DEMETER, since we have refined the simulation of atmospheric and oceanic CO2 stock build-up as well as the corresponding climate change dynamics, and now explicitly include impacts such as ocean acidification. Over the past years DEMETER has been used specifically for the purpose of studying economic opportunities and conditions for CCS deployment, including issues such as potential CO2 leakage (van der Zwaan and Gerlagh 2009; Gerlagh and van der Zwaan 2012). We here build on that work, as well as the broader and still growing literature on this topic (see, for example, Ha-Duong and Keith 2003; Keller et al. 2008; Keppo and van der Zwaan 2012; Teng and Tondeur 2007).
Unlike with DEMETER, we have now substantially simplified the simulation of energy-transition dynamics, as we do not explicitly represent in our new model the replacement of fossil fuels by mitigation options such as renewables, nor do we assume that the costs of mitigation reduce according to learning-by-doing phenomena. Rather, we keep the energy economy relatively simple, by allowing for a shift away from fossil fuels in a more stylistic way. The social optimizer can choose between either continuing with the use of fossil fuels but experiencing the damages emanating from climate change, or shifting away from them to avoid these impacts but at a given abatement cost. An additional degree of freedom for the social planner is that (like in DEMETER) CCS can be introduced as a means to decarbonize fossil fuels, which would permit their continued usage but under extra abatement costs.
New in our current approach is that we represent two types of CCS options, one for which storage of CO2 takes place onshore, the other one offshore. We make several key assumptions for these options, in an attempt to let our model reflect stylistically some of the main features of mitigation through CCS. First, both these types of CCS are characterized by an energy penalty of 30 % (defined as the share of the electricity output of a power plant that needs to be employed to operate the CCS facility), which yields lower levels of CO2 avoided versus the levels of CO2 stored (Gerlagh and van der Zwaan 2006). Second and critically important in a model such as OCEAN, based on publications such as by the IPCC (2005; 2014); Keppo and van der Zwaan (2012) and Finkenrath (2012), we assume that the costs of onshore CCS amount to 50 €/tCO2 (median) with a lognormal uncertainty range of 17–150 €/tCO2 (2.5 % and 97.5 % probability levels). This uncertainty range is especially broad since large-scale deployment for CCS is still practically untested. Similarly, we assume for offshore CCS costs of 75 €/tCO2 (median) with a lognormal uncertainty range of 25–225 €/tCO2 (2.5 % and 97.5 % probability levels).Footnote 1
,
Footnote 2 The difference between the costs of these two types of CCS reflects the fact that offshore activity is more expensive than onshore activity, and that for offshore CCS additional and more advanced technology is required to perform geological injection of CO2 through a layer of sea or ocean water, in comparison to the equipment needed for onshore CCS (even while for the former existing infrastructure such as oil or gas rigs can sometimes be used – see e.g. the Sleipner project (Statoil 2013)). Third, we differentiate between safe and unsafe storage sites, but assume that the type (i.e. safety level) of any particular storage site is unknown in advance. We suppose that the majority of CO2 stored geologically will stay underground forever, which is a necessary assumption if one wants CCS to become deployable on a large scale: we stipulate that the share of storage sites with a perfectly immobilized total quantity of CO2 amounts to 70 % (median) with an uncertainty range of 40–85 % for all injected CO2.
Fourth, we stipulate that for unsafe storage sites, CO2 stored onshore leaks with a rate of 0.33 %/yr. (median) with a lognormal uncertainty range of 0.1–1.0 %/yr.Footnote 3 The reason for this is not because we think that CO2 cannot be stored safely underground: quite on the contrary, we think that sites can be found for which storage will be permanent or close thereto, if they are well managed and monitored (Kharaka et al. 2006; Klusman 2003; Shaffer 2010). But, as explained above, we leave open the possibility that as a result of economic, institutional or political reasons, and/or given the prevailing safety culture, storage activities in some regions or countries may not necessarily be subjected to the measures required for long-term secure containment. Likewise, we suppose that from unsafe offshore storage sites CO2 leaks with a rate of 0.5 %/yr. (median) with a lognormal uncertainty range of 0.1–2.4 %/yr. As described in the introduction, such leakage levels are speculative and little evidence exists to draw upon, but they are deemed plausible by the CCS research community and match the possible figures 100–1000 tCO2/day estimated in several publications (ECO2
2015; IEAGHG 2009; Statoil 2013). On the basis of the fact that injecting is technologically more challenging, and remediating less trivial, for offshore than onshore CO2 storage activity,Footnote 4 we justify our choice for higher leakage rates for the former than for the latter.Footnote 5
Fifth, we assume that CO2 (emitted or leaked) incurs two types of damages to the environment (atmosphere or ocean) into which it is released: tangible and intangible (see e.g. the seminal work by Nordhaus (1994) for a more detailed description of this distinction). We assume that a 3 °C temperature increase leads to a tangible damage cost (emanating from GDP or consumption losses as a result of e.g. sea level rise, extreme weather events or drops in fisheries and tourism) of 2 % (median) of gross domestic product (GDP) with a lognormal uncertainty range of 0.5–8 %. The intangible damages from this temperature rise (such as in terms of reduced human well-being or satisfaction resulting from environmental degradation or biodiversity loss) add another 1 % cost of GDP (median), with a higher lognormal uncertainty range of 0.125–8.0 % to reflect the higher uncertainty for these intangibles.Footnote 6 The tangible costs resulting from an increase in ocean acidification corresponding to a decrease in the pH parameter equivalent to 550 ppmv atmospheric CO2 concentration increase (including for instance the negative effects of ocean acidification on fisheries) amount to 0.1 % of GDP. Given the large uncertainties in this respect, we assume a wide probability distribution of 0.01–1.0 % for these costs. Similarly, intangible costs resulting from a corresponding drop in the pH-value (reflecting e.g. the loss of marine biodiversity) amount to 1.0 % of GDP, with a lognormal uncertainty range of 0.2–5.0 %. Intangible damages are scaled for the base year (2020) and their value increases with income. The pure rate of time preference is set to 1 %/yr. (range 0.5–2 %/yr). This means that the OCEAN model is calibrated so as to be consistent with a robust decline in returns on capital when economic growth slows down, as argued in Piketty and Zucman (2014). The assumed declining returns imply a robust carbon price, robust optimal emission abatement levels, and moderate climate change as the median outcome.
Of course, a model like ours requires many more assumptions, regarding for example business-as-usual GDP and income growth, baseline fossil fuel use and associated CO2 emissions (energy and non-energy related), the radiative forcing contributions from other greenhouse gases, return on capital investments and the intertemporal elasticity of marginal utility, as well as the costs of generic abatement activity. For the details behind the parameter values we assumed for these variables we refer to the appendix in the online Support Information for this article. We ran 100 scenarios with parameter values that span the indicated uncertainty ranges, and thereby perform a Monte Carlo type of analysis. We consider this in our context a particularly appropriate approach, given the uncertainties involved in many parameter values, but we are appreciative of the other possible frameworks that could be used to explore this topic, among which for example Expected Utility or Decision under Uncertainty analysis (see e.g. Held et al. 2009). We next report our results when running our model in a cost-benefit mode for all these scenarios, for a set of generic emissions-, climate- and ocean-related parameters, and for the listed variable values for the use of CCS and potential leakage of CO2.