The cost of stratospheric climate engineering revisited

Abstract

Stratospheric aerosol injection (SAI) has been receiving increasing attention as a possible option for climate engineering. Its direct cost is perceived to be low, which has implications for international governance of this emerging technology. Here, we critically synthesize previous estimates of the underlying parameters and examine the total costs of SAI. It is evident that there have been inconsistencies in some assumptions and the application of overly optimistic parameter values in previous studies, which have led to an overall underestimation of the cost of aircraft-based SAI with sulfate aerosols. The annual cost of SAI to achieve cooling of 2 W/m² could reach US$10 billion with newly designed aircraft, which contrasts with the oft-quoted estimate of “a few billion dollars.” If existing aircraft were used, the cost would be expected to increase further. An SAI operation would be a large-scale engineering undertaking, possibly requiring a fleet of approximately 1,000 aircraft, because of the required high altitude of the injection. Therefore, because of its significance, a more thorough investigation of the engineering aspects of SAI and the associated uncertainties is warranted.

Appendices

Appendix 1: References to the SAI cost of “a few billion dollars”

Here, we document those references made to “a few billion dollars” as being the cost required to geoengineer the climate using SAI. We acknowledge that the range of costs found in the literature is large. For example, the estimate of ~US$0.2 billion/year/(W/m²) by The Royal Society (2009) is one of the cheapest, while the estimates of ~US$50 billion by Crutzen (2006) is at the upper end of the scale. The United States Government Accountability Office (2011) and Rickels et al. (2011) also reported high estimates, although these were not original analyses. Some sources have referred to costs at the higher end of the scale (listed below); however, estimates at the lower end of the spectrum have drawn a disproportionately large response from various circles.

We searched for references to SAI costs in online sources, books, and opinion pieces and we identified sentences mentioning the cost of geoengineering. No formal systematic screening was performed, although efforts have been made to look at important sources.

For newspapers, a query of geoengineering AND (price OR cost OR million OR billion) with the LexisNexis Academic database identified a number of articles (inquiry conducted on May 13, 2014). Note that the query was constructed with the operand OR, and that the addition of the qualifiers million and billion did not narrow the range of identified newspaper articles. For the top 10 outlets listed in the results of this query, we document all articles mentioning the cost of SAI. As described below, many sources refer to low-cost estimates but some do cite high costs. For brevity, we only provide a sentence from each source referring to the SAI cost. Readers are encouraged to consult the original material to understand the context fully.

1.1 Online sources

Collins (2014): “It’s been said that it would be so cheap—in the order of a few billion dollars per year—that most countries could deploy this if they wanted to.”

Robock (2014): “I have calculated that it would only cost a few billion dollars per year to lift enough sulfur into the stratosphere with airplanes to create a cloud to reflect enough sunlight to counteract the global warming we will experience in the next several decades, but with a big ‘if’.” [The “if” is a question on the size of aerosol particles.]

Blackstock et al. (2013): “For just a few billion dollars a year, it appears possible for a fleet of a couple dozen aircraft to spray enough sulfur particles into the stratosphere to lower temperatures by a couple degrees Celsius.”

Gingrich (2008): “Instead of imposing an estimated $1 trillion cost on the economy by Boxer–Warner–Lieberman, geoengineering holds forth the promise of addressing global warming concerns for just a few billion dollars a year.” [This statement by a former Republican congressman of the United States was originally posted on his blog and reproduced in full in a book chapter by Michaelson (2013).]

Morton (2008): “That’s a purchase cost of about $6 billion, and an ops cost more like a billion a year.”

1.2 Book

Keith (2013): “Taking the estimate of one dollar per kilogram delivered to 75,000 ft and assuming one million tons of material per year, the total cost of large-scale geoengineering would be about one billion dollars a year.” [Unlike many other experts, Keith explicitly talks about a new type of aircraft in the same book and thus, his comment should be carefully interpreted, although those quoting him may not realize such nuances.]

1.3 Opinion article

Keith et al. (2010): “At about US$1,000 a tonne for aerosol delivery, [the solar radiation management cost] adds up to just billions of dollars per year.”

1.4 Newspapers

“Annually, it could cost somewhat less than $8 billion—about the price of a major oil pipeline.” (“Arctic thaw could be reversed. Theoretically feasible, but other issues arise” by Bob Weber, The Calgary Herald, December 26, 2012).

“According to Justin McClellan of Aurora Flight Sciences in Cambridge, Mass., whose team evaluated several ways to deliver the sulfates, this would cost about $10 billion per year.” (“Can we engineer a fix to our climate problems?” by Stephen Battersby, The New York Times, November 6, 2012, Health Section, p. E01).

“For well under a billion dollars, a ‘coalition of the willing,’ a single country, or even a wealthy individual could decide to take the climate into its own hands.” (“Geoengineering: Testing the Waters” by Naomi Klein, The New York Times, October 28, 2012).

“Especially the Pinatubo Option: We could scatter particles into the stratosphere with a fleet of high-altitude planes, for the (relatively) low price of a few billion dollars.” (“A high-tech tool kit to fight global warming” by Bill Gifford, June 13, 2010, The Washington Post).

“Keeping the planet cooled steadily (at least until carbon emissions declined) might cost $30 billion per year if the particles were fired from military artillery, or $8 billion annually if delivered by aircraft, according to the Novim report.” (“The Earth Is Warming? Adjust the Thermostat” by John Tierney, The New York Times, August 11, 2009, Section D, p. 1).

“[Cost of stratospheric aerosols] [e]stimated at $25-billion to $50-billion per year.” (“Ideas so crazy, they just might work (temporarily)” by Ivan Semeniuk, The Globe and Mail, December 5, 2009).

“The sulfur would have to be carried aloft by rockets or balloons—and every year or two, at a cost of untold billions of dollars.” (“Invent,” The New York Times, April 20, 2008).

“In a draft of his paper, Dr. Crutzen estimates the annual cost of his sulfur proposal at up to $50 billion, or about 5 % of the world’s annual military spending.” (“How to Cool a Planet (Maybe)” by William J. Broad, The New York Times, June 27, 2006, Section F, p. 1)

Appendix 2: Calculation of cooling efficiency from GeoMIP studies (Kravitz et al. 2011a, 2013)

GeoMIP is an abbreviation for Geoengineering Model Intercomparison Project, which is an international project with a standardized set of experiment protocols designed to analyze the climatic outcomes of geoengineering (Kravitz et al. 2011a, 2011b). The relevant scenarios are G3 and G4. Both of these scenarios envision situations where forcings according to the Representative Concentration Pathway (RCP)4.5 are countered with some form of SAI. In G4, SAI starts in 2020 and it is assumed to amount to a quarter of the material that resulted from the Mount Pinatubo eruption in 1991, injected at a rate of 5 Mt-SO₂/year over an area of 16–25 km near the equator. In G3, SAI also begins in 2020 but it increases with time such that the surface temperature remains approximately constant.

Dr. Kravitz kindly computed the global averages of top of atmosphere (TOA) net total flux for RCP4.5 and G4 for the following models: BNU-ESM, CanESM2, GISS-E2-R, HadGEM2-ES, MIROC-ESM, and MIROC-ESM-CHEM (see (Kravitz et al. 2013) for general model descriptions). The first four models all incorporate bulk aerosol treatments, whereas the remaining two have prescribed aerosol optical depths for stratospheric sulfate aerosols. In the G4 simulations, the injection quantity of geoengineering sulfate aerosols was constant, and the TOA net flux difference between the G4 and RCP4.5 cases reduced with time.

We computed the cooling efficiency for the G4 scenario using a method similar to the diagnosis of effective radiative forcing by Gregory et al. (2004). In their method, the intercept of a temperature-flux line was taken to represent the effective radiative forcing (see their Fig. 1). In their paper, the data points were plotted in the first quadrant, whereas in our case, they plotted in the third quadrant because we considered cooling rather than warming.

We took RCP4.5 as a reference and computed the difference between G4 and RCP4.5 in the net TOA flux change. We compared that difference against the difference in surface air temperature and we regressed the flux change onto the temperature change. Unlike the usual method, which assumes a quadrupling of CO₂, we were dealing with a very small change and thus, we took a decadal average.

Figure 7 shows the forcing changes versus temperature differences, averaged over decadal periods. The results of the two versions of MIROC do not show a simple dependence of temperature change on forcing change and thus, they were excluded in further analysis. Upon performing the regression, we obtained cooling efficiencies of 0.190–0.278.

Fig. 7

Gregory-like diagnosis of effective radiative forcing for Geoengineering Model Intercomparison Project (GeoMIP) G4 scenario. The horizontal axis represents the decadal average of the surface air temperature difference, whereas the vertical axis denotes the net top-of-the-atmosphere flux change

In Fig. 2, we chose 2013 as the year of GeoMIP because the overview paper reporting the initial results was published in that year and because most of the papers on G4 were submitted in that year.

Appendix 3: Regression analysis for lifting cost

To understand the general pattern, we performed a regression analysis in which we regressed the lifting cost onto altitude. We conducted sensitivity analyses by considering the effects of outliers, which we took to be the three data points with the maximum cost, minimum cost, and altitude of >40 km. We also examined the effect of including the estimates from Davidson et al. (2012) because their dataset had no variation in the independent variable because of their focus on the single altitude of 20 km. We also looked at two choices for the dependent variable: the logarithm of altitude and altitude itself. The results are shown in Figs. 8 and 9. The lower-left panel of Fig. 8 describes the regression lines in Fig. 3. The correlation coefficients are generally small for new technologies while those for existing technologies are >0.37. The impact of outliers is significant because their inclusion reduces the cost of existing technologies to such an extent that the order between the two lines becomes reversed for altitude > ~ 40 km. When we take the altitude (not its logarithm) as the dependent variable, the relationship generally degrades. When the data set of Davidson et al. (2012) is included, the slope for new technologies becomes even more negative.

Fig. 8

Results of regression analysis of the lifting cost onto the logarithm of altitude

Fig. 9

Results of regression analysis of the lifting cost onto altitude

Although this exercise is useful for highlighting the general pattern, we emphasize that each study took a distinct method of different quality. We therefore do not utilize the results of the regression analyses in estimating the total costs.

Appendix 4: Sensitivity to discount rates

Although we amortized the cost at a discount rate of 10 % over a 20-year period, the choice of discount rate is always subject to intensive debate in climate change economics. We therefore briefly explore the sensitivity of our results to discount rates. The total annual cost can be written as

$$ T(r)=I\cdot A(r)+O, $$

(A4.1)

where T is the total annual cost (T = (L + P + D)F/E), I is the initial capital expenditure, and O is the cost of operation and maintenance. A is defined as A = r/{1 − (1 + r)^−N}, where r is the discount rate and N is the number of time periods (here N = 20). The percentage difference due to a change in the discount rate is

$$ \frac{\Delta T}{T}=\frac{T(r)-T\left({r}_0\right)}{T\left({r}_0\right)}=\frac{A(r)/A\left({r}_0\right)-1}{1+\left(O/I\right)/A\left({r}_0\right)}. $$

(A4.2)

In the limit of O/I going to infinity, the difference vanishes, i.e., ΔT/T → 0. On the other hand, if O/I goes to zero, ΔT/T → A(r)/A(r ₀) − 1. For r = 0.03 and r ₀ = 0.10, ΔT/T →  ~  − 0.427._.

Figure 10 shows the sensitivity of ΔT/T to O/I with r = 0.03 and r ₀ = 0.10. The data from Davidson et al. (2012), for example, suggest that the range of O/I is 0 to ~200. Therefore, although a lower discount rate would lower the direct cost further, it would not change the order of magnitude.

Fig. 10

Sensitivity of the percentage difference due to a change in the discount rate (∆T/T) to the relative cost of operation and initial capital expenditure (O/I)

Appendix 5: Cumulative production of F-15 fighter jets

This section documents the cumulative production of F-15s. In short, there are various numbers for the cumulative production depending on the variants covered. Boeing’s backgrounder (Boeing Defense Space & Security 2013) reports the cumulative production as in excess of 1,600. Davies and Dildy (2007) provide a detailed dataset of the past production for the F-15A/B/C/D/J/DJ, with a total production figure of 1,198. GlobalSecurity.org, a website that documents numerous defense-related statistics, shows yet another value (GlobalSecurity.org 2011). Their webpage on the F-15 lists the cumulative production as 1,712, including variants not covered by Davies and Dildy (2007). However, their data do not distinguish conversion from production.

In the main text, we take 1,600 as the value of cumulative production in Fig. 6. In the four decades since the initial production of the first F-15 in 1972, average production has been approximately 40, which is the number utilized in Fig. 6. The production numbers of 1,600 in total and 40 per annum do not differ appreciably from numbers quoted in other sources, at least for the purposes of order-of-magnitude comparisons.