Studying geoengineering with natural and anthropogenic analogs
- First Online:
- Cite this article as:
- Robock, A., MacMartin, D.G., Duren, R. et al. Climatic Change (2013) 121: 445. doi:10.1007/s10584-013-0777-5
- 694 Views
Solar radiation management (SRM) has been proposed as a possible option for offsetting some anthropogenic radiative forcing, with the goal of reducing some of the associated climatic changes. There are clearly significant uncertainties associated with SRM, and even small-scale experiments that might reduce uncertainty would carry some risk. However, there are also natural and anthropogenic analogs to SRM, such as volcanic eruptions in the case of stratospheric aerosol injection and ship tracks in the case of marine cloud albedo modification. It is essential to understand what we can learn from these analogs in order to validate models, particularly because of the problematic nature of outdoor experiments. It is also important to understand what we cannot learn, as this might better focus attention on what risks would need to be solely examined by numerical models. Stratospheric conditions following a major volcanic eruption, for example, are not the same as those to be expected from intentional geoengineering, both because of confounding effects of volcanic ash and the differences between continuous and impulsive injection of material into the stratosphere. Nonetheless, better data would help validate models; we thus recommend an appropriate plan be developed to better monitor the next large volcanic eruption. Similarly, more could be learned about cloud albedo modification from careful study not only of ship tracks, but of ship and other aerosol emission sources in cloud regimes beyond the narrow conditions under which ship tracks form; this would benefit from improved satellite observing capabilities.
Geoengineering by means of solar radiation management (SRM) has been suggested as a potential approach (in concert with mitigation of greenhouse gas emissions) to manage climate change (Crutzen 2006; Shepherd et al. 2009; GAO 2011). We focus here on two SRM ideas in particular: the intentional introduction of stratospheric aerosols to scatter some incoming sunlight (e.g., Budyko 1977), and altering the albedo of marine boundary layer clouds by injecting additional aerosols (Latham 1990). Before decisions can be made about implementation, it is essential to improve our scientific understanding of likely positive and negative impacts. Much of this understanding can come from numerical modeling. Outdoor experiments might address some gaps in knowledge, but even small-scale experiments outside a laboratory environment could carry some risk (SRMGI 2011). However, for both of the concepts considered here, there are natural or anthropogenic analogs: volcanic eruptions have provided the motivation for stratospheric aerosol SRM, while observations of ship tracks have provided the motivation behind marine cloud brightening. The processes related to these analogs are also important for understanding climate change itself. Here we discuss using analogs to study SRM.
While volcanic eruptions provide the evidence that increased stratospheric aerosols would indeed cool the planet, there are many reasons for concern about geoengineering with stratospheric aerosols (Robock 2008), with many of these concerns yet to be quantified. Increasing marine boundary layer cloud albedo through injection of sea-salt aerosols to form additional cloud condensation nuclei (CCN) could have different undesired side-effects than stratospheric aerosols (e.g., Jones et al. 2009), and the effectiveness is more poorly understood. For example, the conditions under which adding CCN would increase cloud albedo are not well known (Wang et al. 2011).
A long-term roadmap for geoengineering research (e.g., Caldeira and Keith 2010) would clearly involve more modeling studies than have been done to date, possibly some limited small-scale but open-atmosphere experiments to resolve specific process questions (David Keith and James Anderson, personal communication, 2012), and only if implementation were planned, an initial subscale deployment phase to better understand the climate response (MacMynowski et al. 2011a, b); progress would also be needed in governance appropriate to each stage. However, missing from this description is that much can be learned from a better understanding of natural and anthropogenic analogs, both to directly understand potential consequences, and to evaluate models. This knowledge could minimize or altogether avoid any risky experimentation with the planet. Here we discuss fundamental questions about SRM that can be studied using analogs.
2 Volcanic analogs
The observation that large volcanic eruptions cool the planet was one of the original motivations for suggesting geoengineering (e.g., Budyko 1977; Crutzen 2006), with the eruption of Mount Pinatubo in 1991 for example cooling the planet by roughly 0.5 °C (Soden et al. 2002) by the injection of 20 Mt sulfuric acid into the stratosphere, producing more than 30 Mt of sulfate aerosols (Bluth et al. 1992). However, while it is clear from these natural analogs of geoengineering that “mimicking” a volcanic eruption by producing sulfate or other aerosols in the stratosphere will result in cooling, there are many uncertainties regarding both the effectiveness and the side effects (i.e., the risks). One of the most valuable opportunities for reducing the uncertainties and risks of geoengineering with stratospheric aerosols thus comes from further study of past volcanic eruptions and from studying the climate system response to future volcanic eruptions.
One of the main differences between a somewhat permanent stratospheric aerosol cloud proposed for geoengineering and clouds produced by volcanic eruptions is the lifetime. The e-folding lifetime of stratospheric clouds from tropical volcanic eruption is about one year (Robock 2000), while it is 2–4 months for those from high latitude eruptions (Kravitz and Robock 2011). (This also informs us about the frequency of stratospheric aerosol precursors that would be needed to maintain a cloud in the stratosphere.) The difference in lifetimes means that climate system responses with long time scales, such as oceanic responses, would be different between volcanic eruptions and geoengineering, but rapid responses, such as seasonal responses of monsoon circulations and precipitation would be quite similar, and the volcanic analog would be appropriate. For example, MacMynowski et al. (2011a, b) showed that precipitation response to stratospheric forcing had only a weak dependence on the frequency of the applied forcing, in contrast to the temperature response, which depends on the longer timescales imposed by ocean thermal inertia.
2.1 Lessons from past volcanic eruptions
Volcanic eruption analogs already tell us many things about the potential effects of stratospheric aerosol clouds. These were briefly discussed by Robock et al. (2008), but there are many more examples, discussed here, including additional things that could be learned from more studies. The beneficial impacts include:
Cool the surface, reducing ice melt and sea level rise
It is well-known that global average climate cools after large volcanic eruptions (Robock 2000). After the 1991 Mt. Pinatubo eruption, in addition to the global cooling, Stenchikov et al. (2009) and Otterå et al. (2010) found long-term impacts on ocean heat content and sea level, and Zanchettin et al. (2010) found an impact on North Atlantic Ocean circulation a decade later, so we might expect impacts from SRM also, but would need models and not observations to quantify them.
Increase the CO2 sink
Following volcanic eruptions, observations show an increase of the CO2 sink from global vegetation. The main cause is a shift from direct to diffuse solar radiation (Robock 2000), which enhances vegetation growth (Mercado et al. 2009). But net primary productivity also responds to temperature and precipitation changes, and vegetation adjusts to changing conditions, so the net effect from a continuous stratospheric aerosol cloud needs further study.
However, volcanic analogs also suggest a number of negative effects from a continuous stratospheric aerosol cloud. These include:
Reduced summer monsoon precipitation
The reduction in sunlight after large volcanic eruptions cools land more than oceans. In the summer, this reduces the temperature contrast between warm continents and cooler oceans, weakening the African and Asian summer monsoon circulation and the resultant precipitation. This has been observed after every major volcanic eruption, including 1783 Laki and 1912 Katmai (Oman et al. 2006), 1982 El Chichón (Robock and Liu 1994), and 1991 Pinatubo (Trenberth and Dai 2007). Anchukaitis et al. (2010) showed the average effect on the summer Asian monsoon using tree rings for many centuries. Whether this effect is truly dangerous depends on the proposed SRM strategy, but it would be difficult to design an SRM strategy without negative impacts on precipitation (Ricke et al. 2010).
Destroy ozone, allowing more harmful UV at the surface
Observations following the 1982 El Chichón and 1991 Pinatubo eruptions showed additional ozone depletion because of heterogeneous chemistry on the additional stratospheric aerosols, in the same process that produces the spring ozone hole over Antarctica on polar stratospheric clouds (Solomon 1999). This has also been simulated in response to SRM (e.g., Tilmes et al. 2008).
Produce rapid warming when stopped
Observations show that once a volcanic cloud is removed from the atmosphere, the climate system rapidly warms. If geoengineering were implemented for a long time and then stopped, this warming rebound would produce a much more rapid climate change than the gradual climate change now happening because of increasing greenhouse gases.
Make the sky white
A volcanic aerosol cloud makes the sky whiter, particularly near the Sun, where a large amount of the sunlight is forward scattered (e.g., Plate 3, Robock 2000). Kravitz et al. (2012) showed that this would also be the case for stratospheric SRM. However, it would produce pretty sunsets (Zerefos et al. 2007).
Reduce solar power
The same process that increases diffuse sunlight reduces direct sunlight, affecting solar thermal electricity generation. Murphy (2009) found that for nine solar thermal power plants in California during the summer of 1992 after the 1991 Pinatubo eruption, the summer on-peak capacity was reduced by 34 % from pre-Pinatubo levels, because of a reduction in direct solar radiation.
Perturb the ecology with more diffuse radiation
The same mechanisms that would increase the CO2 sink would affect different plants differently, and the net effect on ecosystems and agriculture is not clear. Certainly there would be changes.
Damage airplanes flying in the stratosphere
Following the 1991 Pinatubo eruption, in addition to direct airplane damage from volcanic ash encounters immediately after the eruption, there was long-term damage to airplanes flying through a dilute sulfuric acid bath, particularly on polar routes where commercial aircraft entered the lower stratosphere. For example, this required more frequent replacement of windows after the 1982 El Chichón eruption (Bernard and Rose 1990).
Degrade astronomical observations
Any cloud that reflects some sunlight back to space will also reflect starlight. Furthermore, it will heat the stratosphere, producing enhanced downward longwave radiation, and could impact stratospheric water vapor content; these would affect IR astronomy. How important these effects would be for astronomical observations remains to be determined. It would be interesting to search for such effects after the 1991 Pinatubo eruption, and determine how such a cloud in the future would affect modern astronomical equipment and stargazing.
Affect remote sensing
A stratospheric aerosol cloud would also affect shortwave and longwave radiation leaving Earth and observed by satellites. After the 1982 El Chichón eruption, the simultaneous development of a very large El Niño was not detected for months, since the enhanced longwave emissions from the warm ocean were masked by the stratospheric cloud (Strong 1984). At the same time, famine warning systems were triggered by erroneous inputs to normalized difference vegetation index calculations.
2.2 What more can we learn from future eruptions?
While past volcanic eruptions inform us of some of the potential impacts of stratospheric aerosol clouds, there are several additional questions that can be addressed by planning for observations of the next large eruption, as well as additional study of past ones. These include:
What will be the size distribution of sulfate aerosol particles created by geoengineering?
Will they remain at the typical effective radius of about 0.5 μm observed after Pinatubo, or will they grow as additional sulfate creates larger rather than more particles? Even though a typical large volcanic eruption is a one-time stratospheric injection, we can learn from the initial processes of conversion from SO2 gas to sulfate particles and then to particle growth. The issue of how particle sizes evolve for geoengineering has been addressed through simulations (Heckendorn et al. 2009; Hommel and Graf 2010; English et al. 2012), but there are limited data to support analysis. It is also important to understand how particle size evolution depends on injection strategy (injecting SO2 or H2SO4) and the pattern of injection (Pierce et al. 2010; English et al. 2012). Such models can be tested by imposing the exact emissions from future volcanic eruptions, if the particle evolution from the eruptions is well monitored.
How will the aerosols be transported throughout the stratosphere?
Under what conditions do tropical injections gradually spread globally? Do injections in the subtropics stay in one hemisphere? What are their lifetimes? How do high latitude injections behave? How does the phase of the Quasi-Biennial Oscillation affect the transport? Does the El Niño/Southern Oscillation (ENSO) phase play a role through tropospheric impacts on atmospheric circulation? What is the dependence on the height of the injections? This work could build on studies of nuclear bomb tests and past eruptions (e.g., Gao et al. 2007).
How do temperatures change in the stratosphere as a result of the aerosol interactions with shortwave (particularly near IR) and longwave radiation?
Is there a response in the circulation to these temperature and resulting geopotential height changes? This question is intimately related to the question above and the next two questions.
Are there large stratospheric water vapor changes associated with stratospheric aerosols? Is there an initial injection of water from the eruption?
How do temperature and circulation changes in the stratosphere affect the tropical tropopause layer, and does heating this layer allow more water to enter the stratosphere? There were not robust observations of large impacts of the 1991 Pinatubo eruption on stratospheric water vapor, but was this a result of a poor observing system?
Is there ozone depletion from heterogeneous reactions on the stratospheric aerosols?
How do changes in other species, such as H2O, NOx, and those containing Br and Cl, interact with the ozone chemistry, and what is the dependence on temperature changes and the location and time of year of the aerosols? Simulations of increased aerosol loading have also found changes in upper tropospheric chemistry (Hendricks et al. 1999).
As the aerosols leave the stratosphere, and as the aerosols affect the upper troposphere temperature and circulation, are there interactions with cirrus clouds?
Do cirrus clouds increase or decrease, and how do these changes depend on the aerosol concentration and particular atmospheric conditions? How can observed cirrus changes be attributed to volcanic effects as compared to changes that would take place in response to normal climate variability? The connection between stratospheric sulfate aerosols and cirrus clouds in the upper troposphere has been studied in the context of volcanoes, with some studies indicating an effect from volcanic eruptions mixed with a signal from ENSO (e.g., Wylie et al. 1994; Sassen et al. 1995; Song et al. 1996), but others finding no impact (Luo et al. 2002; Massie et al. 2003; Lohmann et al. 2003). The issue is important and not yet resolved, but the Kuebbeler et al. (2012) modeling study found that cirrus impacts of stratospheric geoengineering would enhance the global cooling by depleting the cirrus clouds.
How will tropospheric chemistry be affected by stratospheric geoengineering?
What is the impact of the “rain-out” of stratospheric aerosols into the upper troposphere? Will the changing distribution of ultraviolet light caused by ozone depletion have subsequent impacts on the troposphere, particularly through OH and NOx chemistry?
2.3 Differences between volcanic eruptions and stratospheric geoengineering
Volcanic eruptions are clearly analogous to SRM using stratospheric aerosols in many ways, and thus serve as an important component of addressing the uncertainties listed above. However, there are also a few important differences:
Volcanic eruptions are into a clean stratosphere
The most significant difference is that injecting sulfate into a “clean” stratosphere results in a different coagulation problem from a continuous injection scenario. Theoretical studies show that massive volcanic eruptions (Timmreck et al. 2010) or continuous injection (Heckendorn et al. 2009) will result in larger particles than after a one-time injection such as from the 1991 Pinatubo eruption. The larger mean radii expected for geoengineering would result not only in higher concentrations being required to obtain the same radiative forcing, but also more rapid fallout into the troposphere, which would both increase the injection rate required to sustain the desired geoengineering effect and increase the potential for impacts on cirrus and upper tropospheric chemistry.
Volcanic eruptions also include significant ash
Therefore, it may be difficult to determine whether any initial effect observed (or not) on cirrus cloud formation, for example, is due to the ash rather than the sulfate. The lifetime of the ash is shorter than that of the aerosols, so this attribution question is primarily a challenge immediately after an eruption, but very small ash may serve as nuclei for sulfate aerosols and their effects may persist much longer.
The time-scale of radiative forcing is different
This needs to be taken into account in extrapolating between the climate response observed after a volcanic eruption and what would be expected for continuous injection. For example, land-sea temperature contrast and precipitation respond to radiative forcing changes relatively rapidly (Dong et al. 2009), but global mean temperature changes more slowly, and hence the ratio of precipitation to temperature changes should be expected to be much more pronounced after a volcanic eruption than due to continuous SRM (MacMynowski et al 2011b).
Because of the above differences, observations cannot be used as a direct estimate for conditions under continuous geoengineering. Regardless of the data available after an eruption, there will remain uncertainty in the factors listed above. These uncertainties can be limited by modeling or more representative outdoor direct testing, which for some uncertainties may require “tests” large enough to look more like deployment (Robock et al. 2010). Because of governance and other issues, such in situ testing may never take place (Robock 2012).
2.4 Volcanic monitoring
The ability to successfully take observations after a volcanic eruption would be extremely valuable for validating models. However, previous large eruptions have not been sufficiently well monitored. More information is required, for example, regarding the initial aerosol concentrations in order to better validate particle formation, coagulation, and evolution models. Thus we make two recommendations.
First, more can be learned from further data mining from past eruptions; in addition to improving our knowledge, this will also clarify the observational gaps that need to be filled. The focus specifically on the uncertainties associated with geoengineering leads to a different perspective and hence possibly different questions from what might be asked if the goal were solely to understand volcanic eruptions. For example, it is insufficient to know whether a volcanic eruption does or does not have some impact on cirrus, without being able to separate out effects due to ash, or understand the dependence on the aerosol size distribution.
Second is to develop either a rapid response system or system for continuous observations so that we are ready for the next large volcanic eruption, and can gather the data needed to validate models. The evolution in stratospheric sulfate aerosol size distribution occurs over the first few months after an eruption (Stenchikov et al. 1998; English et al. 2011, 2013), underscoring the need for a rapid response capability. Sustained observations would be required from less than roughly 3 months to 18 months following a massive eruption to capture the initial ramp-up, peak, and ramp-down of aerosol concentrations.
To provide data for validating the modeling of particle size distributions and their evolution, a volcanic monitoring system would need to obtain observations during the first few months after an eruption. This means that any rapid response system needs to be available for deployment at any time, with funding in place for the personnel and equipment. This rapid response capability needs to be in addition to sustained background observations (e.g., Deshler et al. 2003).
To be of most use, a volcanic cloud monitoring system will need to measure the spatial peak (highest concentration) of the plume. Limb-scanning satellite measurements, such as SAGE II, did not see the densest part after the 1991 Pinatubo eruption (Stenchikov et al. 1998). For balloon-based observing, this also requires a plume forecast capability (Vernier and Jumelet 2011). Satellite observations will also need independent data on the aerosol size distribution if existing retrieval techniques depend on such assumptions. Stratospheric chemistry observations will require high resolution measurements with stratospheric balloons or high altitude aircraft. Cirrus is adequately observed with existing systems (Sassen et al 2008; Vernier et al. 2009); uncertainties in cirrus impact are thus related to natural variability, and uncertainties in aerosol concentrations in the densest part of the volcanic plume.
3 Ship tracks and marine cloud brightening
Increasing the brightness of marine boundary layer clouds through the injection of aerosols such as sea salt (Latham 1990) has also been proposed as a means of solar radiation management. This strategy derives from the observation of ship tracks, where, depending on conditions, there is a clear cloud signal resulting from the injection of aerosols from the ship exhaust (Christensen and Stephens 2011). However, the complexity of cloud-aerosol interactions results in substantial uncertainties as to the effectiveness of this approach. As in the case of using volcanic eruptions as an analog to stratospheric aerosol geoengineering, there is much that can be learned from analogs. In this case, the principal analogs are anthropogenic, in the form of ship exhaust or emissions from coastal sites, although volcanic plumes in the boundary layer have also been explored (Yuan et al. 2011). A more thorough analysis of existing data would both improve our knowledge and clarify the observational gaps that need to be filled. There are also observational gaps that limit our current ability to assess this approach, such as the entrainment rate, or direct measurement of albedo at high spatial resolution.
The key concept is that increasing the number of cloud condensation nuclei (CCN) while keeping cloud liquid water constant results in more, smaller droplets, and an increase in cloud albedo, the “Twomey” effect (Twomey 1974). However, liquid water path (LWP) rarely remains constant, due to changes in precipitation and entrainment with increasing aerosol (Ackerman et al 2004), and these changes can produce radiative impacts of the same order as those predicted from the Twomey hypothesis (e.g., Lohmann and Feichter 2005). Stratocumulus clouds also tend to naturally “buffer” against processes (such as changing aerosol) that change cloud albedo and precipitation (Stevens and Feingold 2009), for example through changes in entrainment. As a consequence, robust relationships among changes in precipitation, cloud albedo, and cloud coverage have not yet been established from observations. Furthermore, we have inadequate observations to analyze the processes which influence these cloud properties. The challenges in understanding all of the feedbacks involved, and when the introduction of aerosols leads to greater albedo and when it does not, points to the need both for careful data analysis and for greater observational capability.
3.1 Key uncertainties
The sensitivities of marine cloud albedo to specific processes and parameters are poorly understood (e.g., entrainment, LWP, turbulent kinetic energy (TKE), cloud droplet number concentration, cloud fraction), which limits our ability to determine under what conditions the net albedo increases with increased aerosols. In particular, no observational studies are able to measure the albedo sensitivity to entrainment and TKE.
Much of the data analysis to date has focused on ship tracks, as they represent the most visible change due to aerosols. However, exhaust plumes do not always produce ship tracks, and the clouds that are receptive to the plumes span a limited range of stratocumulus conditions, typically less than 1 km cloud top height in a relatively clean environment (Coakley et al. 2000). It is also important to understand the aerosol indirect effect on clouds from other (non-ship track) emissions and pollution, including large smelters and volcanic plumes. Given the larger variability and range of environmental conditions, there could be greater uncertainty in the magnitude of the effect of additional aerosols on cloud albedo outside of the narrow range of conditions where ship tracks are visible.
Assessment of the predicted climate response to the spatially inhomogeneous radiative forcing introduced by selective brightening of marine boundary layer clouds. To offset a significant fraction of anthropogenic radiative forcing using this approach, large changes in radiative forcing would be required over relatively small spatial extent, with unknown climate impact. For example, simulations by Jones et al. (2009) offset 35 % of the radiative forcing due to current greenhouse gases with marine cloud brightening, but found detrimental effects on precipitation and net primary productivity in some regions. There could also be a large impact on drizzle and precipitations along coastlines; further assessments are clearly needed.
3.2 What have we learned, and what are the gaps?
Cloud properties measured in different studies, or by satellite observations (bottom row). Studies include MAST (Durkee et al. 2000), MASE-I & II (Lu et al. 2007, 2009), E-PEACE (Russell et al. 2013), VOCALS (Wood et al. 2011), and DYCOMS-II (Stevens et al. 2003). Measured properties listed here include cloud condensation nuclei (CCN), cloud droplet number concentration (Nd), cloud drizzle properties, CCN chemistry and microphysics, turbulence, entrainment rate, either liquid water content (LWC) or liquid water path (LWP), overall albedo changes, and cloud thickness measurements; measurements of entrainment and albedo are clear observational gaps in most of these experiments
CCN chemistry & microphysics
MASE-I & II
Aerosol-cloud interactions are complex and cloud albedo is not always enhanced by increasing the aerosol concentration. For example, Christensen and Stephens (2012) found that cloud dimming occurred as frequently as cloud brightening when ship tracks were observed in precipitating closed cellular clouds. Cloud dimming primarily resulted from decreases in liquid water path caused, presumably, by the enhanced entrainment of the dry overlying air into the polluted clouds with smaller droplets. By contrast, ship tracks observed in open cells, where the free-troposphere is relatively moist by comparison, almost always exhibited cloud brightening compared to the surrounding unaffected clouds. The extent of LWP adjustments in response to changes in aerosol concentrations remains largely uncertain for low-level clouds as a whole, because these changes are linked to changes in entrainment and moisture in the free-troposphere, and these variables are either not measured at all from space (entrainment) or not measured with sufficient accuracy (moisture) to capture mixing at the entrainment interface.
Despite this progress on exploring the impact of aerosols on observed ship tracks, radiative forcing estimates of these “linear” ship tracks from satellite observations cast substantial doubt on the efficacy of using SRM strategies to brighten low-level clouds. Schreier et al. (2007) demonstrate that the radiative effect can be as large 100 W m−2 at the individual scale of the ship track, however, when integrated over the globe, the annual mean effect is negligible (−0.4 × 10−3 to −0.6 × 10−3 W m−2). Similar results were identified by Peters et al. (2011), in which the properties of clouds were unchanged even near the world’s most densely populated shipping lanes. Although the impact has been shown to be negligible on the global scale, ship tracks can still inform process understanding of aerosol-cloud interactions on the cloud and regional scale. The aerosol indirect forcing in an individual ship track is inferred from space using Moderate Resolution Imaging Spectroradiometer (MODIS)-derived optical cloud properties, which leads to significant uncertainty in partly cloudy conditions, since there is insufficient spatial resolution from current albedo measurements (e.g., Clouds and Earth’s Radiant Energy System (CERES) footprint is ~20 km). Higher resolution (~1 km) satellite-based albedo measurements would improve the assessment of aerosol indirect effects in “linear” ship track observational studies, and thus improve our understanding of aerosol indirect effects at the process level.
Aerosol plumes that do not produce ship tracks but nonetheless affect the properties of clouds after becoming widely dispersed are difficult, if not impossible to detect using current satellite technology. Goren and Rosenfeld (2012) describe a case study in which the emissions from ships affect the properties and increase the abundance of closed cellular stratocumulus for several days. It is anticipated that this may significantly contribute to the global aerosol indirect forcing because sulfur emissions from shipping largely outweigh the natural biogenic production in many oceanic regions, especially in the Northern Hemisphere (Capaldo et al. 1999). Presumably, a small fraction of these emissions goes into producing ship tracks, while the remaining aerosol affects the properties of stratocumulus to an unknown extent. General circulation model simulations (Capaldo et al. 1999; Lauer et al. 2007) indicate that the radiative effect from shipping could be as large as 40 % of the total aerosol indirect forcing due to all anthropogenic activities. Given the large discrepancies in the radiative forcing between satellite observations and climate model results, this is an outstanding problem.
There may be additional opportunities to quantify the difference in the overall cloud albedo from ship traffic. For example, radiative effects may manifest via the gradual phase-out of high sulfur content bunker fuel over the next few decades (International Maritime Organization 1998) or manifest in the remote Arctic ocean regions as ships will have the ability to travel in this area as sea ice progressively melts.
Finally, understanding the climate response to brightening marine boundary layer clouds would benefit from a new Geoengineering Model Intercomparison Project (GeoMIP) surrounding low cloud albedo enhancement. The current GeoMIP study (Kravitz et al. 2011) explores spatially uniform reductions in sunlight or stratospheric aerosols. Since not all models have clouds in the same locations, or clouds receptive to albedo modification, care must be taken as to whether a model intercomparison project is testing the robustness of the model-predicted response to spatially inhomogeneous radiative forcing perturbations, or testing differences between predicted cloud distributions, or testing differences between model parameterizations of cloud-aerosol interaction. The GeoMIP project is currently expanding to conduct such experiments.
Any long-term research strategy for evaluating geoengineering must include as an essential component the evaluation of natural and anthropogenic analogs, volcanic eruptions in the case of stratospheric aerosols and ship-tracks and other emission sources in the case of marine boundary layer cloud brightening. These are imperfect analogs, and will not provide all of the information required to assess effectiveness and risks. However, the ability of models to match observations of analogs would increase confidence in their predictions of geoengineering effects. Thus better evaluation of analogs could minimize the need for open-atmosphere testing of geoengineering.
Current observational capabilities are insufficient to address geoengineering risks. It is particularly important to improve our observational capabilities prior to the next large volcanic eruption, so that our best opportunity to better understand stratospheric geoengineering is not missed. Similarly, improved instrumentation could improve our assessment of the global aerosol indirect effect, in order to understand the potential for marine cloud brightening beyond the narrow set of conditions in which ship tracks form. This is also timely, as changes in shipping fuel may soon provide an unintended experiment, but one where we have not yet adequately characterized the current baseline.
While the questions posed here are motivated by the need to better understand geoengineering, addressing these questions would have major co-benefits to climate science in general, by addressing key uncertainties in the models.
We thank the Keck Institute for Space Studies for funding two workshops at the California Institute of Technology where we discussed topics in this paper, and all of the participants of these workshops who contributed (http://www.kiss.caltech.edu/study/geoengineering). A. Robock is supported by NSF grant AGS-1157525. The work by R. Duren was done at the Jet Propulsion Laboratory, a division of the California Institute of Technology under contract to the National Aeronautics and Space Administration.