This chapter reviews a number of proposed strategies for minimizing the damage and risks from climate change by modifying Earth’s energy budget. The chapter begins with a discussion of idealized studies that provide insight into the general response of the climate system to albedo changes. Two more realistic strategies (stratospheric aerosol injection and marine cloud brightening) are then discussed in greater detail because studies suggest they have the potential to produce a significant cooling and/or they have been discussed more widely in the literature. Other methods that have received less attention or appear to be impractical are discussed briefly later in the chapter, followed by a discussion of observational problems concerning Earth’s radiation budget and climate response to albedo modification that are common to all albedo modification techniques. This chapter concludes with a series of tables summarizing the committee’s assessment of various aspects of these albedo modification strategies.
Although simple energy balance principles, backed up by observations of volcanic cooling, are sufficient to establish that reducing the amount of solar radiation absorbed by Earth can reduce the global mean surface temperature, they do not constrain the geographic or seasonal pattern of temperature that would prevail in an albedo-modified world. These patterns are determined not only by the top-of-atmosphere fluxes, but also by the transport of heat and moisture by atmospheric circulations, the transport of heat by ocean circulations, and various complex regional feedbacks including changes in cloud properties. These processes are represented, with varying degrees of fidelity, in atmosphere-ocean general circulation models (GCMs). Representation of the complex chain of processes linking a specific climate intervention (e.g., injection of SO2 gas into the stratosphere) to the resulting albedo change poses very considerable challenges. Idealized simulation studies bypass the modeling of this complex chain of events, instead directly imposing a reduction in absorbed solar radiation. Earth’s near-surface environment is the product of a complex interacting system involving physics, chemistry, and biology of the land, ocean, and
atmosphere. This real system has far greater complexity than does any model, and thus no model of this system can provide a quantitatively reliable detailed prediction of how Earth will respond to a novel occurrence. Nevertheless, model simulations and theory do suggest some basic properties of the response of the climate system to reductions in the amount of sunlight absorbed.
There is no known way to modify albedo to yield a pattern of top-of-atmosphere solar radiative forcing that is similar (seasonally and geographically) but of opposite sign and amplitude to the radiative forcing pattern due to an increase of CO2 (see, for example, Figure 3 of Kravitz et al., 2013a). A change in albedo has little or no effect at night or in mid- to high-latitude winters, where there is little or no sunlight to reflect, but these areas are influenced by CO2 radiative forcing. A spatially uniform decrease in sunlight also leads to more radiative forcing in the tropics than near the poles, because the annual mean incident solar radiation is greater in the tropics. Even if CO2 and albedo changes could cause the same change in the top-of-atmosphere energy balance, they would cause different changes in the surface energy budget; hence, any albedo modification designed to cancel out the top-of-atmosphere CO2 radiative forcing will cause changes in the surface energy budget, relative to the preindustrial state (Bala et al., 2008; Pierrehumbert, 2010, Chap. 6). The climate response may be geographically more similar than the forcing since the atmospheric and ocean circulation processes that redistribute energy are the same for CO2 radiative forcing and albedo change (Govindasamy and Caldeira, 2000). Idealized simulations can shed light on how the climate system responds to these disparities in forcing. The idealized experiments do not, however, address the question of how closely the targeted reduction in solar absorption can be met through the various proposed albedo modification techniques.
In the hierarchy of attempts to simulate the effects of albedo modification, the most idealized experimental protocol is to reduce the global mean absorption of sunlight by simply reducing the amount of sunlight incident on the top of the atmosphere. This quantity is characterized by a parameter known as the “solar constant,” which is a measure of the power output of the Sun. The amount of solar energy absorbed by Earth in a simulation can be reduced by any desired amount by simply dialing down the value of the solar constant in a model, which is essentially equivalent to reducing the brightness of the Sun. This protocol is easy to implement in any climate model and, therefore, is well suited to multimodel comparison projects. The forcing achieved in solar-constant experiments has a lot in common with that resulting from introducing a very uniform aerosol layer into the stratosphere (Kalidindi et al., 2014), but it has less in common with
the more inhomogeneous forcing resulting from marine cloud brightening or regionally limited modifications of stratospheric aerosols. Solar-constant experiments provide considerable insight into the fundamental climate processes involved in determining the joint response to increased CO2 and reduced solar absorption, but they do not incorporate some important effects connected with the vertical redistribution of heating in that atmosphere, notably the stratospheric heating that would result from increasing the stratospheric aerosol content (Kalidindi et al., 2014). They also do not incorporate the effects of injected substances on atmospheric chemistry, on cloud properties, or on the transformation of direct-beam to more diffuse sunlight.
There is by now a quite considerable literature on solar-constant experiments, which the committee does not attempt to survey comprehensively. Earlier work with sunlight reduction studies is reviewed by Caldeira et al. (2013). The most extensive analysis of solar-constant experiments has been carried out as part of the G1 experiment of the multimodel intercomparisons of the Geoengineering Model Intercomparison Project (GeoMIP) (Kravitz et al., 2013a; see Box 3.1), which allow a search for robust signatures using a standard experimental design. Because the GeoMIP simulations are of limited duration (under a century), the deep ocean does not have time to come into equilibrium with the climate forcing. These G1 and 4×CO2 (a quadrupling of the CO2) simulations therefore do not provide an indication of how the climate would evolve if the albedo modification were maintained for centuries, allowing the deep ocean to respond, although because the changes in surface flux (heat, moisture, momentum) are much smaller than the changes produced by CO2 forcing the ocean response is also smaller. These conclusions should apply generally to all albedo modification simulations done so far, including those with a more sophisticated representation of the albedo modification process.
Here the committee highlights only a few key results of the G1 experiment of GeoMIP. Figures 3.1 and 3.2 show temperature changes produced from a quadrupling of CO2, and the result of a reduction in sunlight sufficient to return the global average surface
BOX 3.1 GEOENGINEERING MODEL INTERCOMPARISON PROJECT (GEOMIP)
More than a dozen modeling groups have participated in a modeling intercomparison project—referred to as GeoMIP—to examine the effects of albedo modification (Kravitz et al., 2011a). The first set of experiments as part of the fifth phase of the Coupled Model Intercomparison Project (CMIP5) focused on four scenarios related to stratospheric aerosol albedo modification (Kravitz et al., 2013a), but other experiments under this framework will add experiments on marine cloud brightening and cirrus thinning.
FIGURE 3.1 Zonal average anomalies in surface air temperature (K; land + ocean average; 12 models), precipitation minus evaporation (mm/day; land average; 12 models), and terrestrial net primary productivity (kg C m−2 yr−1; land average; 8 models) for all available models. All values shown are averages over years 11-50 of the simulations. The x axis is weighted by cosine of latitude. SOURCE: Kravitz et al., 2013a.
temperature to a reference (approximately preindustrial) state. The reduction in sunlight reduces the mid- to high-latitude warming, which exceeds 5°C at the South Pole and 10°C at the North Pole, to about 1°C and reduces the surface temperature change in the tropics from about 5°C warmer (4×CO2 simulation) to 0.2°C to 0.5°C cooler (G1 simulation) than the reference states. This general pattern, which is robust across all solar-constant experiments, occurs because reducing the solar constant in such a way as to offset the global mean CO2 radiative forcing undercompensates this forcing in the high latitudes (where there is comparatively little sunlight to reflect) but makes up for it in the global mean by overcompensating in the more highly illuminated tropical regions. Atmospheric and oceanic heat transports redistribute the excess heating from one place to another, which reduces the geographic inhomogeneity of the tempera-
FIGURE 3.2 Left: Temperature changes produced from a quadrupling of CO2. Right: The resulting temperature change from a reduction in sunlight sufficient to return the global average surface temperature to a reference state (approximately preindustrial state). Results are for all-model ensemble annual average surface air temperature differences (K) averaged over years 11-50 of the simulation. Top row shows December-January-February average, and bottom row shows June-July-August average. Stippling indicates where fewer than 75% of the models (for this variable, 9 out of 12) agree on the sign of the difference (typically where projected temperature changes are small). SOURCE: Kravitz et al., 2013a.
ture response but does not eliminate it. Despite the agreement among models on the latitudinal pattern of temperature responses, there is considerable disagreement among the models in the GeoMIP G1 ensemble as to the sign of the temperature response over much of the tropical land area because there are very small changes in those areas.
Figure 3.3 summarizes the global mean precipitation response in the GeoMIP G1 experiment albedo modified states. Energy is required to sustain evaporation and precipitation must ultimately balance evaporation, so the surface energy balance plays an important role in determining precipitation changes. Reduction in the amount of sunlight reaching the surface tends to decrease precipitation, especially in the warm tropics (Pierrehumbert, 2002, 2010, Chap. 6). Stabilization of the surface layer produced by changes in heating and cooling rates (the result of heating aloft from CO2 concentration increases, and sunlight reduction reducing surface heating) reduces mixing near the surface, causing further changes in both evaporation and precipitation (Cao et al., 2012; Gregory et al., 2004; Kravitz et al., 2013b). As expected from these fundamental theoretical considerations, the combination of CO2 and absorbed sunlight sufficient to
FIGURE 3.3 Global mean precipitation and temperature changes in GeoMIP experiment G1, relative to the preindustrial state. Each symbol represents the results of an individual model in the ensemble. Results for the unmodified 4×CO2 state are also shown, for comparison. The G1 simulations were designed to restore global mean radiative balance, not global mean precipitation and evaporation (on the global mean, precipitation equals evaporation). If the goal were to restore global mean precipitation (and evaporation), the model simulations would have projected some residual warming. For each model, a linear fit (colored line) is derived from annual and global precipitation changes versus temperature changes between 4×CO2 experiment and 1850 conditions using the first 10 years of each simulation. SOURCE: Based on Tilmes et al., 2013.
restore the preindustrial value of global mean temperature reduces evaporation and precipitation relative to the preindustrial state. The amount by which evaporation and precipitation are reduced varies considerably from one model to another, and analysis of the mechanisms accounting for the intermodel spread is a subject requiring further research. One could in principle aim to compensate for CO2-associated temperature changes or precipitation changes (or some combined metric) but one could not simultaneously eliminate both global mean temperature changes and global mean precipitation changes (Ban-Weiss and Caldeira, 2010).
The climate system’s response to the joint effects of an increase in CO2 and a decrease in absorbed solar radiation is complicated by the land-sea contrast. Changes in the hydrological cycle over land are strongly affected by the land’s smaller and varied heat capacity, flow driven by terrain changes, and albedo variations, driving complex circulation changes that transport moisture to and from land masses. Figure 3.4 shows the pattern of changes in the hydrological cycle in the GeoMIP G1 ensemble simulations. Albedo modification affects both precipitation (shown in the top row) and evaporation (shown in the middle row). Net atmospheric water vapor transport to specific locations equals the balance of precipitation minus evaporation (P – E, shown in the bottom row of Figure 3.4).
The precipitation changes shown in Figure 3.4 are regionally inhomogeneous. Reduction in sunlight reduces the CO2-induced increase in extratropical precipitation. These albedo modification simulations were performed with the goal of offsetting top-of-atmosphere radiation imbalance and not for optimizing hydrologic quantities. Because the contour interval was chosen so as to reveal the global pattern, which is dominated by high precipitation and high precipitation changes in the tropics, this figure does not characterize the residual extratropical precipitation anomaly prevailing in the albedo-modified case (upper right panel). Globally averaged root-mean-square (RMS) changes in annual mean precipitation at model grid scale caused by high CO2 levels are reduced by about 55 percent in these albedo modification simulations; over land, these RMS changes in precipitation are reduced by about 50 percent (Kravitz et al., 2013b).
The tropics primarily exhibit reductions in precipitation, except for a narrow strip over the Pacific Ocean. The ensemble of models robustly shows less precipitation and evaporation over the Amazon basin (relative to preindustrial levels), but there is substantial disagreement as to the sign of the precipitation response over Africa. Given that the unmodified high-CO2 state also shows considerable regions of reduced precipitation, the situation could be crudely summarized by saying that a globally uniform reduction in sunlight is better at eliminating CO2-induced increases in precipitation than it is at eliminating CO2-induced reductions in precipitation. In these simulations, sunlight
FIGURE 3.4 Average hydrology changes produced from a quadrupling of CO2 (left column) and the result of a reduction in sunlight sufficient to return the global average surface temperature to a reference state (approximately preindustrial state; right column). Results are for all-model ensemble annual average hydrology differences (mm/day) averaged over years 11-50 of the simulation. Top row shows precipitation, middle row shows evaporation, and bottom row shows precipitation minus evaporation. Stippling indicates where fewer than 75 percent of the models (for this variable, 9 out of 12) agree on the sign of the difference. SOURCE: Kravitz et al., 2013a.
reduction reduces the amount of change in precipitation caused by high CO2 levels, but the pattern of reduced precipitation zones in the sunlight-modified state still differs appreciably from that in the unmodified high-CO2 state.
Figure 3.4 (middle row) also shows that in the sunlight-modified state evaporation decreases as well as precipitation, especially over land. The changes in evaporation have a spatial pattern that is similar to that for precipitation changes, but they are opposite in sign (Kravitz et al., 2013a). Globally averaged RMS changes in annual mean evaporation at model grid scale are reduced by about 39% in these albedo modification simulations; in contrast, over land, RMS changes in evaporation increase by about 10% (Kravitz et al., 2013b), largely due to the biophysical effects of CO2 resulting in the reduced evaporation from land plants no longer being offset by the acceleration of hydrological cycle response to warmer temperatures. The tropics exhibit widespread areas with reductions in evaporation, including over the Amazon basin and central Africa.
The change in precipitation minus evaporation (shown in the bottom row) provides an indication of the change in the amount of moisture imported to land areas, which in steady state is equal to runoff in rivers and streams. Areas in which there is substantial runoff are usually places where there is sufficient soil moisture to maintain plant life. Over much of the land area evaporation changes approximately equal changes in precipitation, with a few exceptions (e.g., drying in some parts of the Amazon and moistening in some parts of Africa). Larger shifts in the net moisture supply are seen in the unmodified high-CO2 state over even broader areas; thus, in these simulations, the sunlight reduction reduces but does not eliminate these effects of high CO2 concentrations. Over the ocean, precipitation minus evaporation is the difference of two large numbers, each subject to modeling challenges; the residual has large associated uncertainties. Over land, maximum evaporation is bounded by precipitation, so the model can be thought of as predicting the fraction of precipitation that evaporates, which is a number ranging from 0 to 1. Over much of the land, absolute magnitudes of changes in precipitation minus evaporation are small, and thus there is considerable disagreement as to sign among the models in the ensemble. Nonetheless, the general implication is that regions experiencing a reduction in precipitation do not necessarily become more arid; rather, the situation could be described as the hydrological cycle spinning down by 5 percent to 10 percent (Figure 3.2), with less rain falling but less rain evaporating back into the atmosphere. Globally averaged, albedo modification decreased the RMS difference in annual mean precipitation minus evaporation at grid-scale resolution by about 66 percent relative to the high-CO2 case without albedo modification; over land, albedo modification reduced RMS differences in precipitation minus evaporation by about 53 percent, despite the fact that these simulations were not designed to optimize the reduction in water delivery to land (cf. Ban-Weiss and
Caldeira, 2010). More research is needed to evaluate the impact of this altered climate state on agriculture, natural ecosystems, and water resources.
Because land responds quickly to insolation changes, the response of the seasonal cycle to albedo modification is expected to be different over land versus ocean, leading to changes in the seasonal cycle of the land-sea contrast which may affect precipitation patterns through their influence on atmospheric circulations, especially in the tropics. Additionally, even when the land-sea temperature contrast approaches equilibrium, the land surface has a tendency to cool more than the ocean (Joshi et al., 2013). When the land surface cools more than the ocean, this tends to cause air masses to ascend less rapidly or descend more rapidly over land and vice versa over the ocean, which would tend to weaken summer monsoonal circulations and thus contribute to a reduction in precipitation over land in response to deployment of albedo modification (Cao et al., 2012). This tendency toward weakening of the monsoons is in the opposite direction of a similar tendency for CO2-induced warming to strengthen monsoons, but the two effects do not precisely cancel out.
Figure 3.5 shows the response of the monsoon precipitation and evaporation in various regions to CO2 with and without sunlight reduction at a level that fully offsets the top-of-atmosphere energy balance from increased atmospheric CO2. The figure confirms that increased atmospheric CO2 concentrations tend to increase the strength of the monsoons, and that albedo modification has the tendency to reduce monsoon strength. These model results indicate that albedo modification at this level may overcompensate monsoonal strength, leaving some monsoons weaker than, but closer to, the preindustrial state (particularly over land) than the world without sunlight reduction. Albedo modification often produces evaporation changes that are similar in magnitude but opposite in sign to precipitation changes (Figure 3.1). Thus, in the albedo-modified GeoMIP simulations, no significant change in precipitation minus evaporation is seen relative to the preindustrial control in most monsoon regions, despite the fact that these simulations were not optimized to achieve this objective (Kravitz et al., 2013b). If the goal were to restore monsoonal strength to a level optimized to match a preindustrial world, the albedo modification may need to be applied at a reduced level, which would likely leave some residual global warming. There is considerable regional disparity in the monsoon response among models, making it difficult or impossible to tune the sunlight reduction strategy so as to optimize the response in all regions. There is also a large spread in predictions of monsoon response. At this level of sunlight reduction, the Indian monsoon response in the albedo-modified state ranges from about a 10 percent increase to about a 15 percent decrease. This underscores the difficulty of predicting monsoon response with the current state of modeling—with or without taking albedo modification into account.
FIGURE 3.5 Impact on the monsoon precipitation and evaporation for seven different regions around the globe. The left two values for each region show the perturbation due to a 4×CO2 increase. The right two values show the effect of albedo modification at a level designed to balance the global top-of-atmosphere energy flux. The length of the whiskers indicates the uncertainty in model response. SOURCE: Tilmes et al. (2013).
The next step up in the hierarchy of complexity is to simulate stratospheric aerosol injection, assuming the stratospheric aerosol optical depth (essentially, a measure of the mass concentration of aerosol particles in the stratosphere) to be horizontally uniform and simply increasing it to produce a negative forcing sufficient to counter the CO2 forcing. This approach is also fairly simple to implement in a wide range of climate models. Other studies, still quite idealized, rescale an externally calculated stratospheric aerosol optical depth, incorporating the effect of inhomogeneity of aerosol distribution, evolution of the particle size, and geographical distribution of aerosols. These idealizations do not account for feedbacks due to changes in stratospheric chemistry, but they do allow for the incorporation of at least some effects of stratospheric heating and a latitudinally and seasonally varying aerosol forcing. An extensive set of simulations of this sort is reported by Ricke et al. (2010, 2012), though these studies did not specifically analyze the effects of stratospheric heating. The results are broadly consistent with the GeoMIP study with regard to the pattern of temperature change and reduction in precipitation, but Ricke et al. (2010) analyzes a broader range of albedo modification magnitudes than was considered in GeoMIP. That study found that, when greater amounts of albedo modification were applied to offset the warming from higher CO2 concentrations, the regional deviations in temperature and precipitation from the preindustrial climate became more pronounced, but in almost all places the changes were much reduced relative to the high-CO2 state in the absence of albedo modification. There were also substantial differences in the character of the climate deviation from the preindustrial state, even between regions as close as India and China (Figure 3.6) projected in these single-model simulations. (This is not the case in many simulations performed with the more idealized solar-constant protocol using many models seen in the lower left panels of Figure 3.4.) The range of albedo modification magnitudes covered in this simulation serves as a reminder that it is possible to choose different targets than simply restoring global mean temperature to its preindustrial value. For example, a small amount of albedo modification would bring the climate state of India and China closer to the preindustrial origin of Figure 3.6. In the earlier (lower-CO2) case, it would be possible to choose a midrange amount of albedo modification, which would restore the temperature in China to its preindustrial value, while leaving the global mean warmer than preindustrial levels. However, this choice still leaves the precipitation in China lower than preindustrial levels, the temperature in India cooler than preindustrial levels, and the precipitation in India higher than preindustrial levels.
FIGURE 3.6 Modeled response to different levels of average global albedo modification over time in India and China. Interannual-variability-normalized regional temperature and precipitation summer (June, July, and August) anomalies (averages for the 2020s minus the 1990s, and the 2070s minus the 1990s) in units of baseline standard deviations for the region including India (triangles) and the region including eastern China (circles). Albedo-modified climates for these two regions migrate away from the baseline in disparate fashions. SOURCE: Modified from Ricke et al. (2010).
Because CO2 is removed from the atmosphere only slowly by ocean uptake and other geological processes, its climate forcing persists for millennia even if emissions cease, and the multimillennial influence becomes stronger as the cumulative amount of CO2 emitted increases (Archer et al., 2009; NRC, 2011b; Solomon et al., 2009). Theoretically, it may be possible to withdraw this CO2 from the atmosphere with carbon dioxide removal (CDR) technologies, but there are currently technical and economic barriers to implementation on a large scale (see companion volume Climate Intervention: Carbon Dioxide Removal and Reliable Sequestration).
In contrast to the long lifetime of CO2 in the atmosphere, the atmospheric lifetime of substances that have been proposed for use in albedo modification are on the order of a year or less (as discussed in detail later in this chapter). Therefore, although it takes relatively little mass of injected aerosol particles (or precursor gases) to cause an albedo change sufficient to offset the radiative forcing due to a doubling or even
quadrupling of CO2, that aerosol mass would need to be renewed more or less continuously, as long as an offset for CO2 forcing was intended.
One can imagine a large number of scenarios in which albedo modification might be deployed (e.g., MacMartin et al., 2014; Wigley, 2006), ranging from ones involving a short-term deployment to ones that require maintenance for millennia. The duration of the deployment affects the kind of climate risks that can be addressed.
Deployment of CDR could help provide an exit strategy within timescales as short as a century or so, for a broad range of albedo modification strategies, but, as mentioned above, deploying CDR at such scales is very challenging. Unless accompanied by CDR, albedo modification strategies whose goal is to limit peak warming in the absence of early and stringent emissions reduction would likely require that the deployment be maintained over the span of time required for natural processes to remove sufficient amounts of CO2 from the atmosphere (which, for high CO2 concentrations, could be millennia) or risk returning to the undesirable climate conditions that prompted the deployment initially. The class of risks associated with long-term reliance on albedo modification in this class of deployments can be called millennial dependence risk.
The Royal Society (2009) assessment rejected strategies requiring millennial dependence, finding that “[b]ecause of uncertainties over side effects and sustainability [albedo modification techniques] should only be applied for a limited period and accompanied by aggressive programmes of conventional mitigation and/or CDR, so that their use may be discontinued in due course” (Royal Society, 2009, Recommendation 3.3). To illustrate some issues associated with deployments aimed at permanently avoiding CO2-induced warming, and to bring the timescale issue into sharper focus, the committee considers the examples of climate intervention proposals aimed at offsetting the long-term warming due to CO2 emissions in the extended representative concentration pathway (RCP)4.5 and RCP6.0 emissions scenarios (Zickfeld et al., 2013). RCP4.5 assumes fairly aggressive emissions controls, though not quite sufficient to keep warming under 2°C; RCP6.0 assumes less restrained emissions. The top panels in Figure 3.7 show the CO2 radiative forcing in the two RCP scenarios. In both scenarios, the rate of CO2 emission peaks on or before the year 2100, the rate of CO2 emission declines sharply thereafter in such a way as to keep concentration fixed for the next 200 years, and emissions cease entirely by the year 2300. Substantial amounts of radiative forcing persist for many centuries after the cessation of emissions. The combined green- and red-shaded regions in the top panels show the amount of radiative forcing that would need to be offset by albedo modification in order to keep the net radiative forcing below 2.5 W/m2, which is approximately what would need to be done in order to keep the CO2-induced warming under 2°C, assuming a midrange climate sensitivity;
FIGURE 3.7 Radiative forcing and climate intervention commitment time. Results in the left column are for the RCP4.5 emissions scenario, while results in the right column are for RCP6.0, which involves higher emissions. The top row shows the CO2 radiative forcing for the two scenarios (thick black line), based on Zickfeld et al. (2013). The solid red line in these panels depicts a target radiative forcing to be achieved by albedo modification, designed to never exceed 2.5 W/m2, whereas the dashed green line corresponds to a strategy in which the radiative forcing is only subject to a cap for the first 75 years, whereafter albedo modification is gradually phased out over the next 75 years. The shaded green region indicates the period of time over which albedo modification is applied in the limited-duration phase-in/phase-out strategy, while the red shaded region shows the duration of commitment in the permanently capped strategy. The middle row shows the corresponding amount of radiative forcing change that albedo modification needs to accomplish, as a function of time. The bottom row translates the radiative forcing trajectories of the top row into global mean temperature using the simplified energy balance model described by Pierrehumbert (2014). The limited-duration albedo modification strategy shown by the green dashed lines is able to slow down the initial stages of the warming but does not visibly reduce the peak warming or delay the time at which the peak warming is attained.
these estimates do not take into account the possible effects of albedo modification on the carbon cycle (see section “Modeled Climate System Responses to SAAM” below). The middle panels show time series of the amount of reduction in solar radiation that would be needed to achieve the target climate and provide an indication of the level of albedo modification effort required over time. Even in the lower emission scenario—for which the unmodified climate exceeds the 2°C target by a small amount—to permanently avoid CO2-induced warming, the climate intervention actions would need to be maintained to nearly the year 2700. To achieve this goal for the RCP6.0 emissions scenario, albedo modification efforts would need to be maintained at a substantial level even in the year 3000, and it would in fact be several thousand years more before the CO2 radiative forcing decays to the point that climate intervention could be terminated without a substantial temperature increase. In a situation where the amount of CO2 emissions mitigation accomplished has proved insufficient to avoid crossing a temperature target on the order of 2°C (or similar), meeting such a target by means of albedo modification would require a millennial or even multimillennial deployment to actively maintain climate intervention without interruption, unless techniques to greatly accelerate CO2 removal from the atmosphere (CDR) are deployed at very large scale. All the extended RCP scenarios used in this calculation assume that anthropogenic CO2 emissions cease entirely by the year 2300 or earlier, implying that either CO2 emission mitigation eventually becomes effective or that the supply of fossil fuel runs out.
Without a near-millennial or longer deployment of CDR, albedo modification could delay but not avoid the crossing of a temperature threshold (MacMartin et al., 2014; Wigley, 2006). By itself albedo modification would only temporarily delay warming, unless the albedo modification effort was continually maintained over the period of substantial excess atmospheric CO2 concentrations, which is anticipated to last millennia. Delaying warming could be useful if the additional time allowed measures to adapt to the eventual warming to be put into place or allowed deployment of CDR methods. It may also be useful in addressing climate damages tied to the rate of warming, though reliance on albedo modification may also introduce risk of making such damages worse if CO2 concentrations are increasing while it is deployed, and the albedo modification is prematurely and abruptly terminated.
The green dashed curves and green shaded regions in Figure 3.7 give an example of a strategy whose goal is to delay, rather than prevent, warming. These provide examples of what can be accomplished with a short-duration deployment. Specifically, the albedo modification follows the same trajectory as the millennial case for the first 75 years (allowing for a gradual phase-in of the procedure), whereafter it is phased out over the next 75 years. The ramp strategy achieves a 25-year delay in the time of
crossing of a 2°C warming threshold in the lower-emission case, and a 20-year delay in the higher-emission case. Smith and Rasch (2012) have explored options for century-scale deployments. Limited-duration deployments might be useful if stringent emission controls have kept CO2 emissions to relatively low levels, when additional time is needed for adaptation, or if significant negative emissions (CDR) are possible.
Because air, land, and the upper ocean respond quickly to changes in radiative forcing, an abrupt termination of albedo modification would result in rapid warming, with global mean temperatures rising within a decade or two to levels close to what would have been experienced without albedo modification (Jones et al., 2013; Matthews and Caldeira, 2007). The possibility of rapid warming is a novel and potentially severe risk not present in the unmodified high-CO2 state, in which temperature increases more slowly over time. As a result, the choice of a climate future in which a high CO2 concentration is compensated by a high degree of albedo modification risks putting Earth’s climate in a precarious state. Phasing albedo modification in or out over many decades, such as might be done to give human and natural systems a chance to better adapt to the resulting temperature change (MacMartin et al., 2014; Wigley, 2006), would reduce the time span over which Earth was subject to termination risk, but an abrupt termination risk will always be present if albedo modification is being used to counter a substantial fraction of the CO2 forcing.
The climatic impacts of abrupt termination were specifically considered by Matthews and Caldeira (2007), Brovkin et al. (2009), Llanillo et al. (2010), and Jones et al. (2013), but rapid post-termination warming was also confirmed by Robock et al. (2008), Jones et al. (2010), and Berdahl et al. (2014), and there are no simulations of abrupt termination that conflict with these predictions of rapid warming. The upper panel of Figure 3.8 shows the warming upon termination in a series of GeoMIP solar-constant simulations (Jones et al., 2013) of the response to increasing CO2 at a rate of 1 percent per year, offset by reduction in solar radiation that is terminated abruptly at year 50. As noted in that study, the inclusion of realistic aerosol effects would not substantially change the rapidity of the warming, because aerosols disappear within 1 to 2 years from the stratosphere. The lower panel, from Llanillo et al. (2010), shows that very similar results are obtained from a highly simplified energy balance climate model and also illustrates that the longer sunlight reduction is used to offset continually increasing CO2, the larger the effect that is caused by termination.
The amount of warming following termination depends on the climate sensitivity—a quantity that is highly uncertain for the actual climate and which varies significantly among models. It is difficult to infer climate sensitivity from observations of a warming climate without albedo modification, and it would be more difficult to do so in a cli-
FIGURE 3.8 Multimodel results for simulation of abrupt termination of albedo modification in GeoMIP solar-constant experiments (Jones et al., 2013). Dashed lines show the climate response to the increasing CO2 without reduction in solar radiation.
mate subject to strong (and possibly uncertain) albedo modification. Hence, it would be difficult for inhabitants of a strongly albedo-modified high-CO2 world to know in advance what magnitude of climate change they would face upon abrupt termination (Matthews and Caldeira, 2007).
Both Jones et al. (2013) and Berdahl et al. (2014) confirm that the rapid warming is accompanied by a rapid loss of sea ice, particularly in the Arctic. Jones et al. (2013) and McCusker et al. (2014) point out that upon abrupt termination some regions simultaneously experience rapid warming and rapid precipitation decreases, increasing the stress on arid regions (though there is only weak consensus among models as to where the stresses are the highest).
Abrupt termination could lead to significant ecosystem, agriculture, and societal impacts that would not have existed had albedo modification never been deployed, but these potential impacts are largely unknown at this time. If the consequences of warming due to CO2 were severe enough to trigger an emergency deployment of albedo modification and no effective adaptation effort was put in place during the period of deployment, it is likely that an extremely rapid onset of a warming of the same magnitude would have even more severe consequences. Few studies so far have specifically addressed the impacts of post-termination warming. Xia et al. (2014) conclude that an abrupt termination would have a negligible effect on rice production in China. They also find that an abrupt termination would reduce maize production
by about 12 percent relative to levels that were achieved during the albedo modification deployment, but these yields are still higher than would have been obtained in the preindustrial condition without CO2 fertilization of land plants. Jones et al. (2013) do not find any marked effect of either albedo modification or abrupt termination on the multimode mean global net primary productivity (which is often taken as a rough proxy for agricultural and ecosystem impacts) despite the massive and rapid climate change; this is difficult to reconcile with robust indications of food security issues in a warming world (e.g., NRC, 2011). However, there is considerable disagreement among the individual models of Jones et al. (2013) as to the baseline net primary productivity, the response to unmodified global warming, and the response to abrupt termination of albedo modification. Furthermore, the multimodel ensemble mean global net primary productivity shows a steady increase even in the control run in which the world warms in response to increasing CO2 without offsetting by albedo modification. Although the mechanism of this increase was not diagnosed, it would be consistent with a dominance of CO2 fertilization effects when interpreted in conjunction with the minimal effect of albedo modification termination on net primary productivity. If so, this is a source of concern requiring further inquiry, because the CO2 fertilization effect in land ecosystem models is very model dependent and subject to considerable uncertainties (Rosenzweig et al., 2014). Overall, there is need for a better understanding of the effects of albedo modification and its abrupt termination on agricultural and natural ecosystems.
The risk of severe impacts of abrupt termination increase with the magnitude of albedo modification deployed. In particular, if CO2 emissions continue during the time over which albedo modification is deployed, and are canceled out by increasing the amount of albedo modification, then the severity of impacts of abrupt termination will steadily increase. It is in futures where CO2 is very high or climate sensitivity turns out to be high that albedo modification is most likely to provide benefits, leading McCusker et al. (2014) to conclude: “We are left with the disconcerting situation in which [albedo modification] is most useful precisely when its associated risks are the greatest.” An unmodified, hot, high-CO2 climate also incurs serious risks. Determining the circumstances under which these risks should be traded for the risks of abrupt or more gradual termination is a challenging problem, which the committee does not address. The surest way to minimize risks of both sorts is to continue and expand efforts to mitigate CO2 emissions, which would minimize the amount of climate change with which any eventual albedo modification would need to cope.
There are many technologies that humanity already relies on which could cause substantial harm if their use were to cease abruptly. However, human history offers no precedent for the maintenance over a millennial timescale of a technological interven-
tion of sophistication and global scope comparable to albedo modification. Further research would be useful to ascertain the ability of society to sustain albedo modification over such a long timescale in the face of other societal, political, and ecological challenges.
Climate intervention using realistic strategies involves atmospheric injection of aerosols or aerosol precursors. Aerosols (solid or liquid particles suspended in the air) of natural and anthropogenic origin are found everywhere in the atmosphere. They affect the planet’s energy budget by scattering and absorbing sunlight, and by changing cloud properties (Seinfeld and Pandis, 2006). They also play a role in the chemistry of the atmosphere and carry nutrients and disease from place to place. Humans have changed the amount of aerosols in the atmosphere through pollution emissions, and by changing natural aerosol sources through land and water use. Aerosols that originate directly from a source (e.g., dust, soil, smoke particles from fires, and bacteria or viruses) are generally called “primary aerosols.” Aerosols that develop from gases (natural and anthropogenic) that condense into a liquid or solid form (e.g., particles containing sulfate, nitrate, and organic carbon) are often called “secondary aerosols.” Aerosols are mostly removed from the atmosphere by dry deposition, sedimentation, or scavenging by clouds.
Aerosol particles higher in the atmosphere are not removed as quickly as those near the surface. Aerosols found high in the atmosphere have a longer lifetime1 than those found near the surface because they are far from clouds and the surface where they would be removed on very short timescales (days).
Aerosols interact with sunlight passing through Earth’s atmosphere. When aerosols scatter sunlight back to space they cool the planet; when they absorb sunlight they warm the air locally but can cool the atmosphere below them. The best estimates of the net effect of atmospheric aerosols are that they cool the planet. One of the broad classes of proposed techniques for altering the Earth’s energy balance involves increasing the number of aerosols in the stratosphere (a layer with a base called the tropopause between about 8 and 18 km above the surface, extending to about 50
1 Scientists usually refer to the average time a particle resides in the atmosphere in terms of a “lifetime” or “residence time” where lifetime is defined as the time required for the concentration of a substance to be reduced by a factor to 1/e times the original concentration.
km). Theory and models suggest that increasing the number of aerosols that scatter sunlight back to space will cool the planet. Scientists have considered deliberately introducing aerosols into the stratosphere primarily because aerosols have a much longer lifetime in the stratosphere (on the order of years) compared to lower altitudes (where lifetimes are on the order of days to weeks). Producing or injecting aerosols in the stratosphere would minimize the amount of aerosols needed to produce a specified amount of cooling because the same amount of aerosols would stay in the atmosphere longer and produce more cooling than at lower altitudes.
Both scattering and absorbing aerosols will reduce sunlight reaching the surface of the planet. A range of aerosols has been considered for modifying the energy budget of the planet (see below section, “Proposed Mechanisms for SAAM”). Most of the methods that propose to use stratospheric particles to cool Earth are likely to produce similar characteristics with regard to their effects on global mean surface temperature and precipitation, but they can differ in important regards with respect to the amount of stratospheric heating they produce and their effects on stratospheric chemistry. The committee’s discussion focuses primarily on injection of sulfate aerosols or their precursors into the lower stratosphere. This is the most-studied technique and is also the one that most closely mimics the way large volcanic eruptions cool the climate.
Formation, evolution, and removal of stratospheric aerosols. Most stratospheric sulfate aerosols are formed as a result of transport into the stratosphere of natural and anthropogenic gases that contain sulfur originating nearer the surface (e.g., carbonyl sulfide, sulfur dioxide [SO2], and sulfuric acid [H2SO4]). Explosive volcanoes also inject SO2 into the stratosphere. These gases undergo a series of chemical reactions that add oxygen atoms to the source gas (through a process called oxidation) which eventually leads to the formation of H2SO4 in the gas phase. In the stratosphere H2SO4 can either nucleate to form new small particles or condense on existing particles, making those particles larger. Particles usually form near the tropical tropopause and some evaporate as they are lifted to higher altitude. Those remaining lower down near the tropopause eventually migrate toward the polar regions where they pass into the troposphere, either transported by the wind through midlatitude tropopause folds or by sedimentation. The average residence time of a particle in the lower stratosphere is approximately 1 year. After eventual transport into the troposphere, the particles undergo relatively rapid mixing processes by weather events, turbulence, and cloud-scale overturning. The aerosols are then rapidly scavenged (timescales of days to weeks) by acting as nucleation sites for cloud ice or liquid particle formation. These
processes are described in more detail in recent textbooks of stratospheric chemistry and summarized in a report on stratospheric aerosols (SPARC, 2006). Figure 3.9 shows a summary from that report of important processes in the lifetime of stratospheric aerosols.
Most of the sulfuric acid gas found in the stratosphere is formed by reaction of SO2 with the hydroxyl radical (OH), the main oxidant of the chemical reactions occurring there. The SO2 itself comes from (1) transport of natural and anthropogenic SO2 from the troposphere, (2) oxidation in the stratosphere of gaseous precursors (natural and anthropogenic), and (3) direct injection of SO2 by strong volcanic eruptions. Most observations (for Pinatubo) and models are consistent with a lifetime for SO2 of order 30 to 35 days (Liu and Penner, 2002; Read et al., 1993). Nevertheless, for large volcanic eruptions, the OH concentration may not be constant but may decrease due to a combination of increased water vapor flux, decreased incident solar radiation, and possibly heterogeneous reactions (Robock et al., 2009a). Modeling studies (Robock et
FIGURE 3.9 The life of natural stratospheric aerosols. The aerosol particles are formed by nucleation in rising tropical air and grow by condensation and coagulation as they are carried aloft. They eventually move to mid- and high latitudes where they may be removed by mixing across the tropopause. SOURCE: Hamill et al., 1997.
al., 2009a) that include coupled stratospheric chemistry find that the lifetime of any given molecule of SO2 is longer compared to studies without coupled stratospheric chemistry because the oxidation rate of SO2 is limited by the lack of reactants (see also Bekki et al., 1996).
There is a well-established theory for the formation (referred to as “nucleation”) of H2SO4 particles from sulfuric acid vapor in the presence of water vapor. This mechanism is thought to be the primary mechanism leading to new particle formation in the stratosphere, although ion-induced nucleation may also play a role (Arnold et al., 1982; Campbell et al., 2014). For a given addition of SO2, the trade-off between new particle formation (leading to more but smaller particles) and coagulation and condensation (leading to larger particles) depends on the temperature and ambient concentrations of gaseous sulfuric acid and preexisting sulfate particles (number and size), mediated by the size of the SO2 concentrations that produce the sulfuric acid (Timmreck, 2012). The concentration of the gases and aerosols that govern these processes is determined by chemical reactions, physical processes (like Brownian motion and particle sedimentation, to name only a couple of processes), and molecular, turbulent, and larger-scale mixing by the winds that govern the aerosol and gas concentrations. Although the basic physics and chemistry that describe new particle formation, condensation of gases on existing particles, particle evaporation, and the coalescence processes that reduce particle number and increase particle size are well understood, subtle details matter a lot in determining the evolution of particle number and mass, and the subsequent role of those particles in the climate system. More work is needed in characterizing these processes in nature (through measurements) and in modeling (through better model treatments and a careful comparison with observed features of aerosols and their precursor gases) before scientists can produce truly accurate models of stratospheric aerosols and their effects on climate.
The effectiveness of possible mechanisms for introducing sulfate aerosols into the stratosphere—injecting SO2 gas that oxidizes to H2SO4—is determined by stratospheric chemistry and transport patterns. There have been some initial studies on this (see below section, “Model Estimates of Aerosol Forcing from SAAM”), but this is still an area that requires substantial research.
Impacts of stratospheric aerosols on climate. Stratospheric sulfate aerosols scatter and absorb sunlight, and they also absorb and emit energy at infrared wavelengths. Their radiative impact depends on the particle size. They are primarily scatterers of sunlight at typical sizes found in the stratosphere, and thus cool the planet, but they can also contribute to local heating of the atmosphere. Even purely absorbing particles in the
stratosphere have a cooling influence on Earth’s surface despite having a heating influence on the stratosphere because the absorbing particles block some of the sunlight that would otherwise reach the surface (Ban-Weiss and Caldeira, 2010). Stratospheric aerosols change the amount of sunlight passing downward through the tropopause and thus have climate effects such as those discussed in the idealized studies above.
Stratospheric aerosols provide sites for heterogeneous chemistry, and some of that chemistry can lead to ozone depletion. Thus, changes in stratospheric aerosol can also affect climate indirectly, by influencing ozone. Ozone is a critically important atmospheric constituent (see WMO  and IPCC [2013a] for modern and comprehensive reviews) in the Earth system. It is one of the major oxidizing agents of the atmosphere, and it participates in many important chemical reactions. Ozone absorbs and emits energy in many parts of the energy spectrum, and its absorption of sunlight produces a notable warming in the stratosphere. It is also a greenhouse gas, absorbing and emitting energy at infrared wavelengths. The heating and cooling produced by ozone change can thus drive circulation changes (IPCC, 2013a; WMO, 2011). Ozone also absorbs light in the ultraviolet region of the energy spectrum (hereafter called UV-B light). Since stratospheric aerosols also scatter UV-B light, reducing the amount reaching the surface, there is the potential for the compensating changes between ozone loss (which will increase surface UV-B) and increasing aerosols (which will decrease surface UV-B) in the total change. The amount of UV-B light reaching the surface has significant implications for surface ecosystems and human health. Increases in surface UV-B light would be expected to lead to increases in skin cancer in humans (see, e.g., McKenzie et al., 2011; Rogers et al., 2010; Stern, 2010).
In scattering sunlight, stratospheric aerosols reduce the direct beam of sunlight and also increase the ratio of diffuse to direct sunlight reaching the surface. This means that while less sunlight reaches the surface (cooling the planet), the light tends to come from more directions, so it penetrates into plant canopies more effectively, exposing more leaves to light, which has impacts on photosynthesis and makes shadows less sharp. Reducing total light reaching the surface tends to reduce light available for photosynthesis, but increasing the diffuse light allows plant canopies to photosynthesize more efficiently. Changing photosynthetic activity can change plant productivity and the capacity of plants to act as a carbon sink. Measurements following the Pinatubo eruption indicate that plant productivity and carbon sink went up (Gu et al., 2003), suggesting that the increase in diffuse light is more important to plant growth than the decrease in the sunlight reaching the surface. The heating and changes to ozone associated with increased stratospheric aerosols can also affect tropopause temperatures with consequent effects on water vapor input to the stratosphere. The added water in the stratosphere affects the climate of the stratosphere and strato-
spheric chemistry, with additional implications for surface climate (Heckendorn et al., 2009). High clouds may be influenced by stratospheric aerosols (Box 3.2).
There are many factors that influence the interactions between stratospheric aerosols and ozone. The chemical interactions generally involve the presence of inorganic chlorine, water vapor, and sulfate aerosols, as noted in a series of studies (Anderson et al., 2012; Drdla, 2005; Drdla and Müller, 2010; Hanisco et al., 2007; Homeyer et al., 2014; Peter and Grooß, 2012; Sayres et al., 2010; Schwartz et al., 2013; Shi et al., 2001; Solomon, 1999), along with the convective injection of compounds from the boundary layer (Hanisco et al., 2007; Pittman et al., 2007; Salawitch et al., 2005; Weinstock et al., 2007).
Increases in stratospheric aerosols might alter the radiative balance and chemistry of the stratosphere, and the Earth system more broadly. These are areas of active research, and recent studies on these topics are described in the below sections (“Observations and Field Experiments of Relevance to SAAM,”“Modeled Climate System Responses to SAAM,” and “Environmental Consequences of SAAM”).
BOX 3.2 EFFECTS OF AEROSOLS ON CIRRUS CLOUDS
Cirrus clouds are high-altitude ice clouds. Thick cirrus clouds have a net negative impact on radiative forcing (Kubar et al., 2007), cooling by reflecting sunlight back to space and warming by trapping outgoing infrared energy through a greenhouse effect. Radiative forcing by thin cirrus clouds is dominated by the greenhouse effect that produces a net positive forcing tending to warm the climate. Observations indicate the net impact of high cirrus clouds is to warm the planet, but the effect of the addition of aerosol particles on this net impact is complex to predict. The net effect of high clouds is a small residual of two large numbers, both of which depend on microscopic cloud properties, and is therefore very difficult to model. Change in the number and size of cloud particles affects cloud lifetime and the balance between the infrared and solar effects of the clouds.
As stratospheric aerosol particles mix into the troposphere, they may influence cirrus clouds in at least two ways. First, they can influence the very complex balance between homogeneous and heterogeneous nucleation processes that produce cirrus ice crystals. The effects depend on both the size of the particles transported to this region from the stratosphere, and the ambient particles, by changing the relative importance of the heterogeneous and homogeneous ice nucleation in the region. It is not clear how cirrus clouds would change if stratospheric aerosol increases were to occur (Cziczo et al., 2013; Froyd et al., 2010). Most model simulations have assumed that homogeneous ice nucleation dominates in cirrus clouds, but there are clearly regions where heterogeneous nuclei are numerous enough to alter this assumption. Second, the radiative heating occurring in the region of stratospheric aerosols can change the stability of the upper tropospheric layers, affecting the vertical velocities that are important to ice crystal formation.
No well-documented field experiments involving controlled emissions of stratospheric aerosols have yet been conducted. Some volcanic eruptions have injected large amounts of sulfur dioxide gas into the stratosphere, and observations of these eruptions and their impact on climate can serve as natural experiments for testing our understanding of albedo modification processes (Robock et al., 2010, 2013). The observed cooling following large eruptions provided much of the initial stimulus for the idea that albedo modification could help offset effects of warming due to anthropogenic CO2 increase, and attempts to model the observed effects of volcanic eruptions can provide some insight into the complexity of the processes and some of the unknowns that still need to be addressed. The climate effects of a single pulse of aerosols such as that produced by volcanoes would differ in important ways from the effects of a sustained effort to maintain a persistent aerosol layer (Box 3.3). Nonetheless, volcanoes provide an excellent opportunity to test and improve our understanding of relevant physical processes. However, there are many challenges and limitations associated with the use of volcanic eruptions as analogues for SAAM, which are discussed in Appendix D, but they do represent the only feasible large-scale experiments (natural or otherwise) in stratospheric attenuation of a large fraction of solar energy. As such, they offer our best opportunity to develop insights into SAAM. Moreover, as “events of opportunity,” they do so without introducing substantial and risky human perturbation to the climate system.
Very large eruptions—the size of El Chichón (1982) or Pinatubo (1991)—produce a detectable climate response that can be used to test simulations of both aerosol forcing and the consequent response of climate, but even smaller eruptions—the size of the Sarychev eruption (2009)—can provide a useful test of our ability to observe and to simulate stratospheric aerosol processes (Kravitz and Robock, 2011; Kravitz et al., 2010, 2011b). Large eruptions can also serve as a test of the effect of increased particle surface area on ozone destruction, of our ability to model the associated atmospheric chemistry, and of other impacts.2 The effect of large volcanic eruptions on Earth’s radiation balance can persist for several years before their concentrations return to background values.
To indicate what is known about stratospheric aerosol effects on the planet, the committee focuses here on the 1991 eruption of Mount Pinatubo, because scientists have the best observational data for it. The Pinatubo eruption, on June 14-16, 1991, injected
2 Other eruptions, such as Tambora in 1815, caused global climatic anomalies that led to widespread crop failure and famine (Oppenheimer, 2003).
BOX 3.3 ARE VOLCANIC ERUPTIONS GOOD ANALOGES FOR STRATOSPHERIC AEROSOL INJECTION?
The short answer is yes and no. Volcanic eruptions that inject large amounts of sulfur dioxide gas into the stratosphere are believed to have much the same effect (at least initially) as proposed methods to engineer the climate by purposeful injection of stratospheric aerosols and, thus, can serve as natural experiments for testing our understanding of albedo modification processes (Robock et al., 2010, 2013). Indeed, it was the observed cooling following large eruptions that provided much of the initial stimulus for the idea that albedo modification could help offset effects of warming due to anthropogenic CO2 increase. Attempts to model the observed effects of volcanic eruptions have provided some insight into the complexity of the processes and some of the unknowns that still need to be addressed. In addition to blocking sunlight, the aerosols absorb incoming solar infrared and thermal heat from below, heating the stratosphere. Thus, the response to the volcanic eruption is not just cooling of Earth’s surface, but also reductions in rainfall over land and a winter warming pattern from the stratospheric heating. However, there remain discrepancies between models and observations that require improved ability to track the aerosol evolution and accurately reflect the radiative transfer that controls the stratospheric heating. Furthermore, there are several differences between volcanic eruptions and purposeful albedo modification that make the volcanoes imperfect analogues. Past eruptions have occurred under conditions of enhanced stratospheric chlorine and bromine concentrations and, thus, have incurred larger stratospheric ozone decreases than might be the case in the future. Eruptions are point-source releases of a range of particles, whereas any albedo modification would aim to produce a more spatially uniform distribution of more uniform aerosols. In addition, eruptions are short-lived phenomena, not lasting long enough to strongly affect, for example, ocean temperatures to the point of altering the heat and density transport processes that control ocean circulation. Because land temperatures respond more quickly than ocean temperatures, volcanoes cause more cooling over land relative to ocean than would be caused by a sustained aerosol layer; this would be expected to contribute to decreased precipitation over land following a volcanic eruption. Albedo modification would need to be maintained for a long time period, with lasting effects on ocean temperatures and circulation, ecosystems, sea ice, and other aspects of the climate system, producing feedbacks not seen to date in volcanic eruptions. See Appendix D for further discussion of the volcano analogy and Box 3.5 below for observational requirements for making better use of volcanoes as natural experiments.
14 to 26 megatons of SO2 into the stratosphere with concentrations peaking near 25 km and reaching as high as 30 km (Read et al., 1993). This was converted to H2SO4 particles over the next 1 to 2 months, with mode radii initially observed peaking at 0.1 μm radius, similar to background aerosol sizes, but developing a bimodal distribution having a distinct second peak near 0.5 μm by November 1992 which lasted until May 1993 (Goodman et al., 1994). Initial concentrations were more than 10 times higher than background.
The volcanic eruption took place in the western Pacific (15.1°N, 120.4°E) and the stratospheric sulfate aerosol plume was observed to extend from 20°S to 10°N after 100 days. After 150 days both SO2 and H2SO4 had reached beyond the tropics in both hemispheres (Read et al., 1993; Russell et al., 1996). Stratospheric concentrations of H2SO4 aerosols remained above background well into 1993. The optical depth of the total stratospheric aerosols had a lifetime (for reduction by a factor of 1/e) of around 1.5 years near 19.5°N (Russell et al., 1996).
Numerous changes were observed following the Mount Pinatubo eruption, including changes in temperatures. Figure 3.10 shows one estimate of the lower tropospheric temperature change following the Pinatubo eruption of 1991 by Soden et al. (2002). Other studies (Canty et al., 2013; Thompson et al., 2009) have estimated that the globally averaged surface air temperature reduction from Pinatubo is somewhat lower (0.2 to 0.4 K).
In addition to reflecting sunlight and changing surface temperature, there are many other impacts. For example, observed effects of large volcanic eruptions on the planet include changes to stratospheric ozone (O3) levels. Column ozone (O3) averaged over 60°S to 60°N decreased by about 4 percent following 1991, but changes in halogens (e.g., chlorine and bromine gases) were also responsible for some of this decline (Chipperfield et al., 2007, Fig. 3-21). Sulfate particles in the lower stratosphere provide surfaces for the chlorine to activate into forms that deplete ozone. Two-dimensional (Tie et al., 1994; WMO, 2003, Section 126.96.36.199) and three-dimensional (e.g., Chipperfield,
FIGURE 3.10 Comparison of the observed (black lines) and model-predicted (dashed lines) global-mean (90°N-90°S) changes in lower tropospheric temperature after the eruption of Mount Pinatubo. The observed anomalies are computed using a 1979 to 1990 base climatology and expressed relative to the pre-eruption value, defined here as the mean anomaly for January 1991 to May 1991 (MSU, microwave sounding unit; No ENSO, observations adjusted to remove the effects of El Niño–Southern Oscillation). The model anomalies are computed for each ensemble pair as the difference between the control and Mount Pinatubo experiments. All of these time series have been smoothed using a 7-month running mean. SOURCE: Soden et al., 2002.
1999, 2003; Stolarski et al., 2006) model studies have shown the chemical effects of volcanic eruptions and it is well known that the presence of enhanced particles in the stratosphere can cause significant ozone loss through heterogeneous chemical reactions, which was demonstrated by studies on Mount Pinatubo (WMO, 2003, 2011). The volcanic effect on column ozone results from heterogeneous catalytic conversion of HCl and ClONO2 to ClO which then, in combination with the hydrolysis of N2O5, titrates NOx from the system. As a result the dominant removal process for O3 is the rate-limiting step ClO + BrO → Cl + Br + O2 (Salawitch et al., 2005; Solomon et al., 1996). Thus, chemical ozone losses from volcanic sulfate injection are largest at times of peak chlorine and bromine, and volcanic impact on ozone at preindustrial halogen levels is estimated to be small or even positive (Tie and Brasseur, 1995).
Dynamical changes resulting from the Mount Pinatubo eruption also contribute to ozone change (Hadjinicolaou et al., 2005). Differences in the effects of Pinatubo between the Northern Hemisphere (NH) and Southern Hemisphere (SH) are not well understood. Models show a SH effect as large or larger than the NH effect, though this is not seen in data (Chipperfield et al., 2007). Stolarski et al. (2006) showed that such effects may be due to interannual variability.
Changes in precipitation following the 1991 eruption were also studied. Trenberth and Dai (2007) examined possible changes in precipitation and associated river runoff associated with the Pinatubo eruption. Global average precipitation decreased by 0.07 mm/day between late 1991 and early 1992 compared to the 1979-2004 average. Global average land precipitation during 1992 was about 10 percent (3.1 standard deviations) below normal while river discharge was also about 10 percent (3.7 standard deviations) below normal. However, this event is confounded by El Niño occurring during the same time period. After removal of El Niño effects on the time series (from 1950 to 2004) using regression, the natural variability in precipitation and runoff is reduced by almost 44 percent and 36 percent, respectively, and effects of Pinatubo stand out much less. However, the 1992 anomalies are still significant at the >99 percent confidence level.
Some studies have suggested that increased aerosol from Pinatubo produced an increase in stratospheric sulfate particles, leading to an increase in optically thick cirrus clouds (Minnis et al., 1993) and in cirrus cloud cover (Wylie et al., 1994), but, ultimately, observational analyses of the aerosol effect on cirrus clouds during Pinatubo are inconclusive, as pointed out by Robock et al. (2013): Ackerman and Strabala (1994) and Minnis et al. (1993) find changes, but Luo et al. (1997) do not. The effect of particles from volcanic eruptions on ice nucleation is still under investigation. Roderick et al. (2001) suggested that Pinatubo also increased the diffuse light entering plant
canopies, leading to increased photosynthetic activity and the capacity of plants to act as a carbon sink.
Numerous recent studies have highlighted the difficulty of simulating the observed evolution of stratospheric aerosols (Auchmann et al., 2013; Foley et al., 2014; Muthers et al., 2014; Thomason and Peter, 2006; Timmreck, 2012; Toohey et al., 2013; Weisenstein and Bekki, 2006), including aerosol size, amount, and location. Models also find it difficult to reproduce other effects on the Earth system, including the diurnal cycle of surface temperature, impacts on the carbon cycle, transport and deposition of aerosol to high latitudes, and changes to atmospheric dynamics (Auchmann et al., 2013; Foley et al., 2014; Toohey et al., 2013).
The ability of models to reproduce the observed signatures produced by volcanic eruptions therefore provides a real challenge to models and a necessary, but not sufficient, test of the ability of models to accurately simulate the processes important to climate and climate change associated with SAAM. Because volcanic eruptions occur relatively infrequently, and stratospheric aerosols return to background values within a few years, volcanic impacts do not persist. Since SAAM introduces a persistent source for stratospheric aerosols, and a persistent forcing, it may involve interactions in Earth system components that are not present following volcanic eruptions, so these simulations are only an incomplete test of the relevant interactions (see Box 3.3). Nevertheless, the simulations provide the single most stringent test of the processes relevant to SAAM available today. More comprehensive and thorough studies using existing observations and improved observations gathered from future eruptions would be extremely useful as testbeds for model evaluation and model improvement.
A somewhat broader perspective on the state of the art in volcanic response modeling is provided by Driscoll et al. (2012), which is the most comprehensive assessment to date of the ability of coupled ocean-atmosphere models to reproduce the winter (December-January-February) volcanic response. This study surveyed the volcanic response after nine different eruptions in all the CMIP5 models which included volcanic forcing. All of the models considered computed the ocean and sea ice response using a dynamic ocean circulation model coupled to the atmosphere, but out of the 13 models discussed, only one computed aerosol properties starting from the injection of sulfur dioxide instead of imposing aerosol characteristics based on observations. Despite the fact that most of the models had the advantage of constraining the aerosol properties based on observations, the ability of the models to reproduce the average winter temperature pattern and circulation response is very poor, although some of this lack of response may have been associated with modeled El Niño–Southern Oscillation (ENSO) frequencies, which cannot be controlled. The multimodel mean
simulation is dominated by a weak cooling, with very little evidence of the observed polar warming (Figure 3.11). This is due to the inability of most of the models to accurately reproduce the atmospheric circulation change forced by stratospheric heating (Driscoll et al., 2012). Based on these results, Driscoll et al. (2012) question whether existing modeling capabilities are adequate for assessing the impact of SAAM climate interventions, though it is also a possibility that the more spatially and temporally uniform aerosol layers that global SAAM schemes aim to achieve would pose fewer modeling challenges.
There is considerable variation in response among the models in the CMIP5 ensemble, and some models do better than others at reproducing some features of the winter response. Thomas et al. (2009a,b) performed a detailed analysis of the winter response to Pinatubo in the ECHAM-5 model. They found improved winter surface temperature responses using observed aerosol properties, specified sea surface temperatures, and quasi-biennial oscillation phase (see also Stenchikov et al., 2004). Nevertheless, some discrepancies between the modeled and observed response pattern remain (see especially Fig. 5 of Thomas et al., 2009b).
Budyko (1974) was the first to suggest a deliberate method to increase aerosols in order to increase planetary albedo by flying aircraft into the lower stratosphere and burning sulfur-bearing compounds. Since that time, a variety of mechanisms for delivering sulfur-containing species to the lower stratosphere have been suggested (Rasch et al., 2008a), including aircraft, rockets, artillery, and pipes elevated to high altitudes carrying aerosol precursors.
In addition, a variety of types of particles have been suggested for introduction into the stratosphere to enhance the planet’s reflectivity. This includes (1) sooty aerosols associated with combustion often called “black carbon” and sometimes discussed in nuclear winter studies (Kravitz et al., 2012b; NRC, 1985; Robock and Toon, 2010; Turco et al., 1990) that strongly absorb sunlight, (2) dust particles that could be viewed as more benign once deposited on the ground (Bala, 2009; NRC, 1992), and (3) artificial aerosols that could potentially be designed with specific scattering and adsorption properties and that can take advantage of light-driven migration of particles to guide them to particular atmospheric locations (e.g., Keith, 2010). Although there are various particle types that could be added to the stratosphere to enhance Earth’s albedo, most of the studies described below discuss sulfate aerosols.
FIGURE 3.11 Composite surface temperature response for the two winters following nine volcanic eruptions: (a) observed and (b) CMIP5 multimodel mean of simulations. ENSO effects have not been removed, though these are expected to be smaller as a result of averaging. Note different scales for the two figures. SOURCE: Driscoll et al., 2012.
Aerosol production efficiency, transport, evolution, and loss vary with altitude, temperatures, and winds, among other factors. All methods that introduce aerosols into the stratosphere are expected to affect the reflection and absorption of energy (the aerosol forcing), which will then vary with time and season, unlike the idealized studies discussed in the previous section. The aerosol mass and number, and subsequent forcing, will be sensitive to (1) the mechanism used to produce and deliver the aerosol; (2) the location of the injection; (3) the vertical and horizontal transport processes that mix the aerosols (timescales of days to years); and (4) the chemistry and physical processes that produce, change, and deplete the aerosols (nucleation, condensation, evaporation and sublimation, coagulation, sedimentation, and scavenging). Figure 3.12 shows an example of the distribution of aerosols and the associated radiative forcing from a modeling study using a simple emission scenario.
Studies involving more realistic aerosol injection scenarios are in their infancy compared to sunlight reduction studies, and details regarding the formulation of the physical processes that control aerosol forcing and response matter a lot to study conclusions. Various modeling approaches have been used to explore SAAM that tend to fall into three distinct classes, or generations, based on their level of complexity in treatment of aerosol processes. First-generation studies used “bulk” formulations, where only total aerosol mass is predicted and the aerosol size distribution is assumed; second-generation studies used “modal aerosol formulations,” where mass is predicted together with limited size distribution information; and third-generation studies used “sectional aerosol treatments,” which attempt to follow the full size distribution.
First-generation formulations include studies by Jones et al. (2010), Kravitz et al. (2012a), Rasch et al. (2008b), and Robock et al. (2008). These studies assumed the source gas for the aerosols was SO2 and generally concluded SAAM could produce substantial planetary cooling. Details (altitude, latitude, temporal injection strategies, and aerosol size) varied across studies but most concluded that less than 10 million tons of sulfur per year (MtS/yr) would be sufficient to counter the forcing associated with a doubling of CO2 concentrations (~4 W/m2). Atmospheric mixing would tend to distribute tropical injections in the lower stratosphere globally, and injections in a single hemisphere at high latitudes would dissipate more rapidly than an equatorial source but generally spread to the subtropics over a season (Robock et al., 2008).
Although many of the first-generation simulations of aerosols did not attempt to model the evolution in the size of aerosol particles, this is an important process
FIGURE 3.12 Example of albedo modification aerosols for June, July, and August from a 20-year simulation for a 2 MtS/yr emission: (a, b) aerosol burden (g/m3 and g/m2, respectively) and (c) forcing (W/m2). The white contour in (a) shows the region where temperatures fall below 194.5 K and indicates approximately where ozone depletion may be important. SOURCE: Rasch et al., 2008b.
because large particles with diameters larger than about 0.6 μm reflect sunlight less effectively for a given aerosol mass (Penner et al., 2001) and fall faster, thus having a shorter lifetime, also making them less effective. Particle size also affects the strength of stratospheric heating, and ozone destruction (via the amount of surface area available for inhomogeneous chemical reactions). More comprehensive treatments of aerosol formation and evolution using second- and third-generation approaches (English et al., 2012; Heckendorn et al., 2009; Niemeier et al., 2011) have followed the early studies. Typically, climate models (e.g., Niemeier et al., 2011) use “modal” representations of particle size evolutions, which may be adequate (i.e., within 25%) if tuned to represent the more complete and substantially more expensive sectional models (Mann et al., 2012; Weisenstein et al., 2007). Clearly sectional models may also have difficulty, in comparison with Pinatubo measurements (see Heckendorn et al., 2009).
Studies with more complete treatments concluded that substantially higher injection rates would be needed because processes treated very simply in earlier studies (condensation on existing particles, coalescence, and accretion) act to produce larger particles than previously estimated (large particles descend more rapidly into the troposphere, where they are removed more rapidly and, as noted above, scatter sunlight less efficiently than small particles). English et al. (2012) summarized estimates for models that included a better treatment for aerosol microphysics and found that the injection rate for SO2 to obtain a 6 MtS burden is five times higher than the injection rate predicted by simulations that assumed prescribed size distributions (e.g., Rasch et al., 2008a). The more comprehensive studies found that an increase in the SO2 injection rate from 1 to 10 MtS/yr produced an increase in the peak column mass of sulfate by a factor of 5 and an increase in the peak aerosol optical depth (AOD, a measure of the aerosols’ ability to attenuate light, which is thus related to the amount of cooling) by only about a factor of 3. AOD was reduced disproportionately for the larger injection rates because those rates produce larger particles. The peak in effective radius at 90 hPa (~16 km) varies from 0.4 to 0.6 μm in the three models studying albedo modification that employed second- and third-generation aerosol microphysics (English et al., 2012; Heckendorn et al., 2009; Niemeier et al., 2011). The more comprehensive treatments indicated that at least 10 MtS/yr (approximately the amount of sulfur injected by the Mount Pinatubo eruption) would be needed annually to maintain a radiative forcing of –4 W/m2, roughly equal to but opposite that associated with a doubling of atmospheric carbon dioxide.
Studies have also explored the sensitivity of the albedo modification strategy to the characteristics of the aerosol source, changing the amplitude, source type (SO2 gas, H2SO4 gas, or sulfate particles), and latitudinal extent (e.g., restricted to near the equator or pole, or extending over a broad band of latitudes, or a hemisphere). More
realistic “plume” simulations that allow for faster rates of coagulation have only been performed in one model (Pierce et al., 2010). English et al. (2012) found, in contrast to Robock et al. (2008), that steady tropical SO2 injection does not produce a hemispherically symmetric albedo modification, but instead produces albedo modification that is higher in the Northern Hemisphere (see Fig. 2 of English et al., 2012). A low bias was also found in their Southern Hemisphere Pinatubo results (English et al., 2013), so this result should be confirmed in other models, but it nonetheless may have important consequences for tropical precipitation (see the discussion of Haywood et al.  in Box 3.4).
The most cost-effective strategy may be to have aircraft deliver a sulfate precursor to the lower stratosphere and inject it there where it is converted to gaseous SO3 or H2SO4 (English et al., 2012; Pierce et al., 2010). The above studies used “sectional treatments” that allow an additional improvement in the representation of aerosol evolution for an increase in computational cost. Pierce et al. (2010) concluded that the direct injection of gas-phase H2SO4 would result in higher H2SO4 aerosol burdens than injecting the same amount of SO2. An important component of that study was the use of a subgrid-scale “plume” model that treated the evolution of particles from just downstream of the source injection until it was diluted to a much larger region for the first 2 days following the precursor emission. English et al. (2012) did not attempt to treat the plume evolution, injecting the aerosols uniformly within model cells of a few-hundred-kilometer horizontal extent, and a few kilometers thick, and they did not find the improvement in efficacy associated with injection of H2SO4 seen in the Pierce et al. study, presumably because this process was neglected. It is clear that the technology associated with the injection (e.g., source, composition, and injection rate) matters, and the treatment of the aerosol distribution as it evolves in the plume downstream of the emissions is also very important. English et al. (2012) also estimated increases in upper tropospheric aerosol content by up to a factor of 100 when 10 MtS/yr of emissions were introduced, with potentially important consequences for high clouds.
The studies also indicated that different scenarios (e.g., latitude, altitude, and source type) with the same overall injection rate can increase the burden of aerosols by roughly 50% (see English et al., 2012, Fig. 6; Niemeier et al., 2011, Fig. 2). This discussion highlights the importance of the treatment of aerosol microphysics for the development of the aerosol size distribution and the sensitivity of the albedo modification for a given injection protocol to highly uncertain aspects of the modeled aerosol microphysics. Modeling of aerosol microphysics is still an area of active research, and more work is needed.
BOX 3.4 REGIONAL ALBEDO MODIFICATION
Several studies have looked at the possibility of doing a regionally focused deployment of albedo modification, in particular in the Arctic in response to the rapidly declining levels of Arctic sea ice. Robock et al. (2008) also explored Arctic injections and found that these scenarios produced much smaller aerosol loading, because the removal rate of aerosols is about four times faster in the Arctic than in the tropics. They found that the rapid horizontal mixing of aerosols in the stratosphere, with a lifetime of months or longer, would make it difficult or impossible to fine-tune the geographic pattern of albedo modification through control of the position and timing of SO2 injection. High-latitude injections would spread to cover a substantial fraction of the hemisphere, though concentrations remain higher in the higher latitudes.The more localized albedo modification did achieve an increase in the amount of sea ice relative to the unmodified high-CO2 case, but the climate response was not confined to the Arctic. They noted the potential for significant changes to precipitation in (Indian and Asian) monsoons, and to rainfall in the Sahel region of Africa. That study identified precipitation changes in those regions, but the differences were generally not identified as significant according to formal statistical tests.
Those signatures are consistent with a more recent study by Haywood et al. (2013), which noted that volcanic eruptions that injected aerosols into the Northern Hemisphere preceded three of the four strongest years of Sahelian droughts, and their model also produced a systematic shift in tropical rainfall patterns due to stratospheric aerosol injection. Northern Hemisphere injections shifted Sahelian rainfall southward, leading to serious drought conditions in the Sahel, and Southern Hemisphere injections shifted rainfall northward (similar shifts in rainfall were also apparent over South America). Such shifts in precipitation in regions of high and vulnerable population could have substantial impacts and much more work is needed to identify the robustness of the response.
A recent study by Tilmes et al. (2014) examined model simulations of idealized regional dimming experiments compared to a business-as-usual emissions simulation. They demonstrated that both local and remote feedback mechanisms are important to the surface energy budget in the Arctic. They found that it was necessary to use a local reduction of solar radiation four times stronger than the global reduction in order to preserve Arctic sea ice area and that even with regional Arctic dimming, a reduction of the oceanic meridional overturning circulation and a shutdown of the Labrador Sea deep convection were possible. They concluded that “Arctic regional dimming does therefore not provide a possible solution for containing Arctic sea ice for a business-as-usual greenhouse gas emissions scenario.”
Although one might also anticipate differences between models in the transport of particles within the stratosphere, there has been little study of this aspect, possibly because of differences in experimental design between studies. Most studies to date have designed their simulations independently, for example using different experimental protocols or different assumptions about emissions. A more careful assessment can be performed through model intercomparisons in which emission
characteristics (e.g., aerosol size, amount, and emission region) are carefully prescribed and treated uniformly between models and simulations. Furthermore, the range of possible choices as to which processes to include and the complexity with which they should be represented makes controlled intermodel comparisons more difficult to carry out and analyze. Compared to solar-constant reduction simulations, realistic aerosol injection simulations are in their infancy, but a recent model intercomparison project—GeoMIP (Box 3.1)—may help with this.
Model results from the GeoMIP experiment G4 (RCP4.5, 5 MtSO2 tropical injection of sulfate each year for 50 years, followed by 20 years of cessation) have been examined by only three models that included interactive aerosols, and one of them appears to have had some inconsistencies (Model for Interdisciplinary Research on Climate—Earth System Model—Chemistry [MIROC-ESM-CHEM]) (Ben Kravitz, private communication). Nevertheless, the two remaining models have been compared (Ben Kravitz, private communication). These results show differences of a factor of 2 in the predicted burden of sulfate between the GISS-E2-R and HadGEM2-ES models over Antarctica in July, but results for the two models are similar over the Arctic and other locations and seasons. This difference may potentially be due to removal processes, rather than transport.
Because of the relatively long lifetime of stratospheric aerosols described in the previous sections, the aerosol distribution and aerosol forcing will eventually spread, and models indicate it would be difficult to restrict the aerosol forcing to less than most of a hemisphere, although it may be possible to achieve some nonuniformity latitudinally. In the scenarios considered to date, aerosol burdens and forcing become sufficiently uniform that many of the idealized studies exploring temperature and precipitation responses to regional and global reductions in solar irradiance are also relevant to understanding the climate response to SAAM. In this section we briefly describe the climate responses that are common to the idealized studies discussed previously, but then we focus most attention on climate responses and issues that are unique to SAAM.
Temperature, water vapor, and precipitation. As in the idealized experiments, model simulations suggest that if stratospheric aerosol albedo modifications were increased to compensate for a forcing from a doubling or quadrupling of CO2, equatorial surface temperatures would be somewhat cooler than an unperturbed planet, polar temperatures somewhat warmer, global averaged precipitation would likely be reduced, and the planetary response to SAAM termination would be much like that described in
the section “Timescale Mismatch, Risks of Millennial Dependence, and Constraints on Strategies for Limiting the Duration of Reliance on Albedo Modification” in Chapter 2.
Robock et al. (2008), Rasch et al. (2008b), and Jones et al. (2010) explored the planetary response to steady tropical injections producing stratospheric aerosol perturbations that were quite symmetric between hemispheres. Using a first-generation bulk model, Robock et al. (2008) found tropical injection at a rate of 5 MtSO2/yr (equivalent to one Pinatubo eruption every 4 years) produced a mean cooling of 0.3°C to 0.4°C relative to the unmodified state, and 10 Mt/yr produced a cooling approximately twice as great; for example, the forcing and response is approximately linear with respect to emissions. (Note that this degree of cooling is not borne out by models that treat more comprehensive particle microphysics [English et al., 2012; Heckendorn et al., 2009; Niemeier et al., 2010].) Jones et al. (2010) used a second-generation bulk aerosol model and estimated a temperature response approximately twice as large for a similar emission scenario. All three studies documented reduced precipitation relative to the preindustrial climate like that seen in the section “Idealized Simulations of the Effects of Albedo Modification” earlier in Chapter 3. Robock et al. (2008) and Jones et al. (2010) noted some effects on monsoon circulations. Recent modeling results as part of the GeoMIP set of experiments show that global temperature and precipitation changes are generally closer to preindustrial values with albedo modification (G3 simulations) compared to continued climate change without mitigation, but that “global temperature and precipitation are still redistributed globally” (Anderson and Ault, 2014). Several studies have explored the idea of regional albedo modification (Box 3.4). The discussion in Box 3.4 is also of relevance to climate interventions which were intended to produce a globally uniform aerosol layer, but which for one reason or another inadvertently resulted in significant regional inhomogeneities.
Clouds. As described in the section describing possible impacts below, stratospheric aerosols may affect clouds, but their impact remains poorly understood. Kuebbeler et al. (2012) noted that increases in stratospheric aerosol loadings will likely lead to an increased upper tropospheric temperature, stabilizing the upper troposphere, decreasing vertical velocity, and ultimately reducing ice crystal nucleation rates and producing optically thinner cirrus clouds. They estimated optically thinner cirrus clouds could exert a strong negative cloud forcing in the longwave which contributes possibly as much as 60% to the overall net forcing. However, their model did not include feedbacks of the stratospheric injection on stratospheric ozone, which is predicted to decrease (see “Environmental Consequences of SAAM” section below) and might lead to decreases in temperature. On the other hand, Cirisan et al. (2013) argued that the net
radiative effect of aerosol-induced changes to number concentrations in high clouds should be small, but this study did not include feedbacks to temperature and humidity in the upper troposphere. Uncertainty in high cloud feedbacks represents a major uncertainty in estimating the climate response of a given amount of stratospheric aerosol injection.
Ozone and indirect radiative effects. Tilmes et al. (2008, 2009), Heckendorn et al. (2009), and Pitari et al. (2014) explored the impact of SAAM on ozone depletion and concluded that SAAM sufficient to counter a doubling of CO2 would delay ozone recovery (due to the decrease in halogens) by a few decades. In one example from these studies, Pitari et al. (2014) in a GeoMIP model intercomparison estimated that in order to counter a fourfold increase in CO2 concentrations, sulfate aerosol surface area density similar to conditions a year after the Mount Pinatubo eruption would be required, and there would be measurable impacts on ozone distributions and surface UV-B radiation. They estimated that if active chlorine (ClOx) concentrations were characteristic of values expected in 2040-2049 that chemical reactions on the sulfate aerosols would decrease the globally averaged ozone by less than 1% (ozone would increase slightly at low and middle latitudes and decrease more strongly in polar regions). These changes are substantially smaller than the ozone depletion measured between 1980 and 2000 from ClOx (McKenzie et al., 2011). They also concluded that any increase in UV-B radiation at the surface due to ozone depletion would be offset by the screening by the aerosols themselves in the tropics and midlatitudes, while in polar regions the ozone destruction effect would dominate the aerosol screening effect, and the surface UV-B radiation would increase by 5% on average, with 12% peak increases during springtime. Because ozone is a radiatively important gas (in the solar and longwave), changes in stratospheric ozone would also produce changes to the tropopause radiative forcing, estimated for the 2040-2049 decade to be less than −0.1 W/m2. Because ClOx would continue to decrease after 2050, the suppression of other ozone-destroying reactions (involving nitrogen) becomes more important than destruction of ozone by ClOx, and SAAM was estimated to increase total stratospheric ozone after 2050.
Tilmes et al. (2009) used a whole-atmosphere model with a fully resolved representation of the stratosphere and concluded that the detailed stratospheric response had an important effect on the geographic pattern of the tropospheric and surface response to stratospheric aerosol injection. In particular, the high-latitude response to stratospheric aerosol injection was much weaker in the simulations with a resolved stratosphere than in simulations that did not adequately compute the stratospheric response. The weakened polar response implies a less effective offset of CO2-induced
polar warming, which is important insofar as preserving Arctic sea ice and permafrost is an often-assumed goal of albedo modification. Stratospheric heating can affect the stratospheric water budget, particularly when the aerosol distribution is significantly nonuniform. Accurate simulation of the stratosphere-troposphere connection requires fully resolved stratospheric dynamics and is currently a considerable modeling challenge.
Sea ice. Berdahl et al. (2014) carried out a limited multimodel study of the Arctic response to two stratospheric aerosol-injection scenarios intended to produce a globally uniform (rather than Arctic-limited) albedo modification. The scenarios were constructed to fix the top-of-atmosphere energy balance at 2020 levels (which already has a positive energy flux into the Earth system) or fix the stratospheric aerosol forcing at 2020 levels while CO2 forcing continued to increase. They found, not surprisingly, that global mean warming and reduction of sea ice continued past the year 2020, because the model experiments were (by design) not intended to entirely counter the radiative forcing by greenhouse gases. In these simulations, aerosol injection delays, but does not prevent, the ultimate loss of September Arctic sea ice. There was also considerable discrepancy among the models as to the effectiveness of the aerosol injection at delaying the loss of sea ice, but further work will be needed to ascertain the source of this discrepancy. This also gives a good indication of the additional kinds of simulations that may become available as GeoMIP2 progresses.
Land biosphere and carbon cycle. Land biosphere models and global carbon-cycle models have been integrated into three-dimensional coupled atmosphere-ocean physical climate models and have been used to assess the likely response of the land biosphere and global carbon cycle to inadvertent human-induced changes to atmospheric composition and climate (IPCC, 2013a). These models project that, under the anthropogenic climate change scenarios considered by the Intergovernmental Panel on Climate Change (IPCC), higher carbon dioxide concentrations would likely increase productivity of the land biosphere nearly everywhere (a result of CO2 fertilization), but human-induced climate change tends to decrease biological productivity in the tropics and midlatitudes (a result primarily of heat stress and secondarily of water stress) and tends to increase biological productivity in the northern high latitudes (IPCC, 2013a, Fig. 6.2). Insofar as albedo modification approaches are able to offset climate change effects of increased atmospheric greenhouse gas concentrations, they would be expected to have no effect on the increased productivity that would be expected as a result of increased atmospheric CO2 concentrations, but they might
tend to increase the productivity of the land biosphere in lower latitudes due to the removal of heat stress in the tropics. These expectations are supported by idealized studies performed as part of the GeoMIP project (Figure 3.13; Kravitz et al., 2013a).
FIGURE 3.13 All-model ensemble annual average differences in terrestrial net primary productivity (kg C m−2 a−1), averaged over years 11-50 of the simulation. For these panels, “abrupt4xCO2” is a climate with a quadrupling of the CO2 concentration, “G1” is a climate with a quadrupled CO2 and a reduction in sunlight sufficient to return the global average surface temperature to a reference state, and “piControl” is the preindustrial climate. Top panel shows abrupt4xCO2-piControl, middle panel shows G1-abrupt4xCO2, and bottom panel shows G1-piControl. Stippling indicates where fewer than 75 percent of the models (for this variable, 6 out of 8) agree on the sign of the difference. SOURCE: Kravitz et al., 2013a.
In climate model projections with a dynamical representation of the carbon cycle, the land biosphere takes up more carbon with albedo modification than it would have in the absence of albedo modification, and because of cooler ocean surface temperatures the ocean also takes up more carbon (Matthews and Caldeira, 2007). Thus, atmospheric CO2 increases may be moderated somewhat (<20 percent; Matthews and Caldeira, 2007) by carbon-cycle response to large-scale albedo modification. These simulations did not consider the increases in diffuse radiation that would be caused by stratospheric aerosols, which would be expected to further increase carbon sequestration by the land biosphere (Mercado et al., 2009). Changes in the total amount of sunlight are anticipated to have much smaller effect on net primary productivity (Bala et al., 2002; Kravitz et al., 2013a; Matthews and Caldeira, 2007).
One concern about albedo modification for the purposes of intentional climate modification is the projection that precipitation would decrease globally (Bala et al., 2007) (see also discussion associated with Figure 3.3). However, at global scale, precipitation must balance evaporation, and the decrease in precipitation is associated with decreased evaporation, resulting largely from a moistening of the boundary layer over the ocean (Cao et al., 2012). An important question for the land biosphere is thus how atmospheric water vapor transport to the land biosphere is affected by albedo modification. This net transport represents the balance of changes in precipitation and evaporation. The results of the GeoMIP project (Kravitz et al., 2013a; Figure 3.4) indicate that “precipitation minus evaporation anomalies are less than 0.2 mm day-1 in magnitude over 92 percent of the globe, but some tropical regions receive less precipitation.” Further discussion of changes to the hydrological cycle from albedo modification is found in the “Idealized Simulations of the Effects of Albedo Modification” section above.
Detailed projections of land biosphere models at regional scale have large uncertainties, but the models indicate the sign of likely responses to various climate forcings. For example, if soils were projected to become parched, the models would project low amounts of net primary productivity. The GeoMIP results (Kravitz et al., 2013a) and results from other modeling groups (cf. Bala et al., 2002; Matthews and Caldeira, 2007; Naik et al., 2003) indicate that, at global scale, albedo modification by stratospheric aerosols in a high-CO2 world would have little detectable effect on land biological productivity in most places but could in some places cause significant increases or decreases in land biological productivity. Relative to the preindustrial state, a high-CO2 world with albedo modification is projected to have higher biological productivity in nearly all land areas, largely due to CO2 fertilization. These projections of changes in biological productivity of natural ecosystems are consistent with projected changes in expected crop yields (Pongratz et al., 2012; Xia et al., 2014). Climate models do not project substantial consequences of sudden termination on the land net primary
productivity beyond what would have occurred had albedo modification never been implemented (Jones et al., 2013; Matthews and Caldeira, 2007), although what sudden termination would mean at the species level remains an open question.
Increased net primary productivity on land is not necessarily a positive outcome for natural ecosystems. Changes in the amount and quality of light, and the patterns of precipitation and evaporation, as well as changes in atmospheric composition and possibly other factors like cloudiness and winds could be expected to disturb natural ecosystems with consequences that at this time are difficult to predict. For example, it is entirely possible that net primary productivity would increase in some areas but that this increase in net primary productivity would be accompanied by the extinction of some native flora and fauna. Furthermore, almost all of the model results described above are based on a limited set of idealized studies, many of which considered dimming the sun instead of actually representing atmospheric aerosols. Many of these simulations did not consider effects of diffuse radiation or include adequate representations of nutrient dynamics. All such simulations are greatly simplified compared to the real world, and further work is required to reduce the uncertainty in these projections.
Acid deposition. Although SAAM would substantially increase the amount of stratospheric sulfate, it is a small source and sink of sulfate compared to other natural and pollution sources that contribute to the acidity of land and ocean and is not expected to have an important impact on planetary ecosystems (see section “Environmental Consequences of SAAM”).
Observational requirements for SAAM should be at a level sufficient to quantify the evolution of the source material introduced to form aerosol particles and the resulting radiative response. This would include quantifying the amount of source material (SO2 or sulfuric acid) injected, its rate and direction of spread with time, the formation of H2SO4, the size of the particles formed, their effect on cirrus clouds, and their effect on Earth’s radiation budget. These requirements are relevant to activities initiated as a result of a concerted world effort or via unilateral and uncoordinated actors. Important impacts on climate are anticipated with albedo modification activities of 1 W/m2 of radiative forcing reduction or less. Detection of this amplitude of SAAM would require determination of Earth’s solar radiation budget to an accuracy of better than 1 W/m2.
The current U.S. aerosol monitoring from space relies on the MODIS,3 MISR,4 and OMPS5 instruments, and the CALIPSO6 mission, although a number of other aerosol products are available on instruments from Europe and Canada.7 The stated accuracy for MISR AOD is about 0.03 or 10 percent, whichever is larger. The MODIS team reports their sensitivity as 0.03 ± 5 percent, which in practical terms is similar to the accuracy of MISR over ocean, since the AOD over ocean is generally low. These accuracies can be compared to the predicted peak zonal average increase in AOD for a 1 MtS/yr injection rate of around 0.05 (English et al., 2012). Such a nearly full-blown experiment would be barely detectable. A modeled 10 MtS/yr injection produced a peak zonal average increase in AOD of 0.2 and so should be easily detectable with current instrumentation.
The OMPS instrument measures SO2 as well as AOD, but it is a limb profiler. The stated limb profiler sensitivity is 3 × 10−6 km−1 for a 1- to 2-km vertical resolution. Thus, this instrument should be capable of monitoring changes of order 0.001 in AOD. However, as this is a limb measurement, it integrates over a path along the line of sight; through the lower stratosphere, for example, the path is effectively 300 to 400 km long, so an aerosol feature would have to be concentrated along the actual line of sight of the limb sounder during occultation to detect something as thin as 0.001 in AOD. The advantage of this instrument, however, is that, in addition to obtaining perturbations to SO2, the approximate altitude of the aerosol layer would be known. This provides a great advantage for validation of model results.
The CALIPSO instrument uses backscattered radiation from a downward-pointed lidar, which can give information on the vertical distribution of the detected aerosols in the fairly narrow region where the lidar is pointing. The European/Japanese EarthCARE satellite mission, scheduled for launch in 2015,8 will also use this technology. Winker et al. (2009) estimated that a single shot from the CALIPSO lidar is not accurate to 0.01 km-1sr-1 so horizontal averaging is used to improve the detection of backscattering coefficients from aerosol layers. However, Kacenelenbogen et al. (2011) compared results from the Version 2 CALIOP AOD retrievals to those from other instruments and found they were significantly smaller than other retrievals.
As noted in the section examining the processes that produce H2SO4, it might be important to also obtain measurements of the aerosol size distribution in order to
3 Moderate Resolution Imaging Spectroradiometer.
4 Multi-angle Imaging SpectroRadiometer.
5 Ozone Mapping Profiler Suite.
6 Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation.
aid in determining the efficacy of injections. However, current remote sensing instrumentation is not very sensitive to aerosol size and is unlikely to be able to pick up a signal from stratospheric injection. Thus, detection of a stratospheric injection signal would depend on the specifics of the observation (see discussion of OMPS detection above, for example). Among the current generation of instruments, we can retrieve about three to five size bins with MISR, provided that the total column midvisible AOD exceeds about 0.15 or 0.2. A multiangle, multispectral, polarimetric imager could improve on current capabilities. With a next-generation instrument, with polarization sensitivity on the order of 0.5 percent, in addition to the 1 percent to 3 percent absolute radiometric calibration similar to MISR and MODIS, we expect greater sensitivity to particle size distribution. Qualitatively, such an instrument would be expected to provide an additional measure (moment) of the particle size distribution (e.g., giving mean effective radius plus size distribution width or variance), but the quantitative sensitivity is not well constrained at this point, and no specific instrument design is slated for building and launch. Aerosol size distributions can be measured from balloon-borne instruments, as was demonstrated after the eruption of Mount Pinatubo, but these measurements are limited in spatial coverage.
The lifetimes of the above instruments and satellites were estimated as part of the Midterm Assessment of Earth Science Decadal Survey Report, which was based on the 2011 NASA Senior Review of each instrument. According to that report, MODIS on Terra is expected to last through 2017, and MODIS on Aqua through 2018 (extended to 2022 in the 2013 NASA Senior Review) (both limited by mission life, not instrument life). MISR is expected to last through 2017 (Terra life expectancy); OMPS on NPP was not covered as part of the Senior Review, but could be expected to last through its design life plus 4 years (the long-term average used for the original Decadal Survey), so it should last through 2019. CALIPSO is expected to last until 2016.9
In addition to the capabilities above, it would be wise to maintain a stratospheric monitoring capability in order to capture information relevant to albedo modification in the event of a volcanic eruption that injected SO2 into the stratosphere (Box 3.5).
A variety of consequences are anticipated to arise from significant changes in stratospheric aerosols. The processes producing these changes are described in the sections
9 Details supplied by Stacey Boland, personal communication.
BOX 3.5 OBSERVATION REQUIREMENTS FOR MAKING BETTER USE OF VOLCANOES AS NATURAL EXPERIMENTS
Observational capabilities must be in place to determine the following quantities in order to make effective use of volcanic eruptions as natural experiments:
- Mass, composition, and vertical distribution of the substances injected into the atmosphere by the eruption;
- Resulting aerosol properties and their evolution in space and time, as well as associated changes in stratospheric chemistry, notably related to ozone; and
- Changes in radiative forcing. This includes top-of-atmosphere measurements of albedo change, perhaps supplemented by ground-based or aircraft-based short-wavelength radiation measurements, but there is also a need to monitor long-wavelength (infrared) changes, since these are involved in aerosol-induced stratospheric heating.
Sufficiently large eruptions will produce a temperature response in the upper atmosphere as well as at Earth’s surface, which will also need to be monitored as a basis for testing simulations of the response of climate to the eruption. The chief impediment to characterizing climate response is separating the volcanically forced response from effects due to natural variability such as El Niño or the Quasi-Biennial Oscillation, and it is unlikely that any improvements over the existing temperature and precipitation monitoring network would significantly ameliorate the problem. It would also be desirable to monitor the response of cirrus clouds to the eruption, though distinguishing between microphysical effects of the volcanic aerosols and cirrus changes arising from the general climate response is likely to be a challenge.
Advanced preparation will be needed if scientists are to make the best use of the next major volcanic eruption. Although Pinatubo is the best characterized eruption to date, ironically our ability to monitor stratospheric aerosols has deteriorated since that time, with the loss of the Stratospheric Aerosol and Gas Experiment (SAGE) II and III satellite-borne instruments. SAGE III was capable of limb-scanning measurements of aerosol optical depth as well as vertical profile measurements of aerosol optical depth. If the SAGE III on ISS launch is successful, some of this capability will be restored. SAGE III on ISS is scheduled to be the first mission launched by the commercial Space-X vehicle in 2015, and to be deployed on the International Space Station (ISS). The ISS platform and its low-inclination orbit are not ideal for aerosol monitoring but would provide some useful capability. Maintaining SAGE III on ISS or a similar capability for the next several decades is a minimal requirement; it is possible that a more economical platform, more specifically targeted to stratospheric aerosol monitoring, could eventually replace the SAGE family.The Optical Spectrograph and Infrared Imaging System (OSIRIS) satellite-borne instrument has been used effectively in the post-SAGE years (Kravitz et al., 2011b), but this instrument is running past its designed lifetime and may not last much longer.
Some capability for monitoring Earth’s radiation budget and the factors that influence it already exists. These include instruments such as the Clouds and the Earth’s Radiant Energy System (CERES) satellite instruments, which measure the various components of Earth’s radiation
budget, and the CALIPSO mission, which measures the vertical structure of clouds and aerosols. Any improvements that could be made with regard to accuracy, coverage, and spatial resolution would greatly enhance the ability to understand the nature of the volcanic response and address shortcomings in the ability to simulate it accurately. Moreover, while the most recent CERES instrument, launched on Suomi-NPP, is expected to operate for at least several more years, CALIPSO, which has been operational for nearly 8 years, is well past its 3-year design life.
There is also a need for a deployable rapid-response observational task force, but any such capability would need to have multiple uses so that the considerable investment required would not lie fallow between major eruptions. Ground-based and airborne lidar instruments—which work by emitting and measuring how much laser light bounces back from aerosols—are valuable for characterizing the volcanic plume and resulting aerosols; lidar has been used effectively in characterizing recent eruptions (Kravitz et al., 2011b). There may also be a role for selective deployment of ground-based and airborne radiometers for the purposes of refining estimates of the amount of solar radiation transmitted through the stratospheric aerosol mass. Some in situ monitoring of stratospheric chemistry, particularly targeted at ozone chemistry, would also be needed. Data collection alone will not be sufficient; there also needs to be an appropriate level of investment in data analysis, running simulations for comparison, and subsequent model development to correct shortcomings.
If there were a standing monitoring capability to rapidly respond to a volcanic eruption, the question would remain as to whether an eruption would be expected in the next few decades. At this point, it is not possible to predict future volcanic eruptions with more than a few days lead time at best and not all eruptions can currently be predicted. Using statistics from the past 1,500 years, there have been 50-year periods with no large eruptions (1912-1963) and 50-year periods with as many as four large eruptions, including the largest, the 1257 Samalas eruption. Analysis of data from 1750 to the present suggests that the time period is too short to give reliable estimates of return periods for large explosive eruptions (Ammann and Naveau, 2003; Deligne et al., 2010).
As such, a rapid response system may be heavily subscribed for the purpose of posteruption observations, or undersubscribed, depending on the amount of volcanic activity. A wise strategy would be to have a dual use for such a system so that it would be available for rapid and sustained deployment immediately following a volcanic event but would also be useful even without substantial eruptions. Such a capability would have significant value for basic atmospheric research, providing data that would improve process models as well as large-scale climate models.
“Idealized Simulations of the Effects of Albedo Modification” and “Modeled Climate System Responses to SAAM” above, and are repeated here for clarity:
- Increased aerosol will affect stratospheric ozone depletion. Current understanding indicates that ozone depletion should diminish in the future as halogen levels decrease.
- There may be impacts on UV-B light reaching the surface, affected by the ozone depletion, and the aerosols themselves. Current understanding indicates the changes would be small.
- If SAAM were employed, there would be changes to precipitation, surface temperature, and soil moisture that may have an impact on ecosystems. Current understanding indicates the changes would be much smaller than those experienced if SAAM were not employed.
- Sunlight intensity would be reduced, but the amount of sunlight arriving from different directions would increase due to scattering on the aerosols (resulting in an increase in the ratio of diffuse to direct sunlight). More sunlight would reach into the plant canopy, increasing photosynthesis, again with possible impacts on natural and managed ecosystems. Sunlight reduction could also affect home heating and solar power facilities.
- Introduction of stratospheric aerosols is likely to slightly increase the acidity of the snow and rain reaching the surface. The effect is estimated to be a very small fraction of the acidity increases associated with industrial pollution today. Thus, any important effects might be counteracted by controlling anthropogenic emissions within the troposphere (Kravitz et al., 2009; Rasch et al., 2008b).
There is also of course the possibility of environmental consequences that scientists have not yet identified. It is interesting to consider how scientists would identify an environmental consequence (including detection and timescale). It should be more straightforward to characterize the impacts on chemistry, light intensity, and precipitation. On the other hand, it will be much more difficult to detect impacts on ecosystems.
To date, there have been no deliberate attempts to deliver sulfate aerosol precursors to the stratosphere with a controlled release and a monitoring program to assess the destiny of the source species as the aerosols form, evolve, disperse, and eventually disappear. As such, all estimates of the technical feasibility are currently theoretical, based on observations of aerosol forcing following volcanic eruptions, modeling studies, and some measurements of plume dispersions behind aircraft and rockets from the early 1970s (Turco and Yu, 1997, 1998, 2012). These studies are not sufficient to provide robust estimates of the development and evolution of the aerosol.
10 See Appendix E for a larger discussion of feasibility.
For reference, artificially duplicating even a relatively small volcanic eruption such as Sarychev in 2009, which ejected 1.2 Tg of sulfur dioxide into the atmosphere, would require a substantial undertaking. The sulfur dioxide loading is roughly equivalent to the total payload capacity of 27,000 flights of an Airbus A330-300 aircraft, and even this comparison understates the difficulty of the injection as commercial aircraft cannot fly high enough to duplicate the required stratospheric injection levels. Specialized aircraft (or other injection platforms) would be needed to carry out the injection. It is unclear at present whether any substantially smaller-scale field experiment involving modification of the stratosphere could begin to compete in scientific payback with what can be learned through assiduous study of the volcanic response (Robock et al., 2010).
The main issue regarding the feasibility of this strategy is associated with an accurate characterization of the aerosol source as it is released into the atmosphere from the delivery mechanism (how much new particle formation, how much vapor deposition on existing particles, and how much coalescence of new particles) as the plume disperses. These characteristics influence decisions about the strategy of delivery and govern the efficacy of the strategy (radiative forcing per unit emission of sulfur). It is also possible that the environmental consequences mentioned above could lead to a decision that the strategy is infeasible.
Robock et al. (2009b) and McClellan et al. (2012) have estimated costs of various delivery mechanisms to take sulfur to the stratosphere, but they did not address the issue of then producing aerosols with a desired size distribution. McClellan et al. estimated costs based on new aircraft designs optimized for delivery of sulfur, followed by in situ oxidation, to be $1 billion to $3 billion per MtS/yr to the stratosphere (20 to 30 km) or $2 billion to $8 billion to deliver 5 Mt to the same altitude range. There are similar estimated costs for hybrid airships that produce a majority of lift force from buoyancy and a smaller percentage from aerodynamic forces, but their large surface area complicates operations in high-altitude wind shear, and development costs were more uncertain. Commercially available aircraft, although poorly suited for high-altitude flight and significantly more expensive per mass of aerosol, could be used to deliver aerosol source species to about 18 km for exploratory work. “Pipes suspended by floating platforms provide low recurring costs to pump a liquid or gas to altitudes as high as 20 km, but the research, development, testing and evaluation costs of these systems are high and carry a large uncertainty; the pipe system’s high operating pressures and tensile strength requirements” (McClellan et al., 2012) make their feasibility very uncer-
tain, and their ability to deliver aerosols distributed across broad swaths of the atmosphere is limited. Costs for rockets and guns appear to be significantly higher than for other systems, but they may also be suitable for exploratory research, or for delivery to very high altitudes. As a general caution, it is noted that many large-scale engineering projects experience higher costs than initially estimated, so all such cost estimates are likely to have significant uncertainties.
These estimates do not appear to account for costs associated with operating in an environment of high concentrations of SO2 and sulfate aerosols, but there is some evidence these issues should be considered. Carn et al. (2009) pointed to an increase in the incidence of crazing of acrylic windows (Bernard and Rose, 1990; Casadevall et al., 1996), forward airframe damage, and accumulation of sulfate deposits (anhydrite and gypsum) in turbines that block cooling holes, causing engine overheating (Casadevall et al., 1996; Miller and Casadevall, 2000), following the El Chichón (1982) and Pinatubo (1991) eruptions. Increases in aircraft damage would presumably increase the cost of deployment.
The cost of a responsible deployment strategy involves not just the cost of aerosol injection, but the cost of observing systems and infrastructure to detect and attribute the magnitude of and response to albedo changes from stratospheric aerosol injection. Estimating the full costs of an observing system and infrastructure to do this was beyond the charge of this committee, but these costs are generally estimated to be significant, as typical satellite deployment costs often run into the billions of dollars.
There are a variety of other issues that have been raised regarding SAAM. These issues are real, and they must be considered and balanced when considering the other consequences, and possible benefits, from SAAM. This section includes several examples but is not a comprehensive list. One example, as pointed out by Robock (2008), is that SAAM would tend to “whiten” the sky (Kravitz et al., 2012a), as well as produce more colorful sunsets by increasing the scattering of sunlight. In addition, changes in direct versus diffuse sunlight may produce changes in ecosystems in the long term. For example, they would be expected to stimulate productivity in the understory of land ecosystems. Changes in UV-B light could also have an effect. Various crops need to be studied, as well as further studies on natural systems, in order to better quantify these types of impacts. Other examples of these types of issues have been compiled elsewhere (Robock, 2008, 2014), and these types of issues may need to be considered as part of an assessment of environmental impacts of SAAM.
There are many component processes that are not sufficiently well understood to produce quantitative characterization of processes important to SAAM, and unambiguous statements about how an intervention by SAAM would affect the planet are thus not possible. Several processes are particularly deserving of attention from both modeling and measurement points of view because they are critical to any implementation of SAAM and are unique to SAAM strategies of climate intervention:
- stratospheric aerosol microphysics (formation, growth, coalescence, and dispersion);
- impacts on chemistry (particularly ozone);
- impacts on water vapor in the upper troposphere and lower stratosphere; and • effects of additional aerosol on upper tropospheric clouds.
Because these processes are simplified and approximated in models, it is difficult for models to produce quantitative (or even, in some cases, qualitative) characterizations of SAAM or any resultant impacts (good or bad) to the planet. More research (measurements and models) would be needed if more precise statements about SAAM and its potential to benefit or harm the planet are desired.
More and better observations would be useful to (1) fill in the blanks in understanding and model treatments, (2) more strongly constrain models, and (3) provide the testbed needed to evaluate model performance. Better models and a better understanding of their limitations would produce more confidence in the predictions. The committee attempts to identify a few obvious opportunities for producing better understanding and the reasons why we think these things are important.
- Because models often disagree, it is important to compare them frequently—to each other (with varying details of complexity) and to observations. This motivates at least four kinds of intercomparison activities:
- Better intercomparison of climate models using varying treatments of aerosol microphysics and employing scenarios that are more strongly constrained (in terms of the type, amount, and altitude of precursor emissions) than have been hitherto performed by the GeoMIP studies would help in understanding model uncertainties and their projection of climate consequences. Historically, GeoMIP has focused most of its attention on solar dimming experiments. It is time to put more emphasis on aerosol formation and evolution, and subsequent impacts on clouds, chemistry, and cli-
mate. When differences are evident it is important to identify the reasons for the difference rather than produce an inventory of model simulation variations.
- There is a variety of climate components that have as yet been almost entirely neglected, and more attention is merited, in particular toward (a) impacts on ocean circulations; (b) consequences to ecosystems from possible UV-B changes; (c) interactions of SAAM with dominant modes of interannual variability, volcanic eruptions, and other unpredictable or unpredicted events; (d) dynamic influences of the stratosphere on the troposphere, as they seem to have the capability for profoundly influencing the nature of high-latitude response, and therefore sea ice and glaciers. Other features (e.g., temperature) have received much more attention, but precipitation features (including monsoons) remain a particular challenge and continued attention is merited.
- Intercomparison between global-scale model formulations of aerosol, clouds, chemistry, and aerosol dispersion and finer-scale models (box and plume models) is useful. Such comparisons would challenge the simplified formulations present in global models with the much more detailed formulation present in the fine-scale models. Only a few such comparisons have been made so far, and the relevant studies differ sufficiently to make identification of common features and deficiencies difficult. More uniform, internally consistent, and comprehensive comparisons would help.
- Comparisons between global models and relevant observations, particularly those following volcanic eruptions, are useful. An increasing emphasis on comparisons with data sets constructed from present and future field studies and satellite data sets that are designed to challenge models could be helpful (see discussion below). Comparisons of model simulations to “de minimus” deliberate introduction of aerosol to assess aerosol microphysics, mixing processes, and impact on local atmospheric chemistry may also be useful.
- The response of the climate to volcanic eruptions is likely to provide one of the best opportunities for challenging a model’s global characterization of SAAM and its impact on the environment. The ability of climate models to simulate the aerosol evolution, and the subsequent response of the Earth system to past and future volcanic eruptions, is a necessary but not sufficient test of any model’s capabilities in assessing climate change. Improved observations discussed below could provide increasingly more comprehensive and stringent tests for climate models.
Field studies, lab experiments, and remote sensing. Although model intercomparison can give a sense of the uncertainty in model predictions, it cannot by itself establish that the models have included the correct physics to the correct level of fidelity. There is a need to develop experiments at the correct scale to test the models and model components and to have the tools available to observe the formation and removal of particles following a stratospheric volcanic eruption. Several actions would be beneficial:
- There is a variety of topics in which field and laboratory studies would help to improve understanding about components critical to SAAM. Some of these studies would probably fall into the category of “de minimus” studies, that is, studies that would have no measurable effect on climate but would provide information that would help in the development, calibration, and evaluation of models and the processes in models.
- At present it is not clear whether a small field experiment involving injection of substances into the stratosphere could resolve the outstanding scientific questions without being of a scale large enough to be considered as deployment (see, however, Keith et al., 2014). For proposals for small-scale projects that inject materials into the stratosphere with environmental risks comparable to ongoing commercial or other permitted activities and that address unresolved scientific issues pertaining to stratospheric aerosol injection, development and peer-reviewed analysis of those proposals should be considered by a transparent deliberative process to aid in developing clear guidelines (see Chapter 4).
- The committee sees opportunities and needs for better measurements in characterizing particle formation, particle growth, particle dispersion, and chemical and radiative consequences that are relevant to SAAM.
- There are also obvious opportunities to make better measurements of volcanic eruptions. The committee suggests that increased attention to satellite measurements of stratospheric aerosols and features that respond to aerosol perturbations would be useful to understanding the consequences of SAAM.
- A rapid-response observational capability to make better use of the next major volcanic eruptions (Box 3.5) would also be very useful in characterizing possible consequences of SAAM. This capability would involve space-borne capabilities for monitoring stratospheric aerosols (which would of necessity be multiple use, since large volcanic eruptions are infrequent) and rapidly deployable ground-based and airborne instruments. As discussed above, associated modeling work is required, particularly with models that resolve stratospheric dynamics and which model the chemistry bridging the injected substances to the formation of aerosols (see Appendix D for further details).
The committee emphasizes that the sociopolitical risks of both modeling and field research be considered, even for experiments that may yield useful scientific information, in light of public perceptions. This is further discussed in Chapter 4.
Low clouds, particularly over dark ocean surfaces, play a very important role in Earth’s energy budget by scattering sunlight back to space that would otherwise reach and warm the surface. Because of the low albedo of the ocean surface and the “whiteness” of ocean clouds that very efficiently reflect sunlight back to space, rather modest changes in cloud albedo, cloud lifetime, or cloud areal extent might produce significant changes to both local and planetary albedo (Slingo, 1990). Low-lying stratocumulus clouds cover 20 percent to 40 percent of the world’s ocean as a fraction of the daytime annual average, as illustrated in Figure 3.14 (Russell et al., 2013).
FIGURE 3.14 Daytime annual average stratocumulus cloud amount (%) over the 1983-2009 period, specifically the subset of low clouds that can be viewed from space without overlying clouds.
SOURCE: Figure adapted from Russell et al. (2013) using data obtained from International Satellite Cloud Climatology Project (ISCCP) D2 monthly means (http://isccp.giss.nasa.gov/products/browsed2.html).
Using simple theoretical arguments based on the work of Twomey et al. (1968), Latham (1990) suggested that it might be possible to deliberately introduce additional aerosols to act as cloud condensation nuclei (CCN) near the cloud base, increasing cloud drop number and changing the properties of clouds in their vicinity to make them more reflective. These ideas have come to be identified as “marine cloud brightening” (MCB). The processes that control this response of clouds to additional aerosol particles remain poorly understood (IPCC, 2007b), even though they are very important regulators of the energy budget of the planet. These low-cloud changes are often (but not always) assumed to occur over rather small regions of the planet, meaning that very large changes in local energy fluxes would be needed to produce a significant planetary-scale change.
Twomey (1974, 1977) calculated that cloud systems with smaller and more numerous drops would reflect more sunlight than systems with bigger and fewer drops, all else being equal (size, cloud depth, and amount of condensed water). This is because the surface area of the smaller drops is larger (for the same volume of liquid water), and light scattering is proportional to surface area. Albrecht (1989) observed that cloud systems with smaller and more numerous drops might precipitate less easily. Later studies examined the possibilities that these more polluted clouds with smaller drops might hold condensed water for longer times, might persist for longer periods of time, and might extend over larger areas than they would if they were composed of fewer, larger drops. All of these mechanisms can influence the planetary albedo.
Liquid drops in warm clouds always originate on aerosol particles, typically through a drop formation mechanism first described by Köhler (1921). The proclivity of aerosol particles to serve as nuclei for drop formation depends on the aerosol size, chemical composition, and surface properties. Larger particles (typically >1 μm in diameter) with compositions that interact easily with water vapor (hydrophilic particles) are called cloud condensation nuclei. Aerosol particles that take up water vapor more readily form cloud drops more easily than those that do not, and larger hydrophilic particles “compete” with other particles, growing to cloud drops rapidly in a saturated air mass and eventually either forming precipitation or evaporating after they are exposed to unsaturated air for a time.
MCB is an attempt to increase the albedo of cloud systems by introducing extra aerosol particles to serve as CCN in air masses that participate in cloud formation. There are many different types of clouds, and each type is driven by subtle but important differ-
ences in the balance of processes that govern cloud formation and evolution. Different cloud regimes are likely to have differing susceptibility to brightening strategies. Various theoretical, observational, and empirical approaches can be used to identify clouds that are susceptible to aerosol brightening, that is, most likely to be brightened effectively by particles (see Figure 3.15) (Oreopoulos and Platnick, 2008; Salter et al., 2008). These and other studies suggest that regions in the eastern subtropical ocean basins typically occupied by “marine stratocumulus clouds” (low, layered clouds over ocean regions) are most susceptible to aerosol changes.
The lifetime of aerosol particles in the marine boundary layer is largely driven by the frequency of frontal precipitation and local drizzling, meaning that it is highly variable but typically 2 to 5 days in the northeastern and southeastern Pacific and the southeastern Atlantic where marine stratocumulus occur frequently (Coakley et al., 2000). The short particle lifetimes make it possible to produce big local changes to the cloud albedo and radiative forcing that vary significantly in space and time, a signature that is quite different from the idealized forcing distributions discussed in sunlight reduction studies and stratospheric aerosol albedo modification strategies where the aerosol forcing can spread globally or across most of a hemisphere.
Conceptually the basic MCB idea is quite clear, but in reality, the processes that control cloud droplet formation are incredibly complex and difficult to include in global models. Aerosol-cloud interactions are one of the major challenges in climate modeling today (IPCC, 2013a).
The first complication is that, within the simple physics described by Twomey (1974), the quantities that are held constant (total water content of the air) and those that vary (aerosol size and composition distribution) contribute to determining the cloud’s maximum supersaturation, a quantity that in turn is affected by the cloud droplet number concentration and the aerosols in the air parcel. This introduces a damping effect by which increased numbers of CCN (at constant specified supersaturation) cause a decrease in actual maximum supersaturation; this decreases the fraction of the available CCN that activate in cloud because the higher CCN number results in a lower maximum supersaturation. Or, in other words, deliberately adding more particles (CCN) reduces the fraction of CCN that can activate to become clouds because there is a limited amount of total water content of the air. Because this effect is instantaneous (i.e., it affects the supersaturation at the same time as it changes the cloud albedo), it is generally considered part of the aerosol forcing rather than a separate feedback. This relationship is evident in aircraft-based measurements of CCN proxies and cloud droplets (Leaitch et al., 1992; Martin et al., 1994). This effect means that as aerosol concentrations continue to increase, the corresponding increase in cloud albedo is reduced,
FIGURE 3.15 The relative susceptibility of marine clouds following Oreopoulos and Platnick (2008). Purple indicates regions where clouds are not particularly susceptible to aerosol effects; red indicates clouds that are susceptible.
but because the supersaturations of warm clouds cannot be measured, there are few quantitative observations that can be used to estimate the magnitude of this effect.
Several other limitations are important to consider. For clouds that reflect nearly 100 percent of the incoming visible radiation, the addition of aerosols has little effect on albedo through the Twomey mechanism of changing drop size. For this reason, large cumulus clouds (typically taller than they are wide) and those associated with storm systems and substantial precipitation are not susceptible to aerosol modification of albedo. There can, of course, be other feedbacks in cumulus clouds that change cloud precipitation, extent, or lifetime (Rosenfeld et al., 2013) that have subsequent effects on cloud forcing. In addition, this process is currently better understood for warm clouds (those containing liquid water rather than ice), so high-altitude clouds that are primarily ice have not been targeted until recently (Mitchell et al., 2011; Storelvmo and Herger, 2014). For these reasons, the focus of MCB has been stratocumulus clouds in the planetary boundary layer, typically occurring in the lowest 1.5 km of the atmosphere.
Because the boundary layer is typically well mixed, buoyancy of a particle plume is not required as neutral buoyancy will result in mixing to the height of the temperature inversion. Timescales for this are estimated to be 1 to 3 hours (Lu and Seinfeld, 2006). One important exception is complex, multilayered boundary layers, in which multiple temperature inversions characterize the lowest stratocumulus layer seen by satellite. In this case, particles will typically only mix efficiently within the lowest layer, and yet albedo is often dominated by the topmost stratocumulus layer (Russell et al., 2013), unless it is sufficiently thin as to allow substantial reflection from lower layers. This results in a reduction in the albedo effect of particles.
To date, observational and modeling work has focused most comprehensively on marine stratocumulous clouds. However, there is still substantial uncertainty on the processes that control MCB potential for effectiveness. Additional observations are likely needed to reduce this uncertainty, and studies that provide controlled (or nearly controlled) experiments in the atmosphere are likely to provide better constraints for comparison to model behavior.
There is ample evidence that cloud albedo is strongly affected by aerosol particles and that mankind is able to influence the albedo of clouds. Figure 3.16 shows an example of “ship tracks,” bright areas of clouds produced by aerosol particles in the exhaust emissions of commercial cargo ships which act as CCN in the marine boundary layer off the coast of California. Ship tracks were first reported in satellite observations by
FIGURE 3.16 Ship tracks satellite image retrieved by NASA’s Terra MODIS instrument. SOURCE: http://visibleearth.nasa.gov/view.php?id=66963.
Conover (1966). These plumes are emitted by large, mostly commercial ships motoring at speeds of 20 to 30 kts and emitting particles at rates of 1019 particles/s with ambient windspeeds of 5 to 15 m/s (Hobbs et al., 2000).
There are existing commercial and experiment-specific examples of cloud albedo modification that can be used to provide both an observational signature of cloud albedo modification and proof of concept of the particle emission and scavenging rates that can be expected in typical marine boundary layers.
Three recent experiments promise to provide essential information on uncertainties associated with cloud albedo modification, including the effects of multilayered
clouds in the marine boundary layer: E-PEACE11 off Monterey in 2011 (Russell et al., 2013), SOLEDAD12 in coastal marine clouds off San Diego (Schroder et al., 2014), and MAGIC13 with ongoing transects from Los Angeles to Honolulu. E-PEACE provided detailed evidence of multilayered cloud structure from ship-based ceilometer and aircraft profiles and 94-GHz Doppler radar. Adding to the information collected on these prior campaigns, MAGIC expands the spatial and temporal coverage of this information on marine boundary layer cloud structure with its year-long measurements on California-to-Hawaii transects. The ship-based cloud radar will provide particularly valuable information on the structure of marine boundary layer stratocumulus clouds.
Table 3.1 summarizes several recent experiments that investigate the effects of aerosols on marine stratocumulus. Some of these experiments, notably MAST, E-PEACE, and MASE I/II, focused on the aerosol-cloud interactions from particles emitted by large cargo ships into marine stratocumulus. Although the engine stack emissions of cargo ships are not efficient as a technique for albedo modification because their potential for cooling is largely offset by the enormous CO2 cost of 100,000 gallons of fuel per day, on track-forming days cargo ships may cause twice as much cooling as warming (using a 100-year time horizon; Russell et al., 2013). As such, they provide observational evidence of both individual and overlapping tracks causing cloud albedo modification. Furthermore, the frequency of track formation over ocean regions provides initial statistics that illustrate how often cloud albedo modification is observed (typically 50 percent of cloudy days in some northeastern Pacific regions [Coakley et al., 2000]) despite the continuous presence of cargo ships in many regions. However, such studies also make it clear that current model understanding and predictive capabilities are not sufficient to know either a priori or by satellite retrievals which cloud conditions are “susceptible” (i.e., support the modification of cloud albedo) and which are not.
Russell et al. (2013) studied cloud interactions using controlled emissions of particles from smoke generators on a vessel much smaller than a cargo ship, burning ~500 gallons of diesel per day rather than 100,000 gallons of bunker fuel per day. One interesting result of this study is that the cloud albedo modification was effective only a very small fraction of the time, even in clouds that are classified by satellite and models as likely to be susceptible. This provides preliminary but nonscalable data on how much additional particle emissions would be needed to achieve the intended effect on planetary albedo compared to what is currently implemented in global models. However, the experiment did demonstrate that simple existing technology
11 Eastern Pacific Emitted Aerosol Cloud Experiment.
12 Stratocumulus Observations of Los-Angeles Emissions Derived Aerosol-Droplets.
TABLE 3.1 Selected Relevant Publications from Previous Aerosol-Cloud Interaction Experiments on Marine Stratocumulus
|Experiment||Publications||Key Findings (for aerosol-cloud interactions)|
|MAST||Russell et al., 1999||Observed changes in drop distributions and LWC profile.|
|(NE Pacific)||Hobbs et al., 2000||Drizzle and LWC changes in ship tracks relative to unperturbed clouds.|
|Frick and Hoppel, 2000||Case studies of four ship emissions that produce ship tracks.|
|Durkee et al., 2000||Test of aerosol-induced ship track hypothesis.|
|Noone et al., 2000a; 2000b||Case studies illustrating background pollution effects on albedo sensitivity.|
|Ferek et al., 2000||Ship emission characterization and size distributions.|
|DECS NE Pacific)||Sharon et al., 2006; Stevens et al., 2005||Rift POCs study; variability in cloud drizzle characteristics due to natural processes and emissions.|
|DYCOMS II||Stevens et al., 2003||Characterization of POCs in nocturnal marine boundary layers.|
|(Nocturnal)||Twohy et al., 2005||CN/CCN/CDN relationships are linear.|
|(NE Pacific)||Petters et al., 2006||CCN closure for marine boundary layer particles.|
|Hawkins et al., 2008||Composition independence of particle activation in the aged boundary layer.|
|Faloona et al., 2005||Entrainment rates and variability in the nocturnal marine boundary layer.|
|van Zanten and Stevens, 2005||Drizzle in nocturnal boundary layer in intense precipitation pockets.|
|CIFEX||Wilcox et al., 2006||CCN increases correlated to CDN and reflected radiation for constant LWP.|
|MASE I/II||Hersey et al., 2009;||Ship tracks had smaller cloud drop effective radius, higher Nc,|
|(NE Pacific)||Lu et al., 2007, 2009; Sorooshian et al., 2007, 2009a,b||reduced drizzle drop number, and larger cloud LWC than adjacent clean regions, but trends were obscured by spatial-temporal variability. Aerosol particles above cloud tops are enriched with water-soluble organic species, have higher organic volume fractions, and are less hygroscopic relative to subcloud aerosols.|
|CARMA||Hegg et al., 2009||Source attribution of CCN and aerosol light scattering.|
|Experiment||Publications||Key Findings (for aerosol-cloud interactions)|
|VOCALS-REx (SE Pacific)||Bretherton et al., 2010||Offshore drizzle not explained by CCN decrease.|
|Feingold et al., 2010||Oscillations in aerosol concentrations correspond to precipitation cycles.|
|Wood et al., 2011||POC regions had enhanced drizzle and LWC.|
|E-PEACE (NE Pacific)||Russell et al., 2013||Frequent multilayered low stratocumulus in the marine boundary layer.|
|Sorooshian et al., 2012||Comprehensive cloud drop chemistry sampling.|
|Coggon et al., 2012||Wide-reaching impacts of ship-emitted particles.|
|Chen et al., 2012||Reversed cloud albedo effect in some ship tracks.|
|Wonaschutz et al., 2013||Hygroscopic growth of organic particles below and in cloud.|
|SOLEDAD (NE Pacific)||Modini et al., 2014||Cloud supersaturation and role of sea salt particles as cloud condensation nuclei.|
|Schroder et al., 2014||Role of black carbon particles as cloud condensation nuclei.|
NOTE: LWC, liquid water content; POC, pocket of open cells; CN, condensation nuclei; CCN, cloud condensation nuclei; CDN, cloud droplet number; LWP, liquid water path; MAST, Monterey Area Ship Track experiment; DECS, Drizzle and Entrainment Cloud Study; DYCOMS II, Second Dynamics and Chemistry of Marine Stratocumulus experiment; CIFEX, Cloud Indirect Forcing Experiment; MASE, Marine Stratus/Stratocumulus Experiment; CARMA, Cloud Aerosol Research in the Marine Atmosphere experiment; VOCALS-REx, VAMOS Ocean-Cloud-Atmosphere-Land Study Regional Experiment. SOURCE: Updated from Russell et al., 2013.
can provide a cheap and effective means of cloud albedo modification with a cooling-to-warming ratio of 50:1, as calculated for a 100-year time horizon (Russell et al., 2013).
Since the global mean reflectance scales with global area, the magnitude of the cooling effect will scale with the area of stratocumulus clouds covered as well as with the residence time of the particles. On the average, particles last 5 to 7 days in the troposphere (Seinfeld and Pandis, 2006). However, empirical evidence tracking particle enhancements from ship tracks suggest a typical lifetime of 24 hours with some ranging to 48 and 72 hours (Coakley et al., 1987).
Particles smaller than 1 μm diameter that are emitted near the surface to influence cloud albedo have a lifetime of just a few days. Because aerosol lifetime near the surface is very short, aerosol emissions will remain relatively close to their source (there would not be time for the winds to blow them more than a few hundred kilometers before they are removed by scavenging or deposition), and aerosols would need to be replenished on an ongoing basis over a large area. The footprint of cloud albedo modification of stratocumulus clouds by controlled emissions could involve just one ship (with speed 10 to 20 kts) that can emit particles that will be spread by the ship motion and the wind over 4 to 6 hours to cover an area of 100 km2, as is illustrated schematically in Figure 3.17 and from satellite observations in Figure 5 of Russell et al. (2013). For this coverage, ships on the ocean surface would ideally trace “racetracks” (or zig-zags) separated by 5 to 10 km (depending on crosswind speed). Each ship would trace out a track visible on the Advanced Very High Resolution Radiometer (AVHRR) and MODIS satellite-borne instruments (both of which have daily coverage). However, as Wood and Ackerman (2013) note, quantitative evidence of the aerosol-cloud effects would need to be provided by simultaneous aircraft and ship-based measurements in clean and polluted areas of the cloud and the boundary layer. To account for the large uncertainties in track width and lifetime, the tracks should likely be engineered 2 to 10 times higher in concentration than model-based estimates. Latham et al. (2012) proposed a larger experiment, using five ships to affect clouds covering an area of 10,000 km2. Moreover, since the biggest uncertainty is the cloud type, a hypothetical large-scale deployment of MCB as a global albedo modification strategy would require a large fleet of vessels to be able to deploy in susceptible areas at short notice. The largest cooling effects could be achieved by staging several fleets around the world that are available for deployment on a daily basis and that can be scaled back to reduce energy and emission expenditures when suitable track-forming conditions are not available.
Recent results demonstrate that while both size and composition affect the efficiency with which particles activate to droplets, larger particles, and particles composed of hygroscopic material, are better CCN. Since surface and mass forces make the energetic (and monetary) cost of smaller, more hygroscopic particles more expensive than equivalently good CCN at larger, less hygroscopic compositions, hygroscopicity per se may not be a limiting parameter. Engineering considerations for aerosol production or delivery issues are likely not the limiting factor for achieving MCB albedo modification strategies (Russell et al., 2013). At typical ambient wind speeds in the clean regions of the Pacific Ocean, the types of emission rates that are required are 1017 to 1019 particles/s (Hobbs et al., 2000).
FIGURE 3.17 Annual mean radiative flux perturbation (W/m2) for albedo modification via (a) stratospheric SO2 injection at 2.5 Mt[S]/yr and (b) increasing cloud droplet concentration to 375 cm−3 in the marine stratocumulus cloud sheets at the eastern sides of the North Pacific, South Pacific, and South Atlantic. SOURCE: From Jones et al., 2011.
Latham (1990, 2002) suggested that seawater might be exploited as a source of small seawater droplets to be injected into the boundary layer, where they could evaporate and form small sea salt particles; sulfate aerosols produced by fertilization of marine biota, and organic aerosols produced by combustion have also been suggested (Wingenter et al., 2007). The methods are discussed later in this chapter in the “Delivery Mechanisms” section.
Although it is clear that humanity can, and does, increase cloud reflectivity through aerosol emissions, in this section the committee identifies some of the reasons for uncertainties in estimates of the amount of brightening that might actually be achieved through inadvertent or deliberate aerosol injections.
There are reports of aerosol effects on cloud fraction (Rosenfeld et al., 2013), but there is no evidence that such effects can be sustained without nonaerosol redistribution of water. There are no modeling studies that explain local increases in cloud fraction (i.e., not regionally averaged increases associated with cloud lifetime due to smaller droplet size, e.g., Ackerman and Strabala, 1994) other than by changes in cloud dynamics that also redistribute water (or heat) from a saturated to a subsaturated region.
Regional scaling. Extrapolating the effects of particles on clouds from the microphysical scale to the regional scale is not linear (Martin et al., 1994). The reason that the microphysical effects demonstrated by Twomey may not scale to regions is that the Twomey phenomenon does not take into account mixing and other processes that can dampen and offset the effects measured on small scales. In this case, the committee considers regional scale to be of the order of 500 km2. A single ship track of mean width 10 km that extends 50 km provides such an area. Russell et al. (2013) calculated that at 15 percent brightening (similar to the reflectance changed estimated by Coakley et al.  for typical ship tracks), the cooling is equivalent to 0.4 nK cooling (average cooling over 100 years is calculated by reducing the cooling effect by the ratio of 12 hr/100 yr; CO2 warming is calculated by linearly equating 280 ppmv CO2 with a warming of 3 K, per Solomon et al., 2009).
Competitive effects. Leaitch et al. (1992) demonstrated that higher particle emissions do not result in equivalent increases in droplet number concentrations, because of suppression of supersaturation and other cloud responses. Moreover, Leaitch et al. (2010) and Chen et al. (2012) have shown that adding particles can also decrease the drop number concentrations (a “reverse” Twomey effect).
Susceptibility. Since the increase in reflectance due to drop size and number is only significant for clouds that are not already sufficiently thick (optically dense) that their reflectance may be modified by aerosol particles, the formation of “tracks” with
aerosol-increased albedo depends strongly on the cloud properties (including superstaturation, updraft velocity, and layer structure) as well as on the background aerosol size and concentration (which is a function of wind speed, seawater composition, and wave conditions).
Producing a realistic representation of clouds and aerosols (and their interactions) that strongly affects the albedo of the planet (and indeed many other aspects of Earth’s climate) is a huge challenge for models, contributing to their identification as one of the largest sources of uncertainty in Earth system modeling. Scientists have attempted to improve the understanding of these features in multiple ways:
- Scientists have developed a range of modeling approaches—from detailed process-level models of aerosols and clouds called “box models,” to eddy-resolving “large eddy simulations” (LESs), to kilometer-scale “cloud-resolving models” (CRMs)—with varying levels of complexity in order to focus on different aspects of aerosols and cloud interactions relevant over small time and space scales, exploring these processes in simulations as short as a few seconds to a few days, in air masses ranging from a few meters to a few hundred kilometers.
- When interested in larger space scales, and longer timescales, scientists represent clouds and aerosols (and their interactions) in Earth system and climate models more simply, by “parameterizing” some of the processes, in order to reduce the cost of the calculation sufficiently to make regional or global calculations for days to centuries viable. Modelers “calibrate” the parameterizations with observations and detailed process models so that they agree approximately, but the appropriate representation of these processes remains an incredibly difficult challenge. Some of the resulting issues that are relevant to MCB are discussed later in this section.
Box, LES, and CRM studies. Bower et al. (2006), Feingold et al. (1998), and Russell et al. (1999) are among those to use a box model to study changes in cloud drop number in the presence of extra CCN. LES models were used by Ackerman et al. (1993) to show that aerosols play a role in preventing the collapse of the marine boundary layer in some meteorological conditions, and that ship emissions might act to prevent that collapse and promote cloud formation. The model study provided an early diagnosis of situations where aerosols promote cloud formation. More recently, Wang and
Feingold (2009a,b,) and Wang et al. (2011) used LES to explore the dynamic response of a marine stratocumulus cloud system to background polluted and pristine aerosol levels and local ship emissions. These studies produced a large number of relevant conclusions for MCB:
- Aerosol particle concentrations played a strong role in influencing whether open (low-albedo) or closed (high-albedo) cellular structures formed (with very strong controls on albedo, precipitation, and cloud lifetime).
- Aerosol particle concentrations also influenced the dynamical structures both within the cloud and in the vicinity of (but outside) the aerosol plume, driving the organization of the clouds both local to emissions and in surrounding areas, leading to a “cloud clearing” on the flanks of the aerosol plume, much like those seen in observations.
- Turbulent motions rapidly mixed surface emissions vertically over a few hours throughout the surface boundary layer (typically less than 1.5 km).
- Cross-wind horizontal mixing of aerosol particle emissions was relatively slow, distributing aerosols laterally over about 20 km in 24 to 48 hours.
- Under some circumstances, ship emissions can actually break up cloud structures leading to reduced albedo, although reduced albedo was less common than increased albedo.
- The presence of drizzle prior to the injection of aerosol particles reduced the efficacy of emissions in changing cloud albedo.
Additional modeling studies also indicate that more particle emissions, and more ships than originally estimated by Latham et al. (2008) and Salter et al. (2008), would be required to produce the desired CCN concentrations and marine cloud changes for these cloud types. Jenkins and Forster (2013) considered the change in buoyancy associated with the evaporation of water from the small seawater droplets that form the CCN and noted a measurable reduction in the efficacy of the aerosol source that would result from droplet evaporation (a 2 percent to 10 percent reduction in the albedo increase). Stuart et al. (2013) used ultrahigh-resolution and plume models to account for coagulation of aerosol particles after they were emitted and concluded that plume-scale coagulation could reduce the efficacy of marine cloud brightening by almost 50 percent.
Global studies. Clouds, aerosols, and their interactions are very difficult to represent at the coarse resolution needed for global climate simulations of months, years, centuries, or millennia, and compromises are necessary to implement such simulations. These compromises make it difficult to represent the shallow boundary layer clouds that are
so important to MCB, leading to identifiable biases and deficiencies in their simulation of these clouds (Bony and Dufresne, 2005; Bushell and Martin, 1999; Lane et al., 2000; Roeckner et al., 2006; Sandu et al., 2010; Stephens, 2005; Tompkins and Emanuel, 2000). Each generation of climate model improves both the representation of cloud and aerosol processes as well as the resolution of the model into grid boxes. The fidelity and plausibility of cloud and aerosol processes and features in climate models are slowly improving (e.g., Boucher et al., 2013; Donner et al., 2011; Kay et al., 2012).
The costs and challenges of cloud and aerosol representations in GCMs have led to two approaches for studying MCB in climate models. In the first approach, some important characteristics of clouds are prescribed, for example, by prescribing cloud drop number in clouds. These characteristics are systematically varied to explore the consequences of cloud changes for climate if scientists had “perfect control of cloud properties.” In the second approach, studies are performed that allow the full range of interactions within the climate model to take place by comparing simulations in the presence and absence of particles added at specified times and locations into the model boundary layer.
In the first class of studies, in which perfect control of cloud drop number was assumed, Latham et al. (2008), Jones et al. (2009), Rasch et al. (2009), Hill and Ming (2012), and Baughman et al. (2012) identified specific ocean regions that were represented in GCMs as particularly susceptible to MCB and then prescribed a cloud droplet number increase (different for each model). This produced changes in the cloud radiative forcing, and they then explored the atmosphere, ocean, and cryosphere responses to these changes. All of these studies increased the reflectance of the modeled subtropical marine stratocumulus regions off the west coasts of continents, as well as in some other seeded regions. One consistent response noted by many of these studies was a persistent cooling of the Pacific, similar to the “La Niña” phenomenon. All simulations indicated global mean cooling and an increase in polar sea ice, in spite of the regional nature of the albedo change.
Jones et al. (2009) increased cloud drop number in three regions of marine stratocumulus (around 3 percent of Earth’s surface area) and found that up to 35 percent of the radiative forcing due to current levels of greenhouse gases could be offset by a very aggressive level of stratocumulus modification (~1 W/m2) that delayed the warming by ~25 years (average reduction in energy reaching the surface of the seeded regions of about 30 W/m2). They also noted significant shifts in important precipitation patterns, with increases in some regions and decreases in others (for example in the Amazon). However, these regional precipitation pattern changes are not found consistently across studies with other models.
Korhonen et al. (2010), Partanen et al. (2012), Jones and Haywood (2012), Alterskjær et al. (2012), and Alterskjær and Kristjánsson (2013) relaxed the constraint of the first-generation MCB studies by exploring responses to additions of sea salt particles at the surface. They all found significant cooling effects on the climate due to cloud albedo being increased by the aerosol indirect effect, but large differences in the spatial distribution of the temperature changes were found. This type of difference in predicted regional responses among models is not surprising because such differences are also seen in comparing simulations of precipitation changes due to global warming, since the processes that control precipitation are very uncertain in GCMs. Partanen et al. (2012) and Jones and Haywood (2012) also assessed the role of the direct radiative impact by the sea salt aerosols and found it to contribute significantly to the total radiative impact.
Some of these responses and feedbacks may be model dependent, and studies have not used a common experimental design, making comparison of the different studies difficult. Alterskjær et al. (2013) attempted to reduce these differences in a model intercomparison using a common experimental design to search for robust responses across three Earth system models (a similar but larger model intercomparison is now taking place under the GeoMIP program [Kravitz et al., 2013a]). In these studies, sea salt aerosol emissions between 30°N and 30°S were increased to offset the forcing from an RCP4.5 scenario between 2020 and 2070. The increased emissions were then terminated to explore the rebound effect. The models studied by Alterskjær et al. (2013) still had significantly different mechanisms for addition of sea salt particles, but forcing amplitudes and forcing mechanisms are closer than previous studies. Some models prescribed aerosol distributions and did not allow cloud processes to remove aerosols; others allowed those interactions to take place and used more complex treatments of aerosol-cloud interactions. Each model accounted for some direct and indirect radiative effects of the emitted sea salt aerosol particles. Each model had significant differences in formulations of aerosol-cloud interactions (some included only the effect of drop radius first studied by Twomey et al. ; others included aerosol effects on precipitation microphysics discussed by Albrecht ) and differing feedbacks, necessitating different increases in sea salt concentrations to cancel the forcing. Each model required a different amount of emitted sea salt aerosol particles increased to counter the greenhouse gas (GHG) warming in the decades around 2060. Some of these differences are summarized in Table 3.2.
For the final decade of the simulations (2060-2070) before terminating the sea salt aerosol particle emissions, the NorESM required an increase by a factor of 3.4 in emissions of the 0.13-μm sea salt particle mode but only a 3.4 percent increase in the total sea salt emission mass flux (equivalent to a fleet of about 7,600 injection vessels, assuming that these have the design and efficiency proposed by Salter et al. ).
TABLE 3.2 Results from Intermodel Comparison Involving Three Earth System Models (MPI-ESM, IPSL-CM5a, and NorESM) Used to Explore Differences among the Models
|Model||Equivalent Sea Spray Emissions (Mt/yr)||Average Surface Temperature (K) and Precipitation (mm) Change from GHG Forcing (2060-2020)||Average Surface Temperature (K) and Precipitation (mm) Change Produced by the Combination of GHG Forcing and MCB Albedo Modification|
|MPI-ESM||316||(+0.9, +0.04)||+0.2, −0.01|
|IPSL-CM5a||560||+1.3, +0.09||+0.2, −0.02|
|NorESM||266||+0.8, +0.05||+0.2, +0.01|
NOTE: The different emissions needed to counter GHG warming are due to differences in the fraction of low clouds in the seeded regions and differences in treatment of the effect of the injected sea salt on precipitation release. MPI-ESM = Max Planck Institute for Meteorology Earth System Model; IPSL-CM5a = Institut Pierre Simon Laplace Climate Model; NorESM = Norwegian Earth System Model. SOURCE: Alterskjær et al., 2013.
The different emissions needed to counter GHG warming are due to differences in the fraction of low clouds in the seeded regions (producing changes in cloud albedo over a relatively smaller area) and differences in treatment of the effect of the injected sea salt on precipitation release (Albrecht, 1989), affecting the cloud lifetime and areal extent. As in the idealized studies described earlier, all three models employing MCB produced reduced evaporation, particularly from low-latitude oceans, and reduced precipitation over low-latitude oceans and storm-track regions compared to the simulations with forcing only from greenhouse gases. But in contrast to studies with uniform sunlight reduction, each model produced increased precipitation, cloud formation, and precipitation over low-latitude land regions in response to the localized cooling over the low-latitude oceans, reducing aridity in many low-latitude land regions as well as in southern Europe (Alterskjær et al., 2013). (This result is consistent with the idealized study of Bala et al.  employing sunlight reduction only over ocean.)
Jones et al. (2011) directly compared the model differences in forcing and response between stratospheric aerosols and marine cloud brightening. Forcing differences are shown in Figure 3.17.
Models consistently indicate that MCB can reduce temperatures. Model simulations show that MCB targeted at susceptible marine stratocumulus will cool preferentially the eastern North and South Pacific and eastern South Atlantic, and will also cool globally and reduce Arctic warming. These results must be viewed with some caution. Cloud models and global model parameterization of marine stratocumulus remain
much simpler than in the real world, and scientists recognize that they do not yet provide robust quantitative predictions of cloud responses to aerosol changes. There are still significant disagreements between model estimates of the Twomey effect compared to estimates from satellite measurements. Global models disagree in their predictions about MCB effects on the spatial distribution and intensity of precipitation, particularly when the MCB global average forcing exceeds 0.5 W/m2 annual average.
There are three types of observations that are needed to track and quantify the radiative effects of particles on cloud albedo: satellite reflectance sensors, described below; in situ aerosol and cloud instrumentation; and logistical metrics. For quantifying ecosystem impacts of cloud albedo modification, monitoring networks for nutrients and biota, as well as case-specific integrated modeling of potential teleconnections, are required. Technology for all of these aspects exists. Experiments that require large-scale, multigroup efforts using national aircraft and ocean research facilities could range in scale from $10 million to $100 million depending on the target region and time of the modification. An example of a small-scale experimental design is provided by E-PEACE, as shown in Figure 3.18.
Variants of this type of study designed to provide additional information about engineering issues and cloud responses to aerosol injection are described in two recent papers proposing field studies that might be used to extend previous work (Latham et al., 2012; Wood and Ackerman, 2013). These studies suggested a series of three staged field experiments that are successively more ambitious. The smallest field experiment would follow particles explicitly designed to be good CCN, monitoring size distribution, chemical composition, and cloud-forming properties close to the injection source, and their destiny as they disperse downwind in the boundary layer. The second would explore possible cloud responses to the injected aerosol using multiple aircraft and ships in a range of conditions and model those specific situations to see whether the models were capable of reproducing the observed aerosol and cloud evolution. The third would examine the impact of multiple injection sources over a limited area (perhaps 100 × 100 km2) to characterize effects on cloud albedo and cloud forcing. Other variants are mentioned by Keith et al. (2014).
The evidence for cloud albedo modification is clear in AVHRR (on NOAA satellites) and MODIS (on NASA’s Terra and Aqua satellites) after proper data post-processing (Durkee et al., 2000). Such signatures are not evident in high-traffic areas (Peters et al., 2011), making regional signatures difficult to detect. Tropical regions also lack consistent signatures
FIGURE 3.18 Illustration of E-PEACE design and observations of emitted particles in marine stratocumulus in July and August 2011 west of central California. The diagram shows the three platforms used in making observations of particle and cloud chemical and physical properties, namely, the R/V Point Sur, the Center for Interdisciplinary Remotely-Piloted Aircraft Studies (CIRPAS) Twin Otter, and the A-Train satellites and GOES. The design included using smoke generated on board the R/V Point Sur that was measured after emission by the CIRPAS Twin Otter in clouds. The satellite was used to measure the changes in reflectance of sunlight due to the effects of the emitted particles on the clouds. The counterflow virtual impactor (CVI) was used as an inlet for evaporating droplets as they were brought into the aircraft, allowing sampling of droplet chemical composition. SOURCE: Russell et al., 2013.
(Peters et al., 2014). With modern visible imagery and ship tracking (e.g., marinetraffic. com), the most obvious evidence for cloud albedo modification can be collected from emissions from controlled ships that “zig-zag” back and forth instead of transiting efficiently from one port to another (typically along standard shipping routes). Ships large enough to emit particles in midlevel seas (such as the R/V Point Sur, 135 ft long and 298 gross tons, http://marineops.mlml.calstate.edu/PS-Specs) are trackable with this existing technology; smaller or fuel-free ships (such as those proposed using Flettner rotors by Salter et al. ) would still be trackable based on required route reporting at ports of call near the targeted region during susceptible cloud conditions. This type of ship activity would be a clear logistical signature of medium- or large-scale MCB deployment.
The CALIOP instrument on board the CALIPSO satellite identifies cloud and aerosol layers using polarized lidars at 532- and 1,064-nm wavelength. Overcast stratocumulus is sufficiently optically thick to extinguish the lidar before it reaches the surface, although in broken or scattered stratocumulus some fraction of the lidar backscatter will originate from the ocean surface.
In summary, the logistical signatures for ship traffic, fuel purchases, and other port activities likely will provide clear evidence of MCB activities. Satellite instrumentation exists to effectively monitor approximate changes in reflectance for large-scale marine cloud brightening activities. However, scientists lack sufficiently high temporal- and spatial-resolution measurements of albedo to enable us to separate the radiative changes of MCB from the natural variability. In situ observational instrumentation exists that could be deployed to effectively observe marine cloud brightening activities but they require expert operators and nonroutine analyses, likely located in open-ocean regions offshore that may be difficult to access.
As described in the previous sections “Observations of Marine Cloud Brightening” and “Modeled Climate System Responses to Marine Cloud Brightening,” there is some potential for undesirable side effects from MCB activities, repeated here for the reader’s convenience. In particular, there is some potential for changes to precipitation patterns and amplitude (Bala et al., 2011; Jones et al., 2011; Rasch et al., 2009) and possibly for interannual variability (Russell et al., 2012), although modeling studies suggest the residual changes are likely less than those for stratospheric aerosol albedo modification and much smaller than for unabated greenhouse gas warming. As in the SAAM and idealized albedo modification strategies, MCB cannot return both temperature and precipitation patterns to preindustrial conditions, and residual temperature changes will also remain; for example, the tropics may cool more than the polar regions (see studies cited above, and Ricke et al., 2010; Tilmes et al., 2013).
MCB activities might introduce changes to the marine and terrestrial ecosystems through changes to clouds and cloud area that reduce the surface flux of sunlight. Changes to the albedo of stratocumulus clouds are likely to substantially alter the surface flux of sunlight; Latham et al. (2008) and Jones et al. (2011) estimated that using MCB at amplitudes sufficient to alter climate would decrease annual mean sunlight reaching the surface by 30 to 50 W/m2 (~20 percent, approximately doubling cloud radiative forcing) locally in seeded regions. These changes in surface energy fluxes are likely to reduce local sea surface temperatures (e.g., Rasch et al., 2009) and their gradients, perhaps influencing important climate modes such as El Niño, and might also change deep ocean upwelling and mixing in the ocean surface layer that delivers nutrients to marine ecosystems, with possible effects on ecosystem services such as fish availability. These marine ecosystems also contribute to the natural aerosol concentrations in near-marine regions that are important in cloud formation (Quinn et al., 2014), so there may be feedback effects as well. Last, the change in sunlight reaching the
surface may influence photosynthesis, and the potential cloud area changes from the brightening are likely to matter most. Changes to cloud opacity are unlikely to influence photosynthesis as strongly as changes to cloud lifetime or areal extent because photosynthesis is only weakly dependent on (direct and diffuse) sunlight intensity. These issues have not yet been explored in models or observations (via ship track or other studies), so potential consequences to ocean ecosystem productivity are very uncertain (Russell et al., 2012). In addition, it is important to recognize that impacts on ecosystems change with the scale of the intervention. Stafford-Smith and Russell (2012) have suggested that regional rather than global deployment of MCB or SAAM (or both) methods might have less serious negative consequences for ecosystems, since the complexity of larger systems provides some degree of resilience.
The use of NaCl or sea salts as the emitted particles would result in increased salt deposition, possibly affecting the salinity of the ocean surface layer in the regions in and surrounding which MCB is deployed. More needs to be done to improve estimates of the impact of deposition on downstream coastal and other continental ecosystems and to evaluate toxicity.
There are important open questions for the feasibility of MCB at deployment scale. Theory, modeling, and observations indicate that the susceptibility of cloud albedo to increases in aerosol particle concentrations saturates, but the point of diminishing returns varies with cloud type and background aerosol amount. The natural variability of clouds is high, and many different cloud regimes exist that may respond differently to aerosol increases, complicating signature detection and making quantitative characterization of cloud susceptibility and effective radiative forcing (ERF) difficult. Since these differences in cloud responses are not well represented in models, observations are needed to improve our ability to quantitatively constrain these differences.
There are only a few situations (e.g., ship tracks) where there is clear evidence that the albedo of a specific cloud has been influenced by local variations in aerosol. In larger-scale cases, estimation of aerosol impacts on cloud properties requires a statistical analysis of a cloud system, generally in the absence of a systematic and quantitative method for varying aerosol concentrations near the cloud system, or a control to monitor similar cloud characteristics in the absence of a perturbation. This means that, to date, all estimates of the feasibility of MCB are restricted to (1) scale-up of simplified parameterizations by global models, (2) process-based models with limited larger-scale interactions or validation, (3) monitoring the response (or lack of response) of
an individual cloud to particles released as pollution in ship of opportunity studies such as MAST (see Table 3.1), and (4) monitoring the response (or lack of response) of clouds to a smaller emission source of particles in a field experiment (E-PEACE) with different physiochemical properties than sea salt, or the combustion particles produced by shipping.
Delivery mechanisms. Although aerosol production and delivery issues are not expected to be the limiting factor for implementing MCB albedo modification strategies (Russell et al., 2013), at least three methods have been considered for delivering suitable aerosols into the marine boundary layer to brighten clouds. The first two methods may prove to be cheaper and have fewer unintended consequences than the third, but they rely on technology that requires development and scale-up.
- Latham (1990, 2002) suggested that seawater might be exploited to produce small seawater droplets and inject them into the boundary layer, where they could evaporate and form small NaCl-dominated particles; Salter et al. (2008) suggested methods and devices that might be used (but do not yet exist) to produce and deliver droplets into the marine boundary layer. Neukermans et al. (2014) and Cooper et al. (2014) discuss a prototype device in the laboratory capable of producing seawater droplets of the appropriate size range that may be able to be scaled up to rates relevant to field studies (e.g., ~1 × 1018 s−1).
- Wingenter et al. (2007) suggested the use of dimethyl sulfide produced by fertilization of ocean biota as a source for CCN, although doubts about the method’s efficacy have been voiced (Vogt et al., 2008; Woodhouse et al., 2008).
- Engine or smoke emissions could also be used as a source for CCN. Freighter emissions producing ship tracks indicate that combustion is an effective source of aerosols, although ship emissions were never designed or optimized for this purpose. E-PEACE (Russell et al., 2013) demonstrated that paraffin oil particles (e.g., material used for skywriting) could also be used effectively. Military-issue “smoke generators” are available that produce these rates at a CO2 cost substantially lower than the exhaust from cargo ships. These typically use a high-boiling-point, unreactive hydrocarbon mixture such as paraffin oil (used commercially for transformers and for sky writing). The National Institute of Standards and Technology designates paraffin oil as environmentally “benign.”14
The inherent problem in designing emissions for MCB is that producing submicron particles requires energy to produce particles with very high ratios of surface area to volume from a bulk liquid or a gas. Particle production from chemical reactions such as combustion uses chemical energy to make submicron particles; nozzle or spray technologies typically use mechanical force (pressure) to make small particles. In order to make MCB cost effective (in terms of both dollars and fuel usage or equivalent CO2 emissions), the energy from particle production must be minimized. This constraint tends to favor particle production from phase changes or chemical reactions in situ (such as condensation of vaporized paraffin oil in a smoke generator) due to the engineering considerations in marine conditions, such as clogging from impurities of source material.
Efficacy. Current estimates of the long-term and large-scale efficacy of the MCB strategy (e.g., the radiative forcing per unit aerosol emission for different marine cloud regions) are generally based on theory and modeling studies, and they are not yet sufficient to provide robust estimates for radiative forcing or to identify limitations of the strategy, consequences to the development and evolution of cloud systems, possible far-field effects, or longer-term climate consequences involving feedbacks. In spite of these uncertainties, estimates have been made. Results are reported in a variety of “units.” Sometimes the measure is expressed in terms of the emission rate (particles m-2) times the area seeded (m2) to achieve an effective radiative forcing sufficient to counter that from a doubling of CO2. Latham (2002) initially estimated an injection rate of particles of ~3 × 106 m−2 s−1 over a surface area of ~77,000 km2.15 Salter et al. (2008) revised that estimate to 1.5 × 106 m−2 s−1 over a similar area. These estimates are equivalent to a local increase in cloud albedo in marine stratocumulus regions of about 0.06, and for the purposes of comparing results between global models and high-resolution models it is sometimes easier to work in units of albedo. But these estimates are largely based on limited-scale observations, and the scale-up is likely to result in diminishing efficiency due to the reduced efficiency of Twomey effects at high aerosol concentrations.
More complete treatments of aerosol cloud interactions in modern climate model studies indicate the need for larger (Korhonen et al., 2010) and smaller emission rates (Alterskjær et al., 2012; Partanen et al., 2012) than the Salter estimate, but these estimates also ignore a variety of cloud and aerosol processes that may be important. Russell et al. (2013) showed that, even in a regime that is considered to have a high
15 An area almost as large as the size of South Carolina (U.S. Census Bureau, 2012); South Carolina’s total area is 82,933 km2.
potential for MCB, the apparent susceptibility of clouds that is represented in models and observed from satellite can be reduced by factors of 2 to 10 due to multiple cloud layers, subgrid-scale drizzle, local clearing, and limited mixing. These studies highlight uncertainties in issues critical to a quantitative characterization of MCB, and the need for laboratory and field work.
Costs. Table 3.3 provides estimates of the potential costs and resources required for various levels of cloud albedo modification activity.
Research beyond the use of computational models is needed to address some of the key open questions on the potential for marine cloud brightening to be useful for albedo modification purposes. The reason is that the uncertainties of cloud susceptibility, scale-up, and feedbacks are not sufficiently understood to be included with confidence in models. These issues produce the largest uncertainty in quantifying marine cloud brightening feasibility and, hence, assessment of cost and risks.
An improved ability to characterize aerosol cloud interactions is needed. Field studies, improvements to model physics, and improvements in the agreement of models with measurements play a key role in demonstrating the understanding of these basic climate processes and help in characterizing MCB potential for albedo modification.
The committee identifies a number of research needs to address the current gaps in understanding of the efficacy and effects of MCB.
Field studies. Previous climate-focused field studies have produced substantial progress in understanding the aerosol-cloud interactions that are of relevance to MCB, but there are still aspects of these interactions that require better characterization. Field studies near existing uncontrolled emission sources provide very useful information and can be evaluated to see the extent to which observed albedo response matches modeled albedo response over some space and timescales. Some issues, however, can be more clearly exposed and understood using deliberate, controlled emission studies. In combination with each other, these observational strategies provide fundamental information on aerosol direct and indirect effects and boundary layer transport that are very important, but crudely treated in current atmospheric models. Together they also serve as a verification and calibration data set for models.
TABLE 3.3 Logistical Footprint at Various Scales for Hypothetical Cloud Albedo Modification
|Likely logistical signature for 1 year||0.0001 W/m2||0.01 W/m2||5 W/m2|
|Dollars required/expended* (expenditure breakdown: 80% for fuel; 10% personnel; 10% aerosol production material/maintenance)||$50k/week||$5M/week||$100M/week|
|Hardware (based on 300T ships with speed 10 kts)||1||100||2,000|
|Footprint (pattern: parallel tracks at 10-km spacing; location: ocean surface in marine stratus cloud regions: SEPac, NEPac, SEAtl)||5 km × 50 km nonoverlapping||100 km × 100 km||20 each 100 km × 100 km|
|People required (seamen, engineers, and technicians)||10 people||1,000||20,000|
|Fuel usage (given current ship technology)||500 gal/day||50,000 gal/day||1,000,000 gal/day|
NOTE: Scaling of costs is assumed to be approximately linear in forcing due to trade-offs between increasing economies of scale and decreasing cloud susceptibility and accessibility. Costs for 5 W/m2 are based on GAO-estimated $5 billion annual cost (GAO, 2010). Costs for smaller-scale deployments are scaled linearly. Costs for 0.0001 W/m2 are comparable to E-PEACE deployment. SOURCE: Russell et al., 2013.
Opportunities to improve understanding of relevant processes that can potentially be revealed much more clearly with small-scale controlled emissions studies include the following:
- Comparing to a control. Monitoring adjacent air masses or air masses prior to and following emissions would serve as an experimental control to contrast with the seeded clouds, and monitoring both the perturbed and control air masses would help identify the sensitivity to preexisting air mass properties (e.g., aerosol amount).
- Tracking changes in a cloud system. Extended monitoring of the properties of aerosols and clouds in regions after controlled emissions of aerosols are released, and in control regions, would provide information about the evolution of the size and composition of the introduced aerosols, and the possibility of dynamic responses to the seeding (evidence for cloud clearing) would also be useful.
- Testing in different regions and seasons. The dynamic responses to particles will vary for different regions and seasons of stratocumulus cloud. The boundary layer properties (including cloud height and thickness, number of layers, degree of decoupling, strength of inversion, subsidence rate, vertical velocities, and entrainment) may all be important factors in the amount of brightening and its persistence. Controlled in situ measurements in different regions would provide much more precise information and insights not available from satellite observations or opportunistic field studies.
- Evaluating differences in emission strategy. Studies using deliberately controlled emissions for hours or possibly for days, covering regions of varying areas, differing release durations and start times, or changing particle types would provide observations of the resulting differences in dynamic responses to seeding, providing information on cloud clearing, sensitivity to diurnal variation in the boundary layer, sensitivity to composition or size distribution of emissions, and so on. These effects probably operate nonlinearly to dampen or increase the brightening. Interactions between multiple adjacent seeded regions may also change the expected brightening.
Model studies. Models disagree with each other, and with observations of clouds, aerosols, and their interactions. These specific studies are recommended:
- Designing model studies to attempt to reproduce the field studies discussed above (particularly the controlled emission studies) could help reveal specific reasons for discrepancies, leading to improved parameterizations.
- Better intercomparisons between climate models using varying treatments of aerosol microphysics, employing scenarios that are more strongly constrained (in terms of the type, amount, and altitude of aerosol emissions) than have been hitherto performed by the GeoMIP studies, would help in understanding the reasons for climate simulation differences that lead to model uncertainties and their projection of climate consequences.
- Intercomparison between detailed models would be useful to resolve critical features and would provide benchmark simulations for the simpler formulations used in global models.
- Comparison between global-scale model formulations of aerosol, clouds, and aerosol dispersion in the subcloud layer, with finer-scale models (LES, aerosol dynamics, and plume models) could be useful. Such comparisons would challenge the simplified formulations present in global models with the much more detailed formulation present in the fine-scale models.
- There has not yet been any exploration of sensitivity of model response to model resolution, or the numerical methods used to solve the equations describing the important processes and their interactions. Studies of these aspects would eventually be important to ensure that predictions of model change are robust.
As with SAAM studies, there are many potential climate impacts from MCB that are essentially unexplored, and more attention is merited with both models and possibly field experiments if they can be done at smaller scales. The committee is specifically aware of a lack of knowledge about (a) impacts on ocean circulations, (b) consequences to ecosystems due to significant reductions in sunlight reaching the surface where MCB is operating, (c) interactions of MCB with dominant modes of interannual variability like ENSO and the Pacific Decadal Oscillation (PDO), and (d) the nature of the remote impacts to precipitation like that found in the U.K. Met Office model discussed previously (Jones et al., 2013). These processes are all likely to operate at longer timescales and be sensitive to forcing on larger space scales and should also be explored.
There are a number of other proposed techniques that are often considered in discussions of climate intervention broadly that also have to do with modifying the albedo and/or radiation balance of the planet. The proposals in this section have generally shown less promise in initial studies, are less developed than the ones described in the earlier sections of this chapter, or are only mentioned in passing in the literature. In particular, not enough is yet known about cirrus cloud modification to warrant a more extensive discussion at this time, although this proposed technique may have potential. Even though time and cost issues may differ among the specific technologies, those differences are at extents that are not yet well quantified due to the limited current state of development.
There have been several proposals in the literature for placing scatterers or reflectors of some kind in space to reduce the amount of sunlight entering Earth’s atmosphere. The options include a large opaque disk, a large transparent prism (Early, 1989), a large sail (NRC, 1992), a large diaphanous scattering screen (Teller et al., 1997), a large iron mirror (Mcinnes, 2002), trillions of small spacecraft (Angel, 2006); and a large ring of space dust (Pearson et al., 2006). The objects could be placed in low Earth orbit or at
the L1 point.16 Several of these ideas require the ability to manufacture in space, making them impractical at the current time. Overall, the committee has chosen to not consider these technologies because of the substantial time (>20 years), cost (trillions of dollars), and technology challenges associated with these issues (GAO, 2011; The Royal Society, 2009).
Several techniques have been proposed as potential mechanisms for increasing the albedo of the planet’s surface, including painting the roofs of large numbers of buildings white, planting crops with higher albedos, covering deserts or other surfaces in highly reflective materials, and generating small bubbles in the ocean to brighten the ocean surface. In general, these techniques are judged to be of low potential use on the global scale because of generally low effectiveness and high costs. Several of these techniques are discussed as “soft geoengineering” (Olson, 2012) because of their low overall risk; that is, the implementation of any of them is easily reversible (e.g., painting roofs back to their original color, replanting original crops, and uninstalling reflectors). There is little to no research demonstrating the practical effectiveness of these techniques and little new research in these areas; the committee summarizes the arguments presented in other assessments.
White roofs. Painting rooftops and road surfaces white in urban areas has been proposed to increase the reflectivity of Earth’s surface (Akbari et al., 2012; Lenton and Vaughan, 2009). This approach would have the potential co-benefit of reducing the need for air conditioning in sunny regions in the summertime, although there are questions about its potential impacts on local moisture and energy transport (Olson, 2012). Although this approach does not require the development of new technologies, it involves large costs, both for initial painting and maintenance, and is limited by the available surface area, which is on the order of less than 1 percent of Earth’s surface. All published estimates in the previous literature suggest that changing planetary albedo by whitening rooftops cannot compensate for a significant fraction of the forcing produced by present or future anthropogenic forcing by greenhouse gas emissions (e.g., GAO, 2011; The Royal Society, 2009).
16 The L1 point is the point between Earth and the Sun where the gravitational attraction between the two bodies is equal, approximately 1.5 million km from Earth toward the Sun.
Bright crops. It has been proposed that specific choices for crop varieties (Ridgwell et al., 2009) or grassland, shrubland, or savannah species could increase planetary albedo (Hamwey, 2007). There are associated risks to making large changes to ecosystems (The Royal Society, 2009) and, even if done on a large scale, current estimates suggest these approaches are limited in the maximum amount of cooling they could produce globally (GAO, 2011; Lenton and Vaughan, 2009). Such methods may produce significant regional cooling potential that could be used as part of local adaptive measures (Ridgwell et al., 2009).
Reflective materials on surfaces. Deserts cover large land areas and generally are found in areas that receive large amounts of incident sunlight. Reflective material placed over large deserted areas could increase the albedo substantially (from 0.4 to 0.8 according to Gaskill ) and potentially have a large impact on the radiative budget of the planet. The costs of such an approach are likely to be very high (The Royal Society, 2009), and although the technology appears plausible, no demonstration of the technology has yet been reported as of the GAO (2011) report. There may be significant maintenance costs for keeping the reflective surfaces clean. There are also serious unanswered questions about how this would affect desert ecosystems as well as atmospheric circulation and precipitation patterns, including potential effects on monsoons (The Royal Society, 2009).
In addition, there has also been at least one proposal to counteract melting polar ice and thawing permafrost by spreading disks of light-colored material to increase the albedo of areas of open water or specific areas in danger of melting, but there are still significant uncertainties about the effectiveness of this approach (Olson, 2012).
Microbubbles. A 1965 President’s Science Advisory Committee report (PSAC, 1965) discussed floating small reflective particles over large oceanic areas to change the amount of reflected sunlight from the surface. Most observers think that this would be difficult to do in practice for many reasons, among them convergence of ocean currents and possible biogeochemical effects. A more recent proposal has been put forward to create microbubbles just under the surface of the ocean that could last for long periods of time to increase the albedo of the ocean’s surface (Seitz, 2011). Such a suspension of voids is referred to as a hydrosol. There is very little published research on this idea, but in theory this approach would have the benefits of being local in scale and easily reversible (Olson, 2012). Evaluating the potential effectiveness of microbubbles requires significant further research, particularly into overcoming and
optimizing variable microbubble yields and lifetimes, as well as further understanding of risks to phytoplankton ecology and biogeochemical cycles (Seitz, 2011).
Modification of cirrus clouds is an alternative to planetary albedo modification methods, the focus of this report. Found in the very cold upper half of the troposphere (typically above 440 hPa, varying with latitude), cirrus clouds are composed almost completely of ice crystals and have a thin wispy appearance. Cirrus clouds absorb a fraction of the long-wavelength radiation (wavelengths of 2 to 25 μm) flowing up from the surface and the lower atmosphere and emit this absorbed energy as long-wavelength radiation upward, lost to space, and downward, contributing to greenhouse warming. Cirrus clouds also contribute to the planetary albedo by reflecting a fraction of the incoming solar (short-wavelength) radiation. Overall, the greenhouse warming contribution (which operates continuously over the whole globe) dominates the albedo contribution from cirrus (which operates only on the half of the globe in sunlight) (Chen et al., 2000; Hartmann et al., 1992; Liou, 1986).
Recent studies have suggested it might be possible to cool the planet by decreasing the opacity, frequency of occurrence, areal extent, and/or duration of cirrus clouds, thus increasing the fraction of the long-wavelength radiation flowing up from the surface and lower atmosphere on to space. While albedo modification techniques would operate only during the day and would be most effective around the equator, cirrus thinning could continuously affect the whole globe (but research shows it is most effective at high latitudes [Storelvmo and Herger, 2014]). In essence, albedo modification decreases the rate of heating of the planet while cirrus modification increases its rate of cooling.
Mitchell and Finnegan (2009) have suggested that the highest and coldest cirrus could be targeted for thinning by introducing aerosols that act as ice nuclei, producing ice crystals that grow rapidly and deplete water vapor, suppressing nucleation and growth of ice crystals that form by other means (homogeneous nucleation). They suggest using bismuth tri-iodide (BiI3) as the ice nuclei, which is nontoxic and relatively inexpensive (Pruppacher and Klett, 1997). Published estimates by Mitchell and Finnegan and most recently by Storelvmo et al. (2013) and Storelvmo and Herger (2014) suggest that small increases to long-wavelength radiation to space could offset the enhanced radiative forcing due to a CO2 doubling.
As discussed by Cotton (2008), the possible adverse consequences of seeding cirrus clouds to increase the outgoing long-wavelength radiation from the lower atmo-
sphere and surface are most likely impacts on the hydrological cycle. Cotton indicates the need for chemical, cloud-resolving, and global models to evaluate the feasibility of this approach and to estimate possible adverse consequences. He judges the feasibility of this approach in terms of implementation strategies as being comparable to seeding sulfates in the lower stratosphere and suggests the costs would be similar to Crutzen’s estimates for stratospheric seeding (Crutzen, 2006).
In a more recent modeling study, Storelvmo et al. (2014) found that seeding of mid- and high-latitude cirrus clouds had the potential to cool the planet by about 1.4 K, and that this cooling is accompanied by only a modest reduction in global rainfall. Intriguingly, and suggestive of the complexity of such modifications, seeding of the 15 percent of the globe with the highest solar noon zenith angles at any given time resulted in the same global mean cooling as a seeding strategy that involved 45 percent of the globe. In either case, the cooling was found to be strongest at high latitudes and could therefore serve to prevent Arctic sea ice loss.
Scientists have only a limited understanding of the physical and dynamic processes influencing formation, maintenance, and dissipation of cirrus clouds. Perhaps most critical to current research, there are significant uncertainties associated with ice nucleation in cirrus clouds and its proper representation in numerical models. Further research is required to be able to assess the potential viability of cirrus cloud modification as a response to climate impacts (Storelvmo et al., 2014). This includes improving the understanding of cirrus clouds through observations, better modeling to understand the role of cirrus clouds in the climate system and expected regional temperature changes from cirrus cloud dissipation, and determining whether cirrus cloud modification is feasible and effective as a climate intervention method with fewer negative consequences than other approaches. Research supporting possible cirrus cloud modification will also be relevant to better understanding the effects of stratospheric aerosol injection—either from volcanic eruptions or from stratospheric albedo modification efforts—because these aerosols will eventually settle out of the stratosphere into the upper troposphere where cirrus clouds reside (Cirisan et al., 2013; Kuebbeler et al., 2012). If deployment were to be evaluated, then development and testing of tailored seeding agents and delivery systems to optimize the dissipation of the cirrus cloud, including addressing the suitability for agents for multiple types of cirrus and understanding of the fate and impacts of seeding agents (evaporation versus falling out), would need to be undertaken.
Social and political challenges to cirrus modification research or eventual deployment are likely to be similar to those faced by proposed albedo modification techniques. These may come from some in the environmental community but also from the many
The success of society in the face of a changing environment relies heavily on an effective observational capability to document and understand change, as well as to inform strategies to address change. The need for a robust observing capability becomes significantly amplified with the implementation of or experimentation with albedo modification methods, given that the indirect effects could be of greater impact than the direct effects, and they may well be unanticipated. The use of an engineered increase in albedo to offset the effects of anthropogenic CO2 increase is fraught with uncertain outcomes that could potentially be much worse than the problem it seeks to address. As a result it is critical that any such undertaking requires a monitoring plan that provides a continually updated assessment of whether the benefits are likely to be greater than the adverse effects. The successful observational strategy would require four elements: (1) monitoring large-scale direct effects, (2) monitoring large-scale indirect effects, (3) intense local process observations to inform models, and (4) capability to detect unilateral and uncoordinated deployment.
A minimal requirement for controlled deployment of a climate intervention involving albedo modification is that one be able to detect and characterize the actual change in albedo achieved by the intervention. This is crucial, because the chain of physical processes linking the controlled injection of a substance into the atmosphere to the resulting change in albedo is so complex, and involves so many stacked uncertainties, that it is unlikely to prove possible to accurately compute the albedo change a priori. It would be incumbent upon those who deploy an albedo modification technique to assess how well the target value is met. Accurate albedo monitoring is also a requirement for a broad class of field experiments aimed at testing albedo modification technologies, though there may also be experiments that yield useful scientific payback without producing a detectable change in albedo.
Satellites are the preferred platform for observation of large-scale albedo changes, because of their near-global coverage, but albedo observations from space pose a considerable challenge. These include converting observations from a single or limited
number of viewing angles to total reflected energy using complex empirically tuned assumptions about the angular distribution of reflected radiation (Loeb et al., 2012); determining the full diurnal cycle based on incomplete sampling by satellites of diurnal variability; maintaining accurate calibrations to account for instrument degradation over time; and merging and intercalibrating observations from different satellites with different orbits at different times, in order to achieve the long-term records necessary do quantify and understand trends.
Any albedo modification, if deployed, should start with an intervention of small magnitude—with a target of perhaps −1 W/m2—in order to gain experience with the consequences of a more modest intervention and its impacts on both to the short-wavelength energy balance and to other aspects of the system before making a decision as to whether the risks involved in scaling to larger values are tolerable; this is the “gradualist” scenario described in Chapter 2 (section on scenarios). In order to provide useful information as to how closely a −1 W/m2 target is achieved, the accuracy of the albedo measurement needs to be significantly better than that, at least 0.25 W/m2. Bender et al. (2006) concluded that albedo monitoring capabilities would have to be roughly an order of magnitude more accurate than they are today in order to assess their importance in the context of anthropogenic climate change. Since that finding is made in the context of an approximate 2.4 W/m2 of radiative forcing by anthropogenic greenhouse gases, it is clear that the current monitoring capabilities fall far short of what would needed in the −1 W/m2 gradualist scenario, let alone smaller-scale field trials, and would be of questionable adequacy even for a full-scale deployment.
Currently, monitoring of Earth’s top-of-atmosphere radiation budget relies primarily on the CERES instrument, which has flown on a series of satellites and is still operational on NASA’s Aqua and Terra satellites and the Suomi NPP satellite at the time of writing. The excess of the top-of-atmosphere mean energy imbalance relative to what can be justified on the basis of ocean heat uptake measurements provided an indication of the intrinsic error in the observation. For the CERES observations, this error (estimated from the excess imbalance) is approximately 5.7 W/m2 (Loeb et al., 2009). The ocean heat uptake has been estimated to be 0.5 ± 0.43 W/m2 at the 90 percent confidence level (Loeb et al., 2012). The order-of-magnitude difference between ocean heat uptake and the satellite-measured imbalance is attributable to some mix of errors in the infrared measurement and the albedo measurement, which results from uncertainties in calibration, measurement of the incident solar flux, instrument spectral response, and angular distribution models. The total uncertainty raises serious questions about the ability of CERES-type instruments to characterize a significant deployment of albedo modification. More work needs to be done on the validity of the data-processing assumptions when it comes to long-term albedo monitoring, and
it would certainly be desirable to develop instrument suites that did not require such extensive corrections.
Measurement error is not the only, or even the dominant, challenge confronting albedo monitoring. Natural variability of albedo is considerable, and it imposes a barrier on the minimum magnitude of induced albedo change that can be detected with a limited-term observation. Considering natural variability limits alone, Seidel et al. (2014) concluded that detection of an abrupt 0.7 W/m2 change in reflected sunlight would be unlikely within a year, even give 5 years of baseline data. They further concluded that detection (let alone characterization) of a 3-month experiment limited to the equatorial zone would require an albedo change three times larger than that produced by the Pinatubo eruption. These conclusions underscore the likelihood that any field experiment aimed at producing a measurable albedo change would need to be large enough to count as full deployment.
Measurement of albedo alone will not generally be sufficient to discriminate between albedo changes due to a climate intervention and those arising from other components of the climate system, such as volcanic aerosols, sea ice, or cloud changes. Isolating the direct effect of a climate intervention would be greatly facilitated by development of a hyperspectral short-wavelength monitoring capability. Beyond quantifying the bulk reflectivity of the surface and/or atmosphere, such observations would characterize reflectivity as a function of wavelength. Such information would provide fundamental insights into the nature of the atmospheric reflectors (i.e., cloud type, water content, optical characteristics, and aerosol radiative forcing) as well as the reflective characteristics of the underlying surface. These spectral signatures, when combined with top-of-atmosphere (TOA) solar irradiance measurements, would provide a detailed understanding of the strengths and limitations of the albedo modification techniques. Hyperspectral imagers can provide additional information on the nature of clouds (e.g., thin cirrus) due to the unique spectral signature of various cloud types and, in the case of snow- or ice-covered surfaces, will allow discrimination of clouds from the spectrally similar (but not identical) underlying snow and ice cover. Information of this type would be valuable in assessing the changes in cloud albedo achieved by boundary layer cloud-brightening schemes, as well as for characterizing unintended effects of stratospheric aerosol injection on upper tropospheric clouds.
Additional insights would be gained from multiangular observations for bulk assessment of cloud vertical structure, and lidar measurements (similar to the CALIPSO mission) for sampling of precise vertical structure of clouds and aerosols.
The capabilities established for monitoring these direct effects would have the added benefit of facilitating detection of deployment by unilateral actors, by detecting
albedo anomalies against a climatological background. To detect such anomalies, however, such capabilities would need to be sustained.
Finally, since the ultimate objective of albedo modification interventions is to lower temperatures at or near Earth’s surface, sustained monitoring of surface temperature would be required. There is a multidecadal history of global and regional surface temperature monitoring from satellites, which complements a distributed ground network. Current global-coverage sensors on polar-orbiting spacecraft include MODIS on Terra and Aqua and the Visible Infrared Imaging Radiometer Suite (VIIRS) on Suomi NPP, all of which build on and improve upon the heritage of the AVHRR system first launched in 1978. Continuity of VIIRS is planned through 2025 on the Joint Polar Satellite System (JPSS) series, and a sustained surface temperature measurement capability into the foreseeable future is essential for understanding the temperature evolution of the Earth system. The importance of such a system would be significantly increased if an albedo modification strategy were to be implemented, as it would be essential for assessing the temperature response at Earth’s surface.
Monitoring albedo determines the proximate cause of the climate change induced by an engineered modification of albedo, but understanding how the climate system responds to this forcing requires additional observations. Albedo feedbacks arising from changes in clouds and sea ice are addressed by the measurements described in the previous section, but beyond that it is necessary to monitor the outgoing infrared radiation, which determines the rate at which Earth loses energy to space. The outgoing infrared flux is affected by the response of clouds, water vapor, and temperature of both the surface and the atmosphere, and accurate monitoring is a crucial part of determining the way a climate intervention has altered Earth’s energy budget. Outgoing infrared observations are provided by CERES and similar space-borne instrument packages aimed at monitoring Earth’s radiation budget.
Because the ocean has enormous heat capacity and is out of equilibrium with the warming atmosphere, closing Earth’s energy budget requires monitoring of ocean heat uptake as well (Hansen et al., 2005). This monitoring is supported by a diverse range of observations of subsurface ocean temperature, but in recent years the Argo float network17 has produced a major improvement in our ability to monitor ocean heat uptake.
Comprehensive monitoring of indirect effects is complicated, because it involves a wide range of climatological processes whose importance may or may not be anticipated. Such processes span a wide range of atmospheric, hydrological, ecological, and other responses. Consequently it is necessary to have a system that observes such parameters as ecosystem health (stress) and dynamics, soil moisture, precipitation, oceanic thermodynamic and dynamic response to a modified energy balance, and other variables. The robustness of the system depends on the risk posture the international community is willing to take. The capabilities necessary to develop an effective system exist today and have largely been deployed, such as the Landsat series of observations, the upcoming soil moisture active and passive (SMAP) mission, microwave radiometers, ocean salinity measurements (e.g., Aquarius), and wind sensors (scatterometers). While this sounds like a call for continued deployment of all the capabilities that have been developed thus to date, it really is a recognition of the fact that avoiding surprises requires vigilance, and the monitoring that ought to accompany the deployment of this global-scale experiment is a commitment to a sustained system that observes all of these critical aspects of the Earth system.
Attributing a credible cause-and-effect relationship requires that scientists have a sufficiently long observation period to distinguish signal from noise and build credible relational statistics, and it also requires that we can develop physics-based linkages between the causes and what we believe are the effects. For this reason, the above observations would need to be sustained for more than a decade, but, more appropriately, through the life of the deployment, since the observed responses will likely be a result of multiple factors and not be stationary in nature. Moreover, because observations only provide information during or after the realization of an outcome, and the real world provides only one realization of a range of possible outcomes, it is critical that process models that capture the physical relationships between the deployment and the response be developed. These models are necessary to provide the insights into the physics that drives direct and indirect responses the forcings. Such insights are necessary in order to characterize and understand the behavior of the climate system response to the albedo modifications, to quantify risks, and to make credible projections. The more quickly and reliably such models can be developed, the sooner the observing system can be scaled back from a comprehensive monitoring system to a more strategic monitoring system targeted at verifying and improving our models. The fact would remain, however, that the better the observing system, the better equipped we will be to understand the implications of our actions.
Detailed understanding of the physics that produces the direct and indirect changes requires detailed process studies that in turn inform diagnostic and predictive models. As a result there is a need, over both land and sea, for a combined in situ and airborne suite of detailed observations on local and—to the extent possible—regional scales. These would complement the large-scale satellite observations described above. One goal of making such observations would be to quantify the forcing agents (e.g., the amount of sulfur dioxide and aerosols in the stratosphere or troposphere) and their evolution and transport over time. Another goal would be to characterize and quantify the response (e.g., the optical characteristics of the resulting clouds, an assessment of the direct and indirect radiative cooling associated with these processes). The specifics of the process observing system would be derived from the modeling objectives. The end goal is to improve model representation of the physics associated with the deployment, such that the secondary effects can be sufficiently characterized and predicted, in order to minimize any adverse effects.
Observing capabilities for detecting unilateral and uncoordinated deployment of albedo modification activities would be relatively straightforward, since the act would be directly measurable. For more insight into the methods used to create the albedo modification and the associated implications, at a minimum, the observational capability identified in the first part of this section above (“Satellite Monitoring of Large-Scale Direct Effects of Albedo Modification”) would be appropriate. For a more comprehensive insight into the effects, the observational needs would be similar to those identified in the second part of this section above (“Satellite Monitoring of Large-Scale Indirect Effects of Albedo Modification”).
Other methods for detecting unilateral deployment, particularly prior to the actual deployment, involve the gathering of intelligence on the movement or use of albedo modification agents (e.g., chemical feedstock transport, manufacturing, and injection facilities).
Monitoring of Earth’s TOA energy budget is at present provided primarily by the CERES suite of instruments, flying on NASA’s Terra and Aqua satellites and the Suomi NPP
satellite. The Terra and Aqua missions are well past their design lives, while Suomi NPP is 3 years into its 5-year design life. With the next CERES instrument planned for launch on the NOAA JPSS platform in 2017, there may be some risk to measurement continuity, which is a very high priority. A number of other Earth radiation budget monitoring projects are anticipated, but maintaining continuity with the CERES record of the past decade is necessary to provide reliable long-term baseline data (Riley Duren, personal communication).
Monitoring of ocean heat uptake at present relies heavily on the Argo float network. This network is supported by a diverse range of international funding sources, but the funding has not been structured to support an operational, as opposed to research-mode, system. 18 Hence, continuity of these crucial measurements into the future is far from ensured.
Some of the most uncertain aspects of climate science have to do with understanding the radiative forcing associated with aerosols. The failure of the current observing capability to quantify the radiative forcing associated with anthropogenic emissions is consistent with the conclusion that the current observational capability to observe and understand climate forcing associated with albedo modification strategies is lacking (see also Robock, 2014). Also lacking is an ability to monitor some of the indirect effects associated with injections: changes to stratospheric chemistry as well as heating near the tropopause and H2O within the stratosphere, for stratospheric injection and changes to cloud optical depth and cloud effective radius associated with tropospheric injection. For example, the MODIS instrument, when combined with observations from the CALIPSO mission, is able to measure cirrus particle sizes which might change as a result of stratospheric injections, but more study is needed to understand whether these current capabilities (in conjunction with current modeling capabilities) are sufficient to attribute an observed change with a stratospheric aerosol injection.
The Pre-Aerosol, Clouds, and ocean Ecosystem (PACE) mission could provide the capability to monitor tropospheric aerosols as well as aerosol-cloud interactions if it were deployed as originally envisioned—with coincident hyperspectral imaging and multiangle polarimetry with spatial resolution of 250 × 250 m2 for selected bands. In addition, such a configuration should allow the retrieval of aerosol heights. However, budget constraints and mission costs are such that the current plans for PACE (still in the definition phase) do not include the polarimeter, and the hyperspectral capability is expected to be scaled back (particularly given the fact that the mission is cost capped). Moreover, the mission is not is not expected to launch until 2019 or
later (Steve Platnick, private communication). If the mission were launched with this combined capability and could achieve an accuracy of the maximum of either 10% or 0.002 optical depth units, much of the albedo forcing and response agents could be well understood.
The SAGE III instrument to be launched on ISS in 2014 is capable of limb-scanning measurements of aerosol optical depth and so will be able to measure the vertical profile of aerosol optical depth at latitudes up to to 51°. At low latitudes the spacing between profiles may be large, so initial detection of an injection may be missed, but once spread zonally should be detectable. The accuracy and precision of the stratospheric integrated column is wavelength dependent, ranging from a few percent at wavelengths ≥676 nm, to ~10% at 525 and 449 nm and perhaps 20% at 386 nm. Three versions of SAGE III were built at the same time, and the SAGE III ISS is the last of the three to be launched.
The observing systems needed to support albedo modification research and controlled deployment are essentially the same as those needed to address fundamental questions concerning the climate system, including estimates of climate sensitivity, characterization of cloud and water vapor feedbacks, aerosol radiative forcing, and response of sea ice and snow cover, all of which occur against the backdrop of natural climate variability. Investment in maintaining continuity of current capabilities, and ultimately improving on their accuracy, is a prime opportunity of a multiple-benefit program that would not only contribute to a better understanding of the consequences of deploying albedo modification interventions, it will also provide fundamental new knowledge about the climate system, which will be essential for meeting the challenges of climate change. It is a no-regrets policy that will be valuable even if albedo modification is never deployed.
This chapter has focused on two anthropogenic actions that are considered to be potentially feasible that could cause Earth to start cooling within a year or two of the initiation of an intervention: (1) introduction of stratospheric aerosols and (2) increasing the reflectivity of low clouds (marine cloud brightening).
It may be technically possible to produce significant changes to the radiative balance of Earth (order 1 W/m2 or larger) via either of these technologies without the need
for major technological innovations. However, albedo modification strategies may introduce major and rapid perturbations to the planet with secondary and tertiary effects on environmental, social, political, and economic systems that are very difficult to predict currently and with effects that could be severely negative. Without further information on these risks, the low initiation costs of albedo modification cannot be balanced against other potential costs and risks of not deploying albedo modification methods.
Looking across the technologies described in this chapter, the committee has identified the following research needs in order to better observe some basic properties associated with Earth’s albedo. Most of these research needs relate to observational capabilities for monitoring Earth’s energy budget that are of a multiple-use nature, address pressing needs in a broad range of climate science besides analysis of albedo modification effects, and do not require any large-scale albedo modification experimentation to yield useful results. Wherever possible, the focus should be placed on “multiple-benefit” research, that is, research that contributes to albedo modification capabilities while contributing to the understanding of climate change and other basic research topics assuming albedo modification is never deployed. Research and observational programs in this category include improved monitoring of Earth’s radiation budget and improved understanding of aerosols and their effect on clouds. An extensive set of recommendations describing modeling and field studies that can be used to improve understanding of relevant processes, and potential consequences from albedo modification, can be found in the earlier sections titled “Summary and Statement of Research Needs for SAAM” and “Summary and Statement of Research Needs for Marine Cloud Brightening” and in Box 5.1 of Chapter 5.
- Because CERES is the prime tool for understanding the top-of-atmosphere radiation budget, a high priority should be assigned to maintaining the continuity of measurement with the CERES instrument package, or with an improved package that can be accurately intercalibrated with CERES during a period of overlapping observations. Since, ultimately, the warming experienced by current and future generations is a direct result of this energy imbalance, sustained monitoring is essential for understanding the evolution of the climate system whether in response to greenhouse forcing or climate intervention. More research is also needed to determine the long-term accuracy of recalibrated and bias-corrected measurements.
- Research is needed on development of a new generation of short-wavelength (albedo) and long-wavelength (outgoing infrared) space-borne instruments that do not require the large bias corrections of current instruments. Development of instruments that could in addition provide spectrally resolved
measurements (“hyperspectral imagers”) would provide an improved basis for discriminating the processes leading to changes in the radiation budget. For support of albedo modification research, hyperspectral short-wavelength measurements are particularly important, but hyperspectral long-wavelength measurements can help discriminate cloud changes and may also be useful in monitoring stratospheric heating due to aerosols.
- Maintaining continuity of the existing Argo float system for continued and sustained monitoring of ocean heat uptake is a crucial part of monitoring the energy budget, as it is the prime source of information about heat exchange between the atmosphere and ocean. Opportunities to expand the system and improve its accuracy should be sought, as well as other opportunities to improve monitoring of ocean heat uptake and storage. Because this heat uptake and storage play a key role in modulating the magnitude and timing of surface temperature change, accurately monitoring these energy exchanges is essential for understanding the response of the climate system to current greenhouse forcing. This need becomes even greater under conditions of climate intervention.
- The observations associated with an intervention, such as hyperspectral measurements, polarimetry, and so on, would provide new data sources, and realizing their full value will require new assimilation and analysis approaches.
- To make use of these types of observations, research is needed on data assimilation and data analysis to improve methods for making optimal use of observations in detecting and attributing the albedo and climate response to deliberate albedo modification.
- Abrupt termination of albedo modification in a high-CO2 world would lead to rapid warming and a host of other rapid changes in climate. There is a need for more research on the impacts of abrupt termination of albedo modification on natural ecosystems and human society. Specifically, it is important to understand what the rates and magnitudes of post-termination warming would be both globally and regionally, what the associated impacts to the hydrological cycle would be, and what the ecosystem responses would likely be. Moreover, research into the relative impacts of a nonintervention scenario and an abruptly terminated intervention scenario, and even a slowly terminated intervention scenario, is needed.
- Finally, if climate-altering deployment of any type of albedo modification strategy were to occur, it would require technology experiments (e.g., tests of delivery systems for aerosols). Because these would be explicitly for the purpose of deployment and experimentation, they might not rise to the level of multiple-benefit research (even though they may produce some improved
understanding of aerosol microphysics). Nonetheless, research in this area would be required in order to responsibly carry out any kind of test or deployment. Development of engineering capabilities required for deployment rather than research should only be developed in the context of a reviewed plan for engineering scale-up of a proposed technique, so that potential “show-stoppers” are evaluated before more tractable but less important ones.
Table 3.4 provides a quick summary overview of the committee’s judgments on aspects such as effectiveness, technical readiness, ramp-up time, duration of effects, cost, ability to detect and monitor, and various risks of the albedo modification strategies presented in this chapter. In each category, the committee has provided an estimate of not only the magnitude of the effect (e.g., high, medium, low, and what those categories mean for that table entry) but also the committee’s confidence in that categorization. The entries on the tables are the product of committee deliberation based on an understanding of the available literature. The table goes into detail for the two strategies that were discussed in detail: stratospheric sulfate aerosol injection and marine cloud brightening.
TABLE 3.4 Table Summarizing the Committee’s Judgments on Various Aspects of the Two Major Albedo Modification Techniques Presented in this Chapter
High Medium Low
|Stratospheric Aerosol Albedo Modification||Marine Cloud Brightening|
|Ability to mask some consequences of greenhouse gas warming, i.e., ability to produce substantial cooling of global mean temperature|
|High: technique could achieve substantial cooling by itself, i.e., a radiative forcing equivalent to a doubling of CO2|
|Medium: technique could be a substantial contributor|
|Low: technique could be helpful but cooling effect is in noise|
|Technological readiness (systems level maturity), technical risk|
|Mature technology (ready to deploy quickly, low technical risk): technology exists at scale|
|Intermediate-maturity technology: prototypes exists, not to scale|
|Immature technology (not ready to deploy quickly, high technical risk): needs prototyping|
|Technological readiness (device level maturity), technical risk|
|Mature technology (ready to deploy quickly, low technical risk): technology exists at scale|
|Intermediate-maturity technology: prototypes exists, not to scale|
|Immature technology (not ready to deploy quickly, high technical risk): needs prototyping|
|Time required to scale to maximum (“irresponsible/uninformed”) deployment with major efforta,b|
|Fast: years (i.e., <10 years)|
|Medium: decades (i.e., 10 < x < 100 years)|
|Slow: centuries (i.e., >100 years)|
|Stratospheric Aerosol Albedo Modification||Marine Cloud Brightening|
|If decision made to deploy, time required to develop informed, well-planned, and controlled maximum deployment with major efforta,b|
|Fast: years (i.e., <10 years)|
|Medium: decades (i.e., 10 < x < 100 years)|
|Slow: centuries (i.e., >100 years)|
|Time for direct radiative effects to dissipate if albedo modification activity is suddenly stoppedc|
|Slow: 1-5 years|
|Medium: 1-5 months|
|Fast: 1-5 days|
|Relative costs of an albedo modification deviced (orders of magnitude; when building at scale)|
|Low cost: order $1 billion per year per 1 W/m2 (i.e., >0.3 W/m2 per billion$/yr)|
|Medium cost: order $10 billion per year per 1 W/m2 (i.e., 0.03 < x < 0.3 W/m2 per billion$/yr)|
|High: order $100 billion per year per 1 W/m2 (i.e., <0.03 W/m2 per billion$/yr)|
|Relative costs of an albedo modification systeme (orders of magnitude; when building at scale)|
|Low cost: order $1 billion per year per 1 W/m2 (i.e., >0.3 W/m2 per billion$/yr)|
|Medium cost: order $10 billion per year per 1 W/m2 (i.e., 0.03 < x < 0.3 W/m2 per billion$/yr)|
|High cost: order $100 billion per year per 1 W/m2 (i.e., <0.03 W/m2 per billion$/yr)|
|Stratospheric Aerosol Albedo Modification||Marine Cloud Brightening|
|Ability to detect unsanctioned albedo modification at scalef|
|Easily verifiable: existing and planned observation systems can verify without retasking|
|Moderately easy to verify: existing observation systems would need retasking or known technology would need to be deployed|
|Difficult to verify: new technology/methods would need to be developed/deployed|
|Ability to measure the radiative forcing of a large-scale, decade-long albedo modification deployment with sufficient accuracy|
|Easily verifiable: existing and planned observation systems can verify without retasking|
|Moderately easy to verify: existing observation systems would need retasking or known technology would need to be deployed; using substantial additional resources employing existing technology|
|Difficult to verify: new technology/methods would need to be developed/deployed|
|Ability to monitor and attribute the climate response of a large-scale, decade-long albedo modification deployment with sufficient accuracy|
|Easily verifiable: existing and planned observation systems can verify without retasking|
|Moderately easy to verify: existing observation systems would need retasking or known technology would need to be deployed; using substantial additional resources employing existing technology|
|Difficult to verify: new technology/methods would need to be developed/deployed|
|Stratospheric Aerosol Albedo Modification||Marine Cloud Brightening|
|Environmental consequences and risks (geographic extent of impact, adverse consequences, co-benefits)g,h|
|Addresses nonwarming effects of CO2 (e.g., ocean acidification, CO2 fertilization)|
|Sociopolitical consequences and risks (include national security)h|
|None / only national issues|
|Binational issues (e.g., one border involved such as United States–Canada)|
|Governance challenges for deployment at scaleh|
|No novel governance challenges|
|Governance challenges likely to be primarily territorial, but with some legitimate interest by other states|
|Potential for substantial adverse effects across international borde or to an international commons|
|How many potential unilateral and uncoordinated actors could have both the technology and resources to deploy at scale|
|Few actors, order 1|
|Medium order, 10|
|High order, 100|
NOTE: In each category, the committee has provided an estimate of not only the magnitude of the effect (e.g., high, medium, low, and what those categories mean for that table entry), but also the committee’s confidence in that categorization.The entries on the tables are the product of committee deliberation based on an understanding of the available literature.
aA “major effort” denotes something on the scale of the Manhattan Project.
b Refers to time from when a decision would be made, but assumes the use of current technologies.
c Does not include secondary effects in climate system, such as changes in precipitation patterns.
d Device refers to a method for deploying some particular albedo modification technique.
e System refers to a device or set of devices capable of altering the radiative energy balance in a measurable way and the associated observing and modeling capabilities for assessing their radiative impact.
f This is likely not a climate signal, but would rather be a logistical signal (i.e., deployment of large numbers of planes to the stratosphere or large numbers of ships) and the resulting stratospheric aerosol cloud (with lidar) and lines in the clouds.
g See sections “Environmental Consequences of SAAM” and “Environmental Consequences of MCB” above.
h Instances where the committee felt the table entries were between values are represented by a symbol that spans both values.
This page intentionally left blank.