Designer Climates?

By Ben Kravitz

"If geoengineering worked, whose hand would be on the thermostat? How could the world agree on an optimal climate?" – Alan Robock (Science, 2008)

When I first started working on this topic in 2008 as Alan Robock's Ph.D. student, I had no idea how to answer these questions, but I knew they were important. I had seen his 2008 study, which showed how solar geoengineering could negatively impact the Indian and East Asian monsoon, which supplies water to literally billions of people. I too was concerned about winners and losers in solar geoengineering and the potential for conflict over optimizing the climate.

Ten years later, I'm still concerned about those issues. However, with the benefits of hindsight and a lot of progress in the research community, I would argue that these questions are the beginning of a dialogue: Is there only one thermostat? And does the world have to agree on an optimal climate, or can the climate be optimized regionally?

Ultimately, asking what solar geoengineering will do depends on what solar geoengineering is designed to do: what the objectives are, how it might be done, who does it, and so on. As examples, solar geoengineering is a very different concept if one is talking about a coalition of nations attempting to use stratospheric sulfate particles to cool the planet versus a local group using marine cloud brightening to save a coral reef. The effectiveness of the two technologies are different, as are the potential side effects, transboundary effects, and geopolitical implications. There is no one solar geoengineering.

This idea of designing solar geoengineering to meet certain objectives has been around for a while. When offsetting global mean temperature change from carbon dioxide with total solar irradiance reduction, one "overcools" the tropics and "undercools" the poles, meaning there are negative and positive residual temperature changes, respectively, as compared to the baseline. Ban-Weiss and Caldeira (2010) were the first to challenge the inevitability of this outcome by imposing different patterns of solar irradiance reduction (fixed aerosol optical depth in their simulations). Irvine et al. (2010) looked at different amounts of solar geoengineering to meet different objectives. MacMartin et al. (2013) imposed different patterns of solar reduction as well as different seasons of that reduction, indicating the potential to meet seasonal objectives like Indian monsoon precipitation or minimum Arctic sea ice extent. All of these indicate that it is possible to meet multiple simultaneous objectives in the climate system with solar geoengineering. There might be many thermostats, not just one, and they might control things other than temperature.

Of course, there remains the question that, given all of the uncertainty in the climate system, how do you make sure you choose the right amount of solar reduction to meet the objectives? In an El Niño year, the planet is warmer, so cooling it might require a bit more solar geoengineering than on average. There are other sources of variability, both natural and human-influenced, that add uncertainty to estimates of how much solar geoengineering might be needed at any given time. Rather than attempt to create a perfect model (which is impossible), one can use a strategy to meet objectives in the presence of uncertainty.

We call this feedback, and we use it all the time in many aspects of our lives. The basic concept is that one uses information about the past observed system state to regularly adjust whatever changes are being imposed. In the case of solar geoengineering, one looks at how far one is from the objective; using global mean temperature as an example, if it's too hot, do more solar geoengineering, and if it's too cold, do less. Jarvis and Leedal (2012) and MacMartin et al. (2014) introduced this concept to the climate literature, borrowing heavily from control theory. Kravitz et al. (2014) demonstrated that feedback can work even if the system is not perfectly known; e.g., what if our estimates of climate sensitivity were off by 50%? This is a critical point because, if solar geoengineering would ever be implemented, the consequences of errors can be drastic, and society doesn't get do-overs – we would have to get it right the first time.

My colleagues and I were the first to combine using feedback and meeting multiple objectives by modifying multiple degrees of freedom in the climate system. We also demonstrated that a feedback algorithm used in one model could be ported to a different, independently developed model and still meet the objectives. However, we used patterns of solar reduction which, while informative, are ultimately of limited utility. First, even though one can impose patterns of solar reduction in a climate model, there is no guarantee that one can achieve the same patterns with stratospheric sulfate particles – one cannot wall off the stratosphere. Also, stratospheric particles have different effects from simply turning down the sun. Particles grow, coagulate, and fall out of the atmosphere, which means that the more you put in the stratosphere, the less bang for your buck you get. They also heat the stratosphere and provide surfaces for reactions, changing wind patterns and atmospheric chemistry, both of which can affect surface climate.

To address these challenges, my colleagues and I used a state-of-the-art climate model to inject SO2 into the stratosphere at multiple locations, applying a feedback algorithm to adjust the injection amount at each location, to meet multiple simultaneous surface climate objectives. This demonstration provided, arguably, the first realistic simulation of how solar geoengineering might be designed, given a set of predetermined objectives. (Note that the actual determination of those objectives can be informed by natural science but is ultimately far beyond the purview of natural science.) To aid in producing a holistic understanding of the implications of this solar geoengineering strategy, this simulation has been publicly released to the climate community.

This raises the question of what things about solar geoengineering can be modified. For stratospheric sulfate particles, one can choose where (horizontal and vertical), when (what time of year), and how much to inject. One can also potentially choose the form of the injection: gaseous precursors like SO2 or directly forming liquid particles. Also, although the discussion in this post thus far has largely focused on stratospheric sulfate particles, there are other methods of solar geoengineering that will undoubtedly cause different surface climate effects. For example, using particles other than sulfate will likely have different effects on surface climate and the stratosphere. Boucher et al. (2017) found that the surface effects of marine cloud brightening generally happen in different places than the effects from stratospheric sulfate particles, indicating the potential to simultaneously use both technologies to expand the space of achievable climates; Cao et al. (2017) found something similar for cirrus thinning. However, adding multiple technologies also adds additional potential sources of uncertainty that need to be resolved.

So ten years following Robock's article, where are we now, and where are we going? For starters, I think we now know enough to ask different questions. It may be possible to meet multiple, simultaneous objectives in the climate system. The community has demonstrated that some things can probably be controlled. There will almost certainly be limits to what can be controlled – solar geoengineering is not a panacea. It seems unlikely that there can be a single solar geoengineering strategy that will meet everyone's objectives, meaning that there will almost certainly be winners and losers. Although models have suggested some results, we cannot yet say for certain whether solar geoengineering will hamper the Indian monsoon, alleviate drought in sub-Saharan Africa, or any particular conclusion about a region. Also, most of the investigations to date have looked at variables associated with physical climatology, most frequently monthly or annual mean global temperature and precipitation, and sometimes regional temperature and precipitation. While informative, these are only proxies for real impacts, such as extreme events (like floods, droughts, and heat waves) or food, water, and energy security. The next phase of research will almost certainly need a prioritization of the most important questions and how those questions can best be answered. If there are any showstoppers, we need to know as soon as possible so resources can be spent elsewhere.

Which leads to my final point about the role of natural science in this discussion. I am a physical scientist, and as such, I have focused this blog post almost entirely on physical science. There are other fields that are crucially important -- from political science, economics, ethics, governance, public health, and engineering to other natural sciences like ecology and those studying agricultural or other specific impacts. Ultimately, natural science will (hopefully) be one of many sources of information about solar geoengineering that can be used by decision makers when assessing the risks of deploying solar geoengineering versus the risks of not doing so. We need to provide that information with continued research aimed at understanding what solar geoengineering can do, and what it cannot.