 I'd like to be here. Thanks. Well, a few disclaimers at the beginning. One is this is in the Stanford Energy seminar series, and I'm really not talking about energy. So if you can leave now, if you insist on energy. And actually, I mean, related to that on the intro, on direct air capture, my only involvement is through this company, Karmash Airing. I'm very proud of it. It's doing pretty well. But I don't do to sort of minimize conflict of interest. I don't really do any academic work on that at all. So I'm really trying to keep the two things completely divided. So I'm going to talk really exclusively about solar geo-sharing and kind of a sandwich talk where I'll start a little bit about some of the broader policy context, then dive into some real technical details, and then come back and talk more about policy context. I guess let me start with really some highest level policy takeaways. One is nothing that I'm going to tell you gets us out of the need to cut emissions. So in the end, if we do not bring emissions effectively to zero, you won't have a stable climate in the long run. And nothing that we know about solar geo-engineering changes that. Solar geo-engineering may be a way to substantially reduce the risks of having a given amount of CO2 in the air. And so it's possible that a combination of emissions cuts and solar geo-engineering might get us to a lower risk climate state than emissions cuts alone. But I don't think we know that with confidence now. That's the most important high level comment, I think. So with that, let me start. So let me start with a high level policy picture of how I think this technology might be used. And then I'll get into details. So climate risk just keep increasing if you keep emitting. The most important single thing to know about climate. And of course, connected with that is the fact that even if you bring emissions to zero, the climate problem doesn't go away. It basically just stops getting worse. I think presumably most of you know that, but it's easy to forget. In some ways, the emphasis coming out of the 1.5 report on the idea there's some kind of sharp threshold at 1.5 or 2 or something, which I don't think there's actually much evidence there is. The idea there's a sharp threshold makes people talk as if somehow if they just stay under the threshold, everything's okay. The answer is the more CO2 there is in the atmosphere, the higher the risk. Almost certainly it grows nonlinearly, but there's no magic safe level where if you just go above it, there's a magic dangerous level. At least I'm not aware of strong evidence to say that that's true. Now in the long run, we could reduce, we could pull CO2 out of the atmosphere by a whole range of methods of which a direct care capture is just one. But I think they're all inherently slow. That takes an immense amount of industrial activity, one of the industrial hubs of the world is burning fossil fuels to provide the world's energy system. If you want to remove that amount of CO2, you've got to deal with gigatons of material. I think there's no way in which there's a magic quick fix that does that with tiny environmental impact and tiny cost. It can be done, but it'll be inherently slow. So the way I think about it is carbon removal is something that you could allow you to gradually pull CO2 concentrations down, but I think of that as something that primarily happens after we brought emissions effectively to zero and then as a long, slow thing on the other side. That's how I think about it being used. So after I'm dead a long time away. When I think about solar geoengineering, I think about ways that we manage that peak, that we manage the period of peak concentrations. And I don't think, as I said, we know for sure that it makes sense to do so, but I think there's evidence, and I'll show you some of that evidence, that solar geoengineering might reduce the risks, the climate risks during the peak concentration. And to me, that's really the question. So again, maybe slightly diverge from John Wyant's framing. I don't think the reason to take solar geoengineering seriously is because we're being less successful at cutting emissions than a lot of us hope we would be, although indeed I actually see lots of reasons to be excited about emissions costs, the extraordinary improvement in the cost of solar being one of them. I think if it makes sense to take solar geoengineering seriously, it's because it actually provides a lot of environmental benefits compared to its environmental harms, which are certainly there. And I don't think that's definitely true, but that's the test. And I think it's the only really interesting test. And to some extent, that test is independent of what is the particular emissions trajectory or concentration. So it might be true that solar geoengineering really reduces risks a lot in comparison to its risks or not, but that would really be more or less as true if we were on a high emission trajectory as a low one. So I won't go into detail, but there are a lot of different ways you could, in principle, do this. I've listed papers on the bottom, so I'm not expecting you to read all this. I'm gonna focus mostly on stratospheric aerosols and there's some, I think, good reasons to focus on stratospheric aerosols, but at this point, there are certainly other things that could be done right up to in principle, space-based technologies that I think might be done in the long, long run, and there are a bunch of other things, including things that interact with high-level, serious clouds, happy to take questions about them, but all focus mostly on stratospheric aerosols. So if you think about the thing we know most about, which is sulfates in the stratosphere, one useful way to divide up this big universe of things, and I should say that another thing I should say in prelude is there's still a comparatively small group of researchers working on this. There's been a strong taboo against research on this topic, and to some extent, one of the real risks is that it may be that some people, including me, have kind of sold ourselves on something that actually isn't as good as it looks, and what we need to do is get this research out into the broader world and get a much bigger set of people looking at it and critiquing it, and only then will we have a better sense of what's real or not. On the other hand, for some of those who might not have been following this, it's important to say there's quite a lot of research now, so there's, depending on how you measure it, I'm like 500 papers on this topic, really much more than there was, kind of exponential increase in recent years of papers and attention. So when you think about the risks in advocacy, one way to divide it up is to think about the risks of any particular way we do solar geoengineering, and so I can think about this as the way we manage radio-forcing, so any particular way we do it has its own particular set of risks, so putting sulfates in a stratosphere can cause, accelerate stratosphere, it owes own loss, it actually may be more importantly, warms the lower stratosphere, which will let more water vapor in, it has various other knock-on effects, it scatters sunlight and can make acid rain, in particular it's a whole bunch of risks that we can all assess with existing tools of atmospheric science and chemistry, and then there's the question of what is the climate response to solar geoengineering? So what we know for sure is that solar geoengineering is not magic anti-CO2, so there's no way in which if you increase the amount of CO2 in the atmosphere, and then use solar geoengineering to perfectly return the earth to say pre-industrial global average temperature, which could certainly be done, because two curves can cross, there's no way that doing that restores the climate exactly the pre-industrial climate, it doesn't, it's not anti-CO2, it's a different climate forcing, and the question is to what extent does it really reduce climate risks and what risk does it increase? Those are the key questions, and this lists some of them. This I should say are the key physical questions. Again, to keep giving you some context, there are lots of physical science questions here, this talk is focusing on the physical science, but in many ways, all the hardest questions about this technology are really about how we go about making decisions about managing a research program and managing decisions about a deployment in a divided world. So let me first talk about advocacy, so about the extent to which these technologies actually could reduce climate risks. So what I'll show you is results from a paper that I and a bunch of co-authors will publish in Nature Climate Change in a few weeks, and it's a particular distillation of that stuff, but in fact, there's a whole bunch of earlier papers that say more or less the same thing. So I don't think we're radically new here, there's actually a big body of papers that support the kind of work I'm gonna show you. There's some particular ways at which I think we've done a slightly improved job in making the comparisons, but the underlying results I'm gonna show you I think should be pretty surprising. We haven't seen them before. So first of all, before we get to our new work, I wanna say I've been involved on and off in climate modeling now for three decades, and people tend to keep claiming that client models with higher resolution are better, and for lots of things that is clearly not true. So it's not clear that if you just turn up, it's clear that if you just turn up the resolution of a client model, you don't know the climate sensitivity better. But I mean, there's some ways in which they really are improving, and what I'll show you is results from a collaboration between Kerry, Emmanuel, and MIT, and a bunch from the GFDL group that use this GFDL model that's got a quarter degree resolution. And on some particular things like precipitation variability, which is the slide I'm showing you here, and tropical cyclones, it really does a substantially better job on getting current climate, which is all we know than previous models, and that's kind of exciting. So this is just, you're showing you the amplitude of the one year maximum rain event, and this shows you roughly what observations look like for the US, and what most models, client models are kind of in this resolution, this is what their ultra high resolution model does, and it really does a significantly better job. So because one of the big questions about solar geoengineering is how it interacts with precipitation and precipitation variability, that was one of the reasons to use this high resolution model. It's the first time a really state-of-the-art ultra high resolution model has been tested on these technologies. This gives you some sense of the specifics of the model and the setup, but I won't say much about that, I'm gonna focus on results. One of the key things we did here is we chose to model a case where solar geoengineering was used to offset roughly half of the warming from CO2, and that's an important choice. Often, most of the previous modeling is just for convenience, effectively assume that solar geoengineering was gonna be used to return climate to pre-industrial, to offset all the radio forcing from CO2. I think there are a bunch of reasons why that's probably not a sensible policy choice. My view is that if it's used at all, it probably will be used sensibly to do less than half, because in general, it looks like the benefits grow roughly, literally, but the risks grow non-linearly, and so you really wouldn't want to do it to try and return climate to pre-industrial. And this gives you some look at some of the reasons why. So this is, sorry, this is in terms of, I cut off the X axis, in terms of percent reduction in global solar insulation where this is 0% and this is two, what you see is that the temperature goes down more quickly than, the precip goes down more quickly than temperature, and that's one of the many ways in which you can't get both things right at once. And what you see on this graph is the normalized RMS differences for some of those key variables, and I'll get to what the variables are in a minute, but they are temperature, the hottest single temperature during the year, the extreme five-day precipitation max and precipitation minus evaporation, which is basically water availability. So let me give you a little bit on the results. First of all, these show you just the overall distributions in two ways over land. So this is the simplest thing, this is surface air temperature anomaly, and this is just looking at all land surface where the red is what happens with two times CO2 and the blue is what happens where you've done this half solar geocase. And so first of all, you can look at the medians and you see that it cuts the median half, which is no big surprise. And you've got two cases here, one of which is all ice-free land area, and the top one for each thing is weighted by population. So one of the things we're trying to do in this paper is inch a little closer to policy-relevant results, and so you could argue that population-weighting is a useful way to do that. We've looked at a few different weightings, but we're not looking at all the globe, we're looking at either land area or population-weighted land area. And the thing you notice here is say if you look at population-weighted land area, so this is the distribution for two times CO2 over the land area, so you get some areas, some small fraction, warm is more than four degrees C and some areas warm a small amount. And you see that not only is the mean reduced, but the extremes have been reduced. So and you see that for temperature, which is no big surprise, but also for maximum temperature, also for P-C. So there's some early papers that suggested that solar geometry could somehow disrupt the Indian monsoon. The paper that said that actually said that in the abstract, but didn't show it in the paper, and there's actually no really strong evidence that it does that. In fact, most of the evidence in paper suggests that it tends to push monsoon strength back towards pre-industrial, but it's an important thing to think about. And so we've started to look at that among other things with P-C, which is basically the water availability and then the paper we've also looked at by season and by region where you can start to pick that apart. And you can see that solar geometry does reduce the change from pre-industrial in P-C, this crucial thing, the availability of water, and also in extreme preset, which is one of the really damaging, which is damaging preset. So that, this is good, if you want to reduce these climate impacts, but it doesn't tell you the distribution of what fraction of the world actually sees its climate made relatively closer to pre-industrial or further away from pre-industrial. In this paper, we've chosen to use the words moderated or exacerbated and it's better, better or worse, because better or worse is not clear that everybody actually wants their climate to be closer to pre-industrial. And we'll get to that in the final slide here. But if you think about it, it's perfectly true, perfectly possible that this plot could be just what it is, and you could still see lots of points that actually moved from here to there, if you like, or rather from here to there, and so had their individual situation made worse. And politically, one of the key things that matters is what fraction of the world actually sees their situation made worse. It's clear that in these models most of the world sees their situation made better, but what fraction sees it made worse? That's the key sort of new thing we went after with this paper using some serious statistical test. And this gives you some answers. So this is answers from high floor, this high resolution model, for these four variables. And again, it's a step forward to use these variables which we think are more policy relevant than people have used before. People often just use precip, but precip's actually a lot less interesting than precip minus evaporation, which is water availability. But the answers for precip aren't much different. So what this is saying is the fraction of the global land surface where things are made worse, exacerbated, for high floor is about 0.4%, well for temperature and temperature stream is zero, but for P minus C or PX it's about 0.4%, but 26 to 40% are made better. What about the other percentage? Not statistically able to decide. So for every point in the world you either put into a bin where statistically made had its climate exacerbated or moderated or you can't tell. So that's what works. Then for geomip, geomip is the geo-insuring model inner comparison project, which is a previous effort that used 12 different lower resolution models. And we've gone back and used that whole data set, rescaling it to pretend that it had done this half geo-insuring case even though it really didn't. And then we've separately taken one of those models and checked the linearity of that rescaling assumption. So then you might say, well, you're still a little worried about this 0.4%. It's made worse off and maybe there's a larger part that we didn't resolve. And so we wanted to really look at what are the parts that seem to be made worse off. And I'll show you that in, first of all, kind of a geeky way and then in a final figure that hopefully it's more digestible. And the geeky way is this, which is these are scatter plots. Oops. Sorry. These are scatter plots that shows you the two times CO2 anomaly on the bottom. So this is the change under two times CO2 in the base case. And then the anomaly in the half solar geo-insuring case. So it's a scatter plot of all points. And what you see is that the places that are exacerbated, that are read, are all the places where the underlying two times CO2 anomaly is itself very small. So essentially the only places are almost the only places in the entire data set where we see that there's statistically significant exacerbation are places where the underlying two times CO2 change was very small. And so then it's very easy to get a big ratio. So that explains something more about these ratios. So now I wanna give this to you in a distribution overland service that I think starts to get closer to a really kind of policy relevant outcome. So these are the so-called SREX regions from IPCC assessment of extreme events. So we used exactly the same standard regions as IPCC uses. And for each of those reasons, we've done an analysis to look for each of these key variables, temperature, temperature, extreme P-minus C and extreme precip. We've coded it for the region as a whole about whether it was exacerbated or moderated and whether it's statistically significant. So blue means moderated, light blue means in the direction moderated but not statistically significant and then reverse for red. So the big news is there are no red points. So there's not a single region in which any of those four variables is statistically significantly exacerbated in this particular model run, which is I think kind of a surprising result. Maybe it's not true and that's the big thing I wanna get into but it's a surprising result. No, I'm not saying that for effect. I think if these results are really representative of the way solar geomissioning could actually work, then the evidence that it could produce a really large reduction in climate risks to humans and the natural environment are substantial. But the big question has to be what are we missing? Is there some way in which we're asking the question the models in which there's a model bias that we're missing? Is there some other way in which we're really getting it wrong? Because the penalty for getting it wrong here is very high. This shows you the results, sorry. This shows you the results when I just, for the case of P minus E and PX, because for T it's obvious and easy, we're showing the distribution over the 12 of the geomint runs. So first of all, I showed it to you just for this ultra-high resolution model. Now I'm showing it to you for each of those and what you see is there are some red here. So in this particular case, which in fact is just where I was on a birding holiday with my family a few days ago by chance, right about here, half of the geomint model suggests that in this region on the coast of Chile, basically, the climate change is exacerbated for P minus E. And that's interesting, but what it turns out is exacerbating this case means made more moist, more water availability. So that's where it really matters whether you think exacerbated equals worse or not. My guess is most people in Chile would not think that a little more water was bad. But on the other hand that the environment in the Atacama Desert, if you like that wants to be like the pre-industrial, would think that a little more water was bad. So these value questions about which direction we want to go really matter. But in any case, the big picture result of this and other similar modeling is that solar geoengineering at least imposed in a simple way, this is with a pure adjusting the solar constant model, but I'll get to that in a second, looks like it really reduces climate change in meaningful impact measures almost everywhere. So what are the ways this might not be true? So one big way is of course, this was for a model that just adjusted the solar constant. Could that be true in the actual world where we're putting say some aerosols in the stratosphere and we can't get the distribution of aerosols exactly right and those aerosols don't just adjust the solar constant but they do other things as well, they scatter light, they change ozone, et cetera. And what is the gap between those two things? You can also look at other variables, droughts, monsoons, sea level, et cetera. So some of this has been done in a bunch of other papers. So a group at NCAR and Doug and Martin at Cornell and others have done a variety of feedback experiments on stratospheric aerosols, we've also done them and I think in a series of papers, we've convinced ourselves that it ought to be in fact not that hard to do adjustments of where you inject aerosols to get a pretty even distribution of aerosols in the stratosphere. There are a bunch of other questions about chemical response and about response of the upper tropospheric cirrus that I think are much more open but I think the question of whether it's technically possible to make a pretty uniform aerosol layer, from my point of view, doesn't look that hard. So my sense is other than the slight change in radiative forcing you get from aerosols versus turning down the sun, a special character of radiative forcing, this is a pretty realistic case. So that's a kind of statement to me of why it makes sense to take this seriously. So first of all, any questions on that? And then I'm gonna go into talking a little bit about experiments. Yep. As a curious comment, you disperse this stuff because of the circulation patterns in the atmosphere. Can you do it so it's like preferentially northern hemisphere or preferentially southern hemisphere? Can you differentially apply this aerosol application? Yeah, so for stratospheric aerosols, what you find is given the stratospheric circulation, there are really sort of about three or four realistic degrees of freedom. So we've actually tried doing a kind of a model where we put the aerosols into every different latitude and then looked at the kind of transfer function if you like, looking at the radiative forcing. And what you find is you can certainly make the northern hemisphere have more radiative forcing and the southern hemisphere have more radiative forcing or you can make the equator have a little more than the poles and that's about the level at which you can really adjust. You can't make a narrow band of radiative forcing just over 30 degrees latitude and not over 25 because of the natural mixing of the stratosphere. Yep. Is there any relevance of fast experiments on seeding the cloud to the brain of the haters over the atmosphere? So that's a really good question and the answer is is there any relevance of work on cloud seeding to this and the answer is in some policy ways, yes, but technically no in the sense that stratospheric aerosols are just a really different problem in lots of ways than what it is that makes cloud seeding work. They're just deeply different problems. But there are connections. So one of the things that's left out of these models a lot is if you put aerosols in the stratosphere they're going to percolate down to the lower atmosphere and in the upper troposphere those aerosols might be ice nucleating and there's questions about what their impact would be on stratospheric cirrus and I think we don't know the answer to and that's an answer that cannot be achieved by modeling because we actually don't know the ice nucleating character of those particles and you have to go do experiments, both lab experiments and eventually in situ experiments. Yep. Just going to switch the first question. Yeah. So that's the perfect policy question. So if it was true that you could adjust the climate locally then there's no hard policy problem. Each region can choose the climate they want and they don't have to bother to coordinate, it's easy. That is absolutely impossible because the world is interconnected flows of momentum and energy and moisture, you can't do that. So even if you could locally adjust the radiative forcing which in principle with some space-based system is at least theoretically possible even if you could locally adjust radiative forcing that doesn't allow you to have local climate adjustment that doesn't affect your neighbor. The climates are inherently interconnected. So when you look at that range of different ways that this might be done, oops, look at the range of different ways that it might be done. One of the reasons I tend to like the idea of aerosols and the stratosphere is they inherently kind of want to be global and they're slow moving, they have this couple of year time constant so they're harder to make local adjustments and I think as a policy case that actually may be better. So some of these other ideas especially marine cloud brightening is inherently local and it only is operable if it's operable at all at about 10% of the world where there's a certain kind of marine stratus cloud that's right and it also has a time constant of just hours. So if you actually built a technology to do this and you did it for several watts per square meter globally you've built yourself a lot of weather control automatically which I think is something you might actually not want to give to humanity right away and you've bought yourself a bunch of asymmetries. So that's one of the reasons I think overall at something that's even might be easier to deal with politically and better environmentally but they all need research. Okay, let me go on and talk a little bit about experiments just to give you a flavor of what's possible. So one of the big questions here is what should be done for research? Because at this point to be clear I don't think it makes any sense to claim we should go ahead and just do solar geoengineering. What I'm claiming is a serious research program makes sense and I want to give you some sense of what might be achieved in a research program by telling you a little bit about the research our group is doing. Not claiming that our group is answering all questions it certainly isn't, it's very small but I want to give you a sense of what sort of research needs to get done and that cannot be just research in the laboratory just research in climate models because models just tell you what we know from the atmosphere now and there's lots that we don't know. So I'll gloss over this and kind of jump to an actual experiment in our lab. So we're thinking about the possibility of putting particles in the stratosphere and those particles will interfere with stratosphere in chemistry with the chemistry of the ozone layer and that interference depends on the detailed chemical interactions of those chemicals in the stratosphere and the particles that we might put there and those involve a bunch of coefficients so-called gamma coefficients basically with the likelihood of reaction when a particle hits a surface that we don't know very well. So I'll give you a sense of where we are right now in terms of laboratory experiments to measure this. This is a classic flow tube experiment that's from Molina kind of old stratospheric chemistry with a mass spec in the back so we have a flow tube coated with whatever kind of particle we're thinking of putting the stratosphere in this calcium carbonate and we flow different gases in there that one that's really hard to synthesize is chlorine nitrate because we want to look at chlorine nitrate reactions in some of these surfaces and I'm not going to jump into the details here but just to say this is kind of what our raw data looks like or Colleen or Jen, my students, raw data looks like. This is her actual half or PhD thesis gamma coefficients or a third of her PhD students thesis gamma coefficients which are hard to measure. This is a difficult real world experiment and then this shows you the ozone column change due to doing geoinsuring with calcium carbonate which is one of the ideas that our group and others have not been looking at and it shows you the differences if you used our laboratory estimates for these gamma coefficients compared to the previous literature estimates and the answer is the differences are pretty substantial. They're not giant but there are differences where you really want to know the answer. So that's just one minor example of why if we were really serious about understanding how solar geoinsuring works you actually have to go and do a bunch of practical work. It's not like groundbreaking science but you actually have to go do these measurements because we don't know them. Why are we interested in calcium carbonate? So that goes back to an earlier paper this is a paper in PNAS that a few years ago where we looked at the way different particles in a stratosphere might work both to make radiative forcing that's for solar geoinsuring and might work to damage or repair the ozone layer. So the particles that people thought about the most are SO2 and so with SO2 if you do say a watt per square meter of radiative forcing depending on when you do it in this century you might reduce global column ozone by close to 10%. That turns out to be a really high end number because that's assuming that we do it with current chlorine loadings and of course with the Montreal Protocol chlorine loadings coming down. So in fact if you do the same radiative forcing with sulfates in 2075 you get much, much lower ozone loss because the chlorine's mostly gone because the sulfate is acting to amplify the ozone destroying properties of the chlorine that we put there. But this red is what happens with calcium carbonate. So calcium carbonate turns out to actually tends to push the ozone layer back towards pre-industrial or tends to increase the amount of column ozone. So the idea here is that with calcium carbonate you might be able to both do solar geoinsuring and also tend to heal the ozone layer which sounds like a good idea. In fact the really interesting part is not ozone. The more we thought about this ozone is how we started thinking about it but the more we thought about it I think ozone is less important than this property which is the amount of stratosphere heating. So one of the, this shows you in a, well in a model that's high quality in some respects and not others. Not a dynamic model. It's a model based on us actually going back to again to laboratory work and fundamental measurements to look at the radiative constants for these various compounds and looking at how much these compounds would actually warm the stratosphere if you put in enough to do two watts per square meter I don't think it is a radiative forcing, I remember right. And what you find is sulfates warm the lower stratosphere by a degree and a half and that's a lot because the lower tropical trouble pause is what regulates the amount of water vapor against the stratosphere which in terms of regulates a bunch of things about stratosphere chemistry and that turns out to be a pretty big perturbation. And some of these other compounds like calcium carbonate or diamond in principle have much, much lower heating rates and I think that's actually at least as big an advantage as the fact that you tend to be repairing the ozone layer. So that's the reason that we actually need to know more experimentally. There's a whole other host of unanswered questions about how you could actually disperse these particles, et cetera. So that's what I showed you as an experiment. We have done this an experiment that under Frank Koitch's leadership we're working towards doing which is a flight experiment that will actually in a balloon gondola allow us to produce a mixed plume from the propellers of the aircraft. It's called a stratosphere controlled perturbation experiment where we'll be able to introduce aerosols and make measurements say of the aerosol nucleation rate or of interactions of aerosols of the background gas. So I'll just show you a little bit about this to give you some kind of overview. Let's give you some look at the sort of current engineering state of it. So you some look at what the experiment actually looks like in terms of scale. So the idea is this experiment moves along and makes a wake of well mixed gas that's mixed by the propellers. Remember the stratosphere is extraordinarily stratified. So if you just release some material in stratosphere and you haven't mixed the stratosphere up with a propeller you don't get a well mixed volume and you can't do an experiment on it. So the propellers in this experiment do two things. They move the craft at kind of walking speed literally meter and a half a second dragging the balloon behind it. So they move the craft but they also make an aerodynamic wake and that aerodynamic wake is basically our beaker into which we can introduce materials and do a controlled experiment. So this gives you some of the kind of objectives. One objective is just understanding stratospheric mixing better. It turns out that we actually don't know enough about background stratospheric mixing rates and there are ways in which we think we could use experiments like this to improve our knowledge of mixing rates. There's questions about what the actual interactions are in the stratosphere. So we can in a lab take reagent grade materials and the lab make experiments with sulfur but what do you know? After years of us assuming that a whole lot of the aerosols and the stratosphere were all sulfur there's now evidence that maybe there's more organics there than we thought. There's really interesting evidence from Dave Fahey's group at NOAA that suggests that actually there's a lot less SO2 than there ought to be in the stratosphere and so it's possible that we're seeing more organics there. That's a different chemical system. So the point is what you do in the lab is not necessarily what's in the atmosphere and you actually have to go make controlled measurements to understand what's there. Why are we doing these experiments? It's surprisingly hard to reproduce stratosphere conditions in a lab. A lot of these reactions to stratosphere involve radicals. The radicals are formed each solar day and then they go away at night. So actually getting a 24 hour measurement tells you a lot about stratosphere chemistry but you can't really or it's very hard to make truly wall-less laboratory environments. It's really hard to measure these things in the lab and also as I say we don't know precisely what's in this stratosphere. So that gives you some kind of high level sense of what our group is thinking of doing for experiments but I think the larger question here again is if the world was really serious about knowing about this, the experiments I just showed you are just one of hundreds of experiments you'd want to do to really start to give you coherent answers to a whole bunch of questions you'd want to know that would help inform knowledge about the risks and efficacy of solar engineering. So now let me say a little bit more going kind of back to policy about how this might fit into long run climate policy and I'll give you some versions of the curves that I gave you in that first slide but now coming out of an integrated assessment model. So for those of you who lived in the integrated assessment world, this is from DICE, this is the Nordhaus model. It's the kind of original, simple integrated assessment model and I like it because it's so simple. It's got a pretty limited number of equations. It's just, it models a world which doesn't exist. It models a world with one unitary decision maker who's optimizing over the next centuries which of course is not the world we live in but I think that world is a useful, talking about that world is a useful tool because there are some respects in which optimizing welfare is a useful way to think about the way the world might be and I think the model is appropriately simple and in the host of integrated assessment models people have not, I think before this, done a model in which we are freely optimizing both solar geoengineering and carbon removal as well as mitigation and climate damages in a single format. So we try to do that all in DICE, I think, for the first time. So first of all, I'm just showing you a baseline and each of the next plots I'm gonna show you, you'll have the same four axes. So this is all going after 2250 which I know is absurdly far in the future but if you really wanna think about carbon removal I think it's relevant to go out that long which is why we did. So the first one show you emissions. Below it shows you the temperature change. This shows you the change in rate of forcing watts per square meter and this shows you the GDP percentage damages. So that's generic DICE. So this is the generic DICE and just to be clear in this generic DICE, DICE solves things by having a temperature rise about four degrees C compared to pre-industrial. So DICE is in the current standard version of DICE the optimal answer for how much to cut emissions leave temperatures about four degrees above pre-industrial. When it balances the overall cost of cutting emissions against the risks of climate change given its discount function. I'm not saying necessarily I think that's the right answer but that's the standard answer in DICE. So what we did was add, first of all add carbon removal in the simplest way we could. We added carbon removal by just assuming that the mitigation cost function in DICE keeps on going beyond the mitigation of 100%. And so effectively we're assuming that it's a smooth and monotonic function so the cost of removing the last ton of CO2 of squeezing the last ton of CO2 emissions out of the economy is the same as the cost of removing the first ton from the atmosphere. That's the only change we made in DICE and suddenly you get some curve for how mitigation works and then when we added with solar geoengineering. Now for solar geoengineering what do we know about the cost of solar geoengineering? What do we know about the risks? Our answers that we're handling in it we're making a fundamental kind of arm-waving parameterization that we think is quite conservative and let me tell you what it is. First of all and this will sound like salesmanship we're assuming the cost of doing solar geoengineering is zero. It won't really be zero but I think that's actually a very fair approximation. The cost is so low of doing it that it's not likely to play a big role in the decision. It's all about the risks. So what do we assume about the risks? We assume in this model that the risks of doing solar geoengineering at a level of radiative forcing big enough to offset the damages of two times CO2 were just as big as the damages of two times CO2 which there's actually no paper to back that up. That's I think a very high estimate of solar geoengineering damages. So we're to repeat that. We're assuming that if you did enough solar geoengineering to offset two times CO2 your damages from whatever the risks are of solar geoengineering, aerosols in the stratosphere are as big as the damages of two times CO2 which is not the kind of thing I showed you in the modeling study above but let's just suppose that it's true. You might say why would you bother to do it then? The answer is because the damage functions here are quadratic. So even if you have something that's just as bad as CO2 you'd still like to do some reduction because you've got these quadratic damage functions and you'll reduce the overall cost if you use some of it. So I don't think we know that answer at all. It's a purely parametric answer but it's a useful parametric answer. We can adjust to get you some sense of how these three policy instruments play together. I think the thing you get out of a model like this is not a specific numerical answer. I don't believe the answer about the amount of mitigation here is the right about the amount of solar geoengineering here. Let's say it's the right amount of solar geoengineering. What I think this model does tell you that's useful is something about the sequencing of these things in time. I think that's probably a relatively robust answer. And there's an important answer here. If you ask many climate policy experts right now they would list in terms of preferability, cutting emissions first, carbon removal second, and solar geoengineering as a very distant third in terms of rank ordering things in terms of how we should think about them as part of the climate portfolio. Many people assume it's also the same answer in time that you should first cut emissions, then do carbon removal, then do solar geoengineering. I think that latter assumption is wrong. So as I think, and I think this model backs it up in a strong way, that the right way to think about the time sequence is in fact cut emissions, solar geoengineering, and then carbon removal. It might still be that it doesn't ever make sense to do solar geoengineering. That depends on the balance of risks and benefits. But if it does make sense, it makes sense to do it, as I showed in that first slide, to cut off the peak. And it makes sense to use carbon removal as a way to pull emissions slowly back down. And that's the thing I think you see in this integrated model, and I think there's some good reasons for it. So this shows you that result. I'm happy to give you more answers and questions. And that shows you the result for the 2C target. The 2C target, of course, is an arbitrary target. This is inherently an awful model that doesn't have the 2C target built in. But this shows you what happens if you force the model to meet the 2C target. This model has no uncertainty. So let me show you a new result from literally calculations myself. And Kate Rickey, along with Oliver Wharton, providing some input, are just sort of playing around with right now, which is using some models that do have climate uncertainty, climate sensitivity uncertainty built into them, to look at what are the trade-offs in terms of the probability of staying under the kind of arbitrary 1.5-degree C target based on two simple parameters. So one parameter is the rate, the percent per year, which we count fossil fuel emissions. So this is where I want to get away from the complexity of big integrated assessment models and just make a simple parameter where nobody here has any, you're all free to choose whatever value of the parameter you like. So just to be clear about what's in the carbon model here. We simply take the last 30 years of CO2 emissions and linearly extrapolate them in the future. So assume the baseline is just a continuous extrapolation, linear extrapolation of CO2 emissions. And then we assume that starting in 2020, those emissions are reduced by a certain percent a year. Quite a lot of other papers have made this. This is a very convenient parametric way. So if you reduce emissions by 10% a year, you get this very rapid drop off. So emissions basically go to zero in 15, 20 years. So that's this number, percent per year reduction. And of course what you see, and then what's plotted is the probability of staying under 1.5. So what you see is if you do no solar geoengineering, even if you cut emissions by 10% or 12% a year, more or less equivalent to having a global war, you still actually don't have that good of probability of staying under 1.5. This is in a model that allows no overshoot. So it's the rigid interpretation of 1.5 and there's some calibration issues about the way that edge is, but it's roughly right. And then this shows you another parametric thing which is how quickly you're ramping up solar geoengineering, again arbitrarily starting in 2020, but we can try starting different times, where the ramp up rate is in watts per square meter per decade. So again, you're assuming a solar geoengineering is not like turned on as a massive switch, but that you're ramping it up slowly and the issue is how quickly you're ramping it up. And then you're looking at the trade off between those two things in terms of the probability of staying under 1.5. And what you find, and what I think is important here is what I consider to be, in some senses, relatively small ramp rates for solar geoengineering at 10th or two tenths of a watt per square meter per decade, which would have peak solar geoengineering, it's still of order of one watt per square meter, reduces the probability of getting under 1.5 or improves the probability of getting under 1.5 really dramatic. And you can produce the same kind of results for two or whatever you have. So, inevitably, what I was getting to there was stuff that was effectively about what happens if we deploy this technology. But I wanna come back with one last slide before I take questions and point out that the real question that we face now is not about deploying solar geoengineering. It's about whether we move from where we are now, which is where a small group of sort of enthusiasts essentially have run a little sort of out of their bootstrap solar geoengineering research program. So we had Harvard, we now have formally a Harvard-wide program signed off by the provost, but it's still $2 million a year class program, a bunch of different faculty involved. And that's just, that's one of the larger programs in the world, and of course that's tiny. And so the question is really, do we have a serious solar geoengineering research program? Do we take this thing collectively as a community from being this little thing that a few people are interested in and sort of build it into the mainstream of climate and environmental science? So we think about an integrated way in the way we think about climate policy, and we think about an integrated way in terms of how we think about its environmental impacts and how we think about its technology and so on. That is what I think is the real question is, do we build a serious research program? And that's a matter of strong debate. So there are people who I respect a lot in the climate science and policy community who really think it's a bad idea. Who think we should not have a research program. And their fundamental reason is fear of what I think is often called the moral hazard of that's not quite the right name. Let me give you, I think, the more functional name for it, the functional explanation. The reason that people oppose research is fear, and I think a well justified fear, and I think it's in fact a certainty, that some forces that want to block emissions cuts will exploit this research program to try and block emissions cuts. And I think that is going to happen. I don't think that's a reason not to do research at all, but I think it's important to be clear right about the fact that that will happen. So that forces like large fossil fuel companies or whoever that want to argue against emissions cuts will use results from a research program. They have cherry picking them to show how great solar chair sharing is and arguing that we don't need to do emissions cuts or we should put off decisions, that's gonna happen. From my point of view, that is a reason to think about, we should think hard about how, as a community of researchers, we deal with that reality. We should think about the way we develop this information, but I don't think that fact alone is a reason not to develop information. Happy to answer more questions about that. Let me give you a little kind of toy diagram about the way I think in a very simplified way about the decisions about starting a research program. So should we start a research program? If we did start one, I hope it would be international, open access, multi-disciplinary, I'm sure all this won't be true, non-commercial, but that's what I'd like to see. And I don't think that's crazy. Of course, it may produce negative findings. That is we may find that it really works much less well than I thought. Maybe there's some deep thing in climate models that makes the results look as good as I showed you earlier and it's a particular bias in climate models we hadn't thought of and we find it. And then we really say this just doesn't work as well as the early advocate stuff. Then nothing much has happened. Maybe we find that the kind of stuff I've been telling you is broadly true with lots of new information. Then we get to make, and again, who's the we? Then lots of different we's are on the world because the world isn't run by one we. But then there are decisions made about deployment and those decisions are more informed by facts. And to be clear, even if we do deployment, there's risks, there's risks to deployment, there's risks to not deployment. There's no risk-free answer here. If we have no research program, which has really been effectively a decision over the last decades, decision has been to say, we're going to keep a lid on this thing. That doesn't mean that there will be no deployment decisions. These basic ideas here are old. There are going to be countries that are going to face very strong climate damages in the next decades. And I think it's essentially a certainty that they will at least consider deployment of this even if we agree there's no research program. So then they're going to make an uninformed deployment choice. Obviously this is much more binary than real world. So I made this a very kind of binary, simple, yes, no thing. The real world is more fuzzy. I get it. I'm just trying to find a simple way to communicate this. My view is simply that providing more information makes us more able to make sensible decisions. I'll try one last way to say that, which is simply that my generation, I'm 55, will not make the decisions about solar generation deployment. Those will be made by the next generation or the generation after that. But we're going to make decisions about how much information they have to do it. Because research programs take a long time. Developing really robust knowledge about how to do this is going to take a decade or two. And if we decide to keep on doing what we're doing now, which is having no serious research program, then they'll make decisions on the basis of the kind of ignorance we have now. If we have a really serious research program, they'll have a lot more information. They won't know everything. There's no possible way to do that. Even if we have a big research program, there will still be deep uncertainties about exactly how well this works, about exactly how to do it. But there will be less uncertainty than there is now. And my view is that we ought to have a really serious integrated research program. So I could take questions. So thanks, David. We're running a little bit long, but we do have time for a few questions. Let's do you student first. Great, right in front. And then John Doich is an elder student. They're deeply different points. So first of all, let me unpack some things you said. So you said carbon removal or geoengineering. I think they are so different that it's not useful to think about them as two forms of geoengineering. So this problem is hard because many entities can do it. It's so cheap, stratosphere, geoengineering, that basically almost any government could do it. And the effects are global. But there's no particular local risk. The risks are kind of global and globally distributed. That's a particular kind of government's problem. Carbon removal, more or less there's local risk. So anything you do for carbon removal will have some local environmental risks, but more or less there's a pure global benefit. So that structure is just like mitigation. So for mitigation, if you spend money to cut emissions, there's some local impacts of spending the money to cut emissions and there's a global benefit. So carbon removal more or less in terms of political problem looks to me like emissions mitigation. The problem is basically how to share the cost burden of doing it and the local environmental impacts burden of doing it. But solar geoengineering is a very different governance problem, so both governance problems. In some ways, I'll give you some ways in which it's an easier and harder problem. In some ways it's an easier problem because it's cheaper. So one of the fundamentally hard things about cutting emissions is simply that it's expensive. Nobody really wants to pay. Solar geoengineering doesn't have that problem, but it does have other problems. So there's problems of risk, there's problems of weaponization, there's problems of what if countries really disagree about how much we should do. I think there are actually a bunch of ideas now, including ideas in literature about coalition formation and about how those things would happen, but there's no magic answer. I'll answer one question that comes up a lot which is worry about the fact that if we've gradually ramped it up. So it's often said that once you start you have to keep going forever. That's not true at all. You can certainly start and then ramp down slowly. And you can certainly start and ramp up slowly to say three tenths of a watt per square meter, learn a whole lot about risk in the stratosphere and still have a very small termination shock, three tenths of a watt isn't that big. But it's true that if you ramp it up to like three watts per square meter or two and you stop suddenly you've got a risk, for sure. But that's one where I think that politics is really interesting. I think in practice it's very hard politically to stop what's you've started because even countries that oppose deployment would see it in their self-interest to maintain the ability to deploy if other people stopped because it's risky to stop once you've got it established. And Oliver Morton has a billion analogy which is the GPS system. So we now have, I guess we're coming on to four globally independent GPS systems. The reason is purely because the Europeans don't want to rely on ours. Ours is a free public good, but the Europeans want to have their own and the Chinese have their own and the Russians have their own now. Let's do John, another student and then Arun, maybe we're probably out of time. So John Deutsch. I have a question about your two dimensional font. Mm-hmm. It's not empty, one for five, three. What do you think the next two is for the budget more or not that? Is that really able to make some of your points? Yep. I think the problem is, here's, this is what's tough. So if the kinds of results I showed you in that first section are corrected, if it's really true this only costs of order, billions of dollars a year, then the ratio of cost to benefits, monetized cost to benefits is actually like vaccines. It's like thousands to one, literally. You can do the math, you've got that kind of memory. But I think all of us suspect that it can't really be that good. So I think there's stuff we're leaving out that we don't know yet, which is the reason to do research. So I don't, my guess is it's not gonna end up looking that good. And I think if you just look at that monetized cost benefit, I basically, it looks too scarily good, I guess is the answer. But yes, you're right, we do have to look at that. Okay, I'm looking for a, but I guess what we really need is a risk axis. Looking for another student question. Students, students. So you talked a lot about the risks of temperature and changes in temperature and precept, and sort of along the lines of what you're just saying, what do you think are the sort of most important other risks that your research program would study? Let me divide them into two categories. To me, the biggest risks I really worry about are risks of gross misuse. And that's where I think the choice of which technology makes a big difference. So if for example you just do stratospheric aerosols in one hemisphere, you get big asymmetries. You move the ITCC, the region of intense rainfall, creating droughts on one side, that would be terrible. And I think no kind of rational disinterested single entity who was trying to maximize utility, the world would ever do that, but it might happen. And so those to me are really the biggest risks. And I think we can think a little bit about what technologies are more or less prone to misuse. And some of my kind of offhand comments about weather control and local manipulation ahead in that direction. In terms of sort of more technical unknown unknowns, I'm most concerned for stratospheric aerosols about upper troposphere. And the reason is for a given flux of stratospheric aerosols, obviously you should think all the way down to health impacts at the ground. Actually, Seb Eastam, a student of Steve Barrett at MIT has now done that. But for a given flux of stratospheric aerosols, it turns out the impact on aerosols in the boundary layer is really pretty darn small. It's not zero, but it's small. But the impact on the upper troposphere is pretty big. And I think we really don't know enough about what that does say to serious clouds. And that gets into real questions where we just don't know the science about what fraction of serious clouds are driven by ice nucleon or not. So that's the thing where I think it's clear that a well-structured research program could answer that question, but we just don't know the answer. And there are ways in which actually it could make things better or worse. And we don't know what the sign is. Last question to Arun. A couple of questions. There you go. Thank you. Thank you. So this is a question. What is the most fire-resistant and only the most high-resistant? And that is the most hot-resistant question. What is the most hot-resistant and only the most hot-resistant and only the most hot-resistant? Yeah, yeah, yeah. So I think the biggest thing we learn from volcanoes is a level of in what I obviously everybody has their own view about plausible. But what I view as plausible employment scenarios where geo-insuring is part of a portfolio of solutions and you use a watt per square meter or so, then if you do it effectively, you're talking about adding maybe 2 million tons a year if you were doing sulfur, the stratosphere, per year at the peak. And I think the big thing we learn from volcanoes is not really climate response or something. It's a kind of double check on unknown unknowns. So a reasonable point of view that you might say to me is, hey, David, your stratosphere as a high-speed is pretty smart, but in the end, am I ever going to trust all these models? And I think the big thing you learn from Pinatubos, we know that putting 8 million tons in a single year didn't do anything really terrible. And I think that gives you a lot of confidence that putting 2 million tons a year is likely to be not that far out of the range of variation that our models mean so it doesn't mean that we're exactly right, but the fact that we hit it with an 8 million ton hammer in one year means that if you put a million tons a year in or 2 million tons a year, you're still inside the range of natural variability we've observed and modeled and you're less likely to find some huge unknown unknown. So you also asked about how do you stop? Can you take particles out? I think the main way you stop is simply by waiting. The particle lifetime is about a year and a bit. It's not that long. And I think the point is you never start by turning on the switch and putting in 2 million tons a year, that would be nuts. You start by ramping it up very slowly, putting more in some area, looking at the stratosphere chemical response, monitoring it with limb sounders and with in-situ measurements and sort of working your way up. And then if results don't look good, you just let it settle out. There are potentially ideas for removing particles. We published a paper on counter geoengineering with some kind of wacky ideas, but I don't really believe them that much. They seem very science fiction. With that, let's thank David one last time. Thank you.