University Place

Thank you all for being here today. I am Larissa Back, an Assistant Professor in the Department of Atmosphere and Ocean Sciences and the lead organizer of this event. This is the sixth annual Department of Atmospheric and Ocean Sciences Len Robock lecture, supported by the estate of Len Robock. The goal of this lecture series is to bring dynamic, high-profile speakers to speak to a general audience about topics related to atmosphere and ocean sciences.

We chose tonight's speaker, Dr. Kerry Emanuel due to the unique perspective and expert knowledge he has on climate science and hurricanes. He's certainly high-profile; he was named one of Time Magazine's 100 most influential people in 2006 and is a member of the National Academy of Sciences. He's made numerous contributions to hurricane science, energetics and, including more recently, developing a promising technique for inferring tropical cyclone activity from coarse grain output of climate models or re-analyses.

Dr. Emanuel is visiting us from MIT where he is a Cecil and Ida Green professor in the Department of Earth, Atmospheric and Planetary Sciences. He's also a co-founder of the Lorenz Center, which is an MIT think tank devoted to understanding climate. Emanuel is working on tropical cyclones and various aspects of moist convection in the atmosphere.

He's the author of two excellent books, "The

Divine Wind

the History and Science of Hurricanes" and "What we Know About Climate Change." Let's welcome our speaker to the sixth annual Len Robock lecture to discuss "The Problem of Hurricanes and Climate". (applause) Thank you very much, Larissa, for that great introduction. It's wonderful to be here with you all tonight, and I get to do something I like to do very much which is to talk about hurricanes. Even though it's the middle of winter, or at least it still feels like the middle of winter.

So, let me begin with this. Back in the end of the 19th century, the Swedish chemist Svante Arrhenius started to worry about what would happen if we increased the carbon dioxide content of the atmosphere, which he felt we would do by burning fossil fuels and he did some calculations. And those calculations suggested to him that for every doubling of carbon dioxide, you'd raise the surface temperature by about 4 degrees C, which is still within the bounds of, you know, a hundred years later what we estimate the propensity to be. I would argue that Svante Arrhenius' prediction is coming true, unfortunately.

So what you see on this graph is the global mean temperature of the planet in green, and the blue curve is proportional to the natural log of the CO2 content of the atmosphere which is, in turn, proportional to the radiative forcing in the system. And although there are a lot of fluctuations that are natural, pretty much the temperature is following Arrhenius's prediction. There aren't scientists around, credible atmospheric scientists, who don't believe this is true. We're warming up the planet.

We're performing an enormous experiment. We're taking a big gamble with future generations. We don't know how it will turn out. One of the interesting sub-problems of this, is what's gonna happen to hurricanes.

And, hurricanes are, of course, very destructive around the world, if you've been paying attention to the news, the horrible news coming out of the South Pacific with tropical cyclone Pam, and so forth. So, what I'd like to do tonight is to, why did this come out, out of order, I don't know, is to first give you a quick overview of hurricanes, so that we're on the same page, and to talk about the basic theory of hurricanes which I'm going to use to guide a discussion of their sensitivity to climate, and ask what have hurricanes been like in the past and how might they be affected by global warming. And at the end, if there's time, I'm going to talk about this very interesting idea, is how do you actually go about making estimates of risk in the current climate, but also in the changing climate. So just a brief overview of tropical cyclones.

We'll begin with a basic question, "What are we talking about here?" There is actually some confusion on this, what's a hurricane and if you look up in a glossary, it will tell you, hopefully, that it's a tropical cyclone. Okay, okay, that tells us what it is. But, it's a specific kind of tropical cyclone that has sufficiently strong winds and occurs over the North Atlantic or eastern North Pacific. So, it's a regional name given to a subset of a class of phenomena called tropical cyclones that's given to it in the Atlantic and Eastern Pacific.

So, of course, you want to know what's a tropical cyclone. This is one of the official definitions of it. If you look it up in a glossary, it's a rotating storm system between 100 and 1,000 kilometers in diameter, low pressure, strong winds, spiral bands of thunderstorms, and so forth. I think you've all seen nice pictures of them, and I'll show you a few in a minute.

Now the word "'hurricane" comes to us from the tribes that inhabited pre-Columbian Caribbean region. The Mayans, the Tainos and the Caribs. They had words like "Huracan", "Juracan", "Hunraken" which was their God of Evil. And also, invariably, played a role in their creation myths.

And at the left, there, you see an image of Hunraken lifted off of an ancient Cuban vase. And you see this sort of spooky-looking skull and where there should be ears, there are these spiral bands which look remarkably like the symbol that modern forecasters put on weather maps to denote the presence of a tropical cyclone. It's interesting, because there's a suggestion that the ancient Mayans and Tainos, and so forth, understood that hurricanes are rotating wind systems. So they're whirlwinds, basically, which Western science didn't figure out until the middle of the 19th century.

Now we don't know that for sure, but there are a lot of images like this, and they had the correct sense of rotation for the Northern Hemisphere. So, it's just something we can think about a little bit. And they, of course, the word was brought back to the Old World by Columbus and his successors. They made deep impressions on artists, including Shakespeare, who wrote hurricanes into several of his plays, including his last one, "The Tempest", which, we think, he formed out of accounts of a big hurricane that lead to the founding of Bermuda in 1609.

Well, here's a pretty photograph from space of Hurricane Floyd which you see approaching Florida here, although it ended up going up into the Carolinas, doctored by the scientists at NASA to give it this kind of three-dimensional perspective. You see this whirling mass of clouds a few hundred miles across spiral bands of thunderstorms. The heaviest wind, the strongest rain, the deepest clouds are arranged in kind of a doughnut around an eye which is often clear. And here's a closer picture of the eye of Hurricane Igor in 2010.

So you can quite clearly see this annulus, which we call the "eye wall," and this clear funnel-shaped region called the "eye." Usually there's some clouds in the base of that eye. Since the 1940s, we have routinely flown specially-equipped aircraft into hurricanes to make measurements. Here are two that are currently used by NOAA. They're Orion P-3s, four turboprop engines.

They penetrate most of the storms in the North Atlantic that threaten land. But nowhere else in the world besides North Atlantic, unfortunately, is this done in any sort of routine way. And if you fly in, you might see a scene like this. This is the eye of an Atlantic hurricane, and it's like standing in a Roman coliseum, except the coliseum is 10 miles high and maybe 20 or 30 miles across, and it's white.

And you see the blue sky of the stratosphere up here. Sometimes you see a cascade of ice crystals like a waterfall flowing down the inside. Now, this is a really magnificent sight, and you can't do it justice with a two-dimensional picture. So when I retire, I'm determined to start a hurricane safari operation, okay?

(audience laughing) And if you'd like to sign up, come and see me. If we look at the climatology of these storms, you see a map here of the tracks of tropical cyclones globally between 1945 and 2006 color-coded by something called the Saffir-Simpson scale which measures how strong the winds are. Here's the familiar belt of hurricanes in the Atlantic that the Mayans and the Tainos knew about coming up and curving off, sometimes to the North and East, sometimes heading up into the Gulf of Mexico or further south. A very active belt in the eastern North Pacific which you usually don't hear about because, fortunately, these storms rarely affect land.

Every once in a while one will make it out to Hawaii. A very active belt in the western North Pacific here and across the Bay of Bengal and the Arabian Sea, and then another belt in the Southern Hemisphere from the Central Pacific across Australia and the Indian Ocean. You'll notice that there aren't any hurricanes close enough to the equator. That's because they're rotating vortices.

You need to be able to take the earth's rotation axis and project it onto what you call "up" locally and not get zero. But at the equator, there's no projection. So the moral of the story is if you, or your retirement project, which is another one of mine, is to sail around the world and you're afraid of hurricanes. Well, you just stick to the equator and you'll be fine, though you may have other problems.

(audience laughter) And you might want to reflect on the fact that there are very few storms in the South Atlantic. So if you want to discuss that, if there time for questions, we can talk about that. Almost invariably, somebody asks that question. These are creatures of the summer and the early fall.

So this is the number of storms per month in the Northern Hemisphere, read off of this upper axis. And in the Southern Hemisphere, read off of this. So if you look at these two graphs, there are more in the Northern Hemisphere over there and they tend to come a little bit later in the year, but basically summer and early fall and not very many in the middle of winter, okay? Now let's get into a little bit of the theory of this because I think this will help us understand how hurricanes respond to climate change.

We don't do this entirely empirically. We actually use all the information that we have available to us, including that that comes from theory. And a hurricane is, in essence, a pretty simple example of an engine. A heat engine that turns heat energy into mechanical energy.

In this case, the wind. And so, here's how it works. This is a cartoon, and it's meant to be a kind of cross-section through a hurricane. So this is the center of the storm, the rotation axis, and then this is the distance from the center.

This is altitude. This is meant to be the eye wall cloud. And if you were to be so unwise as to go for a hot air balloon ride starting at point "A" here, you would be swept in toward the center of the storm, spiraling around it at the same time. And when you got to the eye wall, you would be thrust up very rapidly into these deep thunderstorms in the eye wall.

And if you survived that experience, you'd be ejected out the top. And over a period of several weeks, although in practice the air actually will get wrapped up in other weather systems, but in models at least, over a period of several weeks, it will sink back down to "A". And the lurid colors here are a measure of the heat content of the air that we call entropy. And you'll notice that as you go from "A" to "B" the entropy is increasing, going from cold colors to warm colors.

And that reflects the big source of heat that powers the hurricanes, which is transfer of heat from the ocean to the atmosphere. That's what makes hurricanes run. And it's made possible by the fact that the atmosphere in the ocean in the tropics are not in thermal equilibrium. And that's reflected by the fact that the air near the surface isn't saturated with water.

Water can evaporate from the ocean, and does. And that evaporative potential is what runs hurricanes, and it's why they die as, sort of begin to die, as soon as they go on land, okay? That potential, in turn, turns out to be because we live in a greenhouse world. Greenhouse gases in the atmosphere, most importantly water vapor, don't let heat get out of the ocean.

They sort of radiate it back in. But the ocean is absorbing all this energy from the sun, so it has to get rid of the heat to be in balance. And it does so, mostly, by evaporating water. To do that, it has to have this evaporative potential, and that's the energy source for the hurricanes.

Anyway, if you come into "B", you go up. You notice you go up a line of constant color. That's constant heat content, we call that an adiabatic process. Then all the heat it got from the ocean is ultimately radiated away in the form of infrared radiation, space, and then we have another adiabatic leg.

Well, it turns out that these four legs, isothermal expansion to low pressure, adiabatic expansion, isothermal compression, adiabatic compression, are almost magically the four legs of the most efficient engine that you can possibly, theoretically build, that operates between two fixed temperatures, as discovered by the 19th century polymath, Sadi Carnot, and now called the "Carnot Cycle". So if anybody's had Thermo 101, they've been told about the Carnot cycle, but they probably haven't been shown an example. This is a natural example, and it's almost spooky that it is so efficient at converting the heat energy of the ocean into wind. My mother told me I should never present an equation after dinner.

And I have to now apologize to her 'cause I'm gonna do this twice, just twice, okay? Knowing that, you can actually write down an equation. You don't have to understand this. You just have to know it exists.

It's the E=MC of the hurricane world. It tells us the maximum wind speed, in this case the maximum wind speed squared you can have, is a function of somethings that go on in the environment. This is just, this thing in box you don't really need to worry about is approximately one. The surface temperature of the ocean.

The temperature way up at the top of the hurricane, we call the outflow temperature. And then this is the real driver, that term in parenthesis, it's the thermodynamic disequilibrium between the ocean and the atmosphere. And so you can calculate that pretty easily from atmospheric data. And here is a map showing the annual maximum value of this quantity in the current climate.

So at every point, we look at the whole seasonal cycle. We pick out the maximum value and then we contour that. And so, you can see that the maximum, oops, I pushed the wrong button. Push it again, let's see if I can go back to this.

Yeah, here. The maximum wind speeds are 80 meters per second. That's about 170 knots or 190 miles an hour, or something in that. And in the western Pacific and regions of high potential in the Atlantic and eastern Pacific, and essentially no potential at all where the water's cold.

We don't observe hurricanes forming when this potential is less than about 40 meters per second, this light green color. We just don't observe that. So, we can map that out. And one of the things we can ask is, "How does that quantity change with the climate?" Well, does it have anything to do with real storms?

So this is just a chart, a plot showing the number of storms that achieve, the number of storms that achieve this fraction of the theoretical intensity. This curve is for just weak tropical storms. This one is for hurricanes and typhoons. And you can see the vast majority don't, but a few get close.

So this thermodynamic limit means something. But most storms don't get close to it. It's like saying that the speed limit on the highway is, let's say you upped it to 120 miles an hour. Well, you still find most people not driving that fast, and a few nuts who are driving the speed limit, right?

Something like that. So, how does this quantity, it's one of the easier things we can ask, is we can do the calculation. How does the speed limit for hurricanes change when we change the climate? Well, we've done a lot of work on that over the years.

And one of them, one of the things we can do is to run a really simple model called a single-column model which represents the whole atmosphere as a single pillar of gas, probably the oldest kind of climate model but still very useful. And this is this potential wind speed as a function of the number of doublings of carbon dioxide. And you'll notice that it first rises pretty steeply, but then it levels off. And so you can't keep doing this indefinitely by putting greenhouse gas in the atmosphere.

What happens is, up here you have so much greenhouse gas that there's essentially no infrared radiation leaving the ocean at all. And then there's nothing more you can do. You keep putting up gas, you'll raise the temperature but you won't raise this potential anymore. But there are other things you can do.

So this is the same curve, this red one I just showed you. We graphed it versus, in this case, ocean temperature rather than doubling CO2, when you double CO2, or triple it or so forth, the ocean temperature goes up and so does the potential intensity. We're just graphing one versus the other. But if instead of doing carbon dioxide, you did sunlight.

If you just turned up the volume of the sun, it goes up along this curve. So it rises faster, and that keeps going. There's no limit to that, if you keep turning up sunlight. If you turn up the conversions of heat in the ocean itself by changing ocean currents, you can get it to go up a little bit faster.

The really fast way to do it is to reduce the trade wind speeds at the surface. That is the background wind in which the hurricane exists. If you lower the wind speed, you have to have a more, a greater evaporative potential to get the heat out. So, that's how you get a really strong storm is if you change the climate so that it's less windy.

For whatever reason. Now, here is a map showing an estimate over the last, well, over the thirty-year period from 1980 to 2010, of the trend in potential intensity from climate data. And where you see reds, it's going up. Where you see blues, it's going down.

Where you see white, the slope is not significant in a statistical sense. And you can see over most of the world, according to this particular re-analysis, this is going up in the tropics. If you can see the difference between light and dark red, you'll see that in the Pacific region, especially, it's going up most rapidly in the sub-tropics, that is at the periphery of the zones that now generate hurricanes and less rapidly in the deep tropics, which is kind of interesting. That's not quite true in the Atlantic, though, but it's true in the Pacific.

If you look at climate model projections for the next hundred years, assuming that we don't do anything about the increase in CO2, you see kind of a broadly, similar pattern that's projected. This is in meters per second per decade, like last chart. It's going up. It's projected to go up, particularly in the sub-tropics, less so in the deep tropics, and to go down in many places in the Southern Hemisphere.

Although to be fair, these are places that don't have hurricanes today. They're getting even less favorable for hurricanes. And along the coast here, we don't see hurricanes developing and certainly not off California. So in some places, the favorable places get more favorable and the unfavorable places get less favorable according to this projection.

So if you look at a variety of climate models, the patterns are different in their details. But these broad features I just described, maintain. And what that means is that we think the speed limit is going up, and we think it will continue to go up. It doesn't mean that all hurricanes will get more intense.

But it means that the most intense ones of the future climate should be more intense than the most intense storms of the current climate. Now, of course, it's not just thermodynamics that influence hurricanes. There are other things that are known to. The fact that the background winds change with height, we call that wind shear.

The rotation of the large scale of winds at lower levels, we call that vorticity, and we can come up with a kind of empirical equation, never mind all the math here, that's just for the atmospheric science people, I threw this slide in, that looks at this genesis potential index over time. And we are gonna look at not the parts that depend upon wind shear, but just the parts that depend on the thermodynamics. And we do that, we see for the, there's this simple model. Actually as you warm the climate, the number of events should go down, not up, which is kind of interesting.

And a lot of people who worked on this problem using many techniques have come to that conclusion that the actual number of storms go down. Now that might seem like good news. In some ways, it is. The problem is that almost all the damage that's been done historically by hurricanes has been done by the granddaddies, the really big ones.

Whereas, the smaller ones are much more numerous. So sort of the general consensus now is that there'll be fewer weak hurricanes, which are most of them, but more strong hurricanes. And that's not good news. The other thing that is a strong consensus about is that a given hurricane of a given size and intensity will rain more, just because there's a lot more water in the atmosphere when you're at high temperature.

This is the same model I talked about before. It's a predicted trend over the next hundred years. And this genesis index, it's negative in a lot of the deep tropics, but it's positive at the periphery of the tropics which suggests that maybe hurricanes will start to expand pole-ward. What can we learn now that we've done the theory?

And I'm not going to expose you to very much more of that. I'm sure you'll be relieved, my mother certainly would be, and talk about what does the record actually show? What can we learn from hurricanes in the past? Well, here's something interesting to get the conversation started.

This is a record, in red, of the power of North Atlantic hurricanes going all the way back to 1870. It's a measure of power or how much energy they dissipate. And the blue curve is the ocean temperature of the tropical Atlantic where hurricanes swarm in summer. And we don't actually believe this red curve much before about 1950, because the hurricane data is pretty bad when we go back.

But we do believe the ocean temperature curve. And you can see that there's sort of a nice correlation, first of all. Except this period, from about 1939 to about 1945. Now the older people in the audience that might strike a bell.

What was going on there and we can talk about that, of course, it's World War II. But the other thing I want to draw your attention to is there's sort of two things going on. On a long time scale there's this dip, there's a rise, a dip and then an abrupt rise. From the short time scale, you have these fluctuations that have a period of about a decade which we don't really understand.

But they're quite regular. And you see them both in the hurricane power and in the ocean temperature, rather remarkably. Alright, so that's what the North Atlantic data. Now I should have said when we talked about climatology, North Atlantic hurricanes are only about 11 or 12% of the world's tropical cyclones.

Most of the tropical cyclones by far on the planet are in the Pacific and Indian oceans. But these guys get the press. But that's not why I'm showing you that. I'm showing you this chart because we have the best, by far, the best data on North Atlantic hurricanes.

Everywhere else, it's very spotty. We don't even know how strong Hurricane Pam, that occurred a few weeks ago was to within a reasonable margin of error. We're doing a terrible job measuring hurricanes, I would argue. We know that they exist.

Now there's been some very nice work done here in the Department of Atmospheric Sciences here at Wisconsin. And this is a paper that came out just a couple years ago by Jim Kossin and collaborators showing, based on satellite data, trends in hurricanes globally, trend in hurricane intensity by something called a quantile. So these are the weak storms over on the left. These are the stronger storms on the right.

There's a margin of error shown by gray here, but you can see that there is some tendency for an increase in the incidence of strong storms. And this is something that Jim and his colleagues have done more recently that is even much more profound and a much stronger signal, in some ways. And that shows, I don't know I dropped the time scale here, but I managed to do that, trends over time in the latitude or the distance from the equator, in this case, at which hurricanes, tropical cyclones reach their peak. The blue is based on satellite data.

The red's based upon other observations, for the most part. And you can see that over time. What's the span of years? I've forgotten the-- Yeah, 1980 to 2010.

That the latitude at which the storms reach their peak intensity has been progressively moving northward in the Northern Hemisphere and southward in the Southern Hemisphere. That's a pretty strong signal. And we'd like to understand what's going on there. It's nice to have Jim Kossin here.

I can ask him when the thing gets dropped from the axis. Thank you, Jim. I'm going to now do something a little bit new for the sake of the atmospheric people, but I think the rest of you will find it interesting. I'm going to propose a hypothesis that might very well be wrong, otherwise, it wouldn't be a hypothesis.

And that is that this longer period signal that I referred to earlier, this, the long period variations, particularly in the 20th century, were mostly due to man but not carbon dioxide. But from another kind of pollution that we emit in the form of particles called sulfate aerosols. And I'm gonna show you some evidence for that. So let me sort of present the idea.

We know from other work done by large numbers of other researchers that in the late 20th century, the radiative effects of these man-made aerosols sort of had globally comparable magnitude to the change in radiation due to the change of CO2 over that century, basically. So they're sort of the same ballpark, not something we know. But another thing we know, if you remember a graph I showed earlier, is that per unit radiative forcing for, per one watt per meter squared. The short wave forcing has about twice as much effect on this potential intensity as the CO2 forcing.

It's much more effective to change the short wave forcing. And that's what aerosols do, they affect sunlight. They don't really affect the infrared radiation leaving the planet. So, we know that.

A short wave forcing should be a little bit more important or somewhat more important for hurricanes. And from work that's been done on aerosols by a bunch of different groups, we know that in the tropical Atlantic, a lot of the inter-annual variability in the aerosols is thought to be owing to European, sulfates of European origin. If you look at the sulfates in places like Barbados, they're coming mostly from Europe, not from North America. And they are thought also to interact with natural aerosols coming from Africa to give you a fairly large signal.

So we're gonna start with that. See, I'm gonna blame Europe for a lot of things tonight. Here is a record of the emissions of sulfate aerosols by European countries going all the way back to 1850. It's kind of interesting.

It's a record of fossil fuel combustion, up to a point. Then it's a record of regulating pollution after that point. So there's this gradual increase through World War II. And then after World War II, a huge increase which corresponds a big industrial expansion, population growth in Europe and so forth.

And then regulations start to kick in. We had the Clean Air Act in the U.S., we had a bunch of parallel acts in Europe. And although fossil fuel combustion hasn't gone down, this particular kind of pollution has been controlled. So it's plummeted like that.

So I want you to remember this curve. It peaked in the 1970s. It peaked in the United States in the 1970s. And I know that in my office at MIT in the 1970s on a hot, humid, hazy day, haze is largely sulfate aerosols.

I couldn't see the buildings in Boston across the river from my office, typically. And today that's no longer true. It can be hot and humid and I can see those buildings clearly. It's really, if you pay attention, you notice these things.

Alright, so we're gonna take that record, and we're gonna form a hypothesis that the sort of long time scale variability of the 20th century is owing to both the sulfate aerosols and the long wave radiative forcing of CO2. As we're gonna try to predict that signal knowing only CO2 and sulfate aerosols. One thing we don't know very accurately is how strongly the aerosols actually affect sunlight at the end of the day. And we're gonna also hypothesize that that shorter period variation, the residual once you've taken out the long period of variation, is a natural oscillation that has nothing to do with radiative forcing or little to do with it.

For those of you in the know, nominally equivalent to something called the Atlantic multi-decadal oscillation. And to test that idea, we're gonna just take that record of hurricane power and separate it artificially into long period variability and short period, or quasi-decadal, variability. We're gonna just simply try to fit the longer period to two quantities, the carbon dioxide and European sulfate emissions. We're going to use those two things as predictors and try to predict that long period signal.

So, let's do that. And then once we've done that, we're going to look at the residual and see what the pattern, the spatial pattern of that residual looks like. So to start off with, this is the correlation between power dissipation and this thermodynamic speed limit. The potential intensity in all that correlation, not surprisingly, is in the belt where hurricanes tend to form in the Atlantic, and it's pretty large there,.6 or R squared is.6, but, you know, it's not much influence of what's going on out here.

And here's what happens when we take the long period signal, which is the blue curve. It's just a long period component of the storm power, and try to fit it. We got this fitted curve which is in green. And the contribution of sulfate aerosol to that green curve is this red curve.

And the contribution of carbon dioxide is this purple curve. So you can see the really big effect in the North Atlantic, according to this analysis, was this sulfate pollution. It's just much more effective at changing the thermodynamic speed limit of hurricanes. But there has also been, over this period, some increase that can be laid to carbon dioxide.

That's kind of interesting. Now if you take that fitted curve and you subtract it from the actual hurricane power, you get this pattern of short period variability. And it has, it turns out to have a peak power at about six years, but I think it's probably, that's probably a little bit shorter than what's really the case. It's maybe more like ten years.

But it's a much shorter period, and it looks a little bit random. And if you look at the spatial pattern of that, this is just correlating that variability with power dissipation. I'm sorry, with the potential intensity. You see this interesting pattern with a big maximum around the Azores off the coast of Africa, another one up here off of Spain, and nothing going on here in the western Atlantic.

If you look at climate models, here are the climate models' renditions of the amplitude of this kind of a higher frequency mode. This is 25 recent generation couple climate models doing this kind of dance. And their periods range all over the place, from fifteen to a hundred years. But their spatial pattern is that.

It looks pretty similar in many respects. So we think we're seeing that. And in the climate models, it's demonstrably a natural fluctuation, I didn't tell you. Those particular climate models were not forced with either sulfate aerosols or carbon dioxide, they just ran freely and naturally produced that chaotic variability.

So there's a lot going on in the Atlantic. But I think the big hurricane drought that we demonstrably had in the '60s through the '80s or early '90s may very well have been because of European aerosols. And when they cleaned up their act, unfortunately, the hurricanes came back. And, no, I'm not advocating they go back to burning dirty coal.

Although, Germany seems determined to do that right at the moment. Alright, so let me now turn to the story of hurricane risks, because I think it's very important. If we start to get a handle on being able to predict how hurricanes might respond to climate change, the next step is how do we respond to that? Do we do nothing?

What should we do about it? So, what are the risks? Well, there's the obviously one is wind. Wind, of course, all by itself does a lot of destruction, as you see if you saw the terrible pictures from Tropical Cyclone Pam in the South Pacific.

Hurricane Andrew in 1992, almost all the damage was from wind. But, in most hurricanes, it turns out the most destructive and lethal aspect of it isn't directly the wind, but the storm surge which is driven by the wind. It's an inundation that occurs usually about the time the center of the storm goes over. It's flooding, and it kills a lot of people.

In East Pakistan in 1970, a storm surge associated with a tropical cyclone killed literally a half a million people, a half million, okay? In one go. And we hope that will never happen again, because that country, by the way, is now called Bangladesh. And the fact it's Bangladesh, East Pakistan is partially a result of that storm, by the way.

It's an interesting historical story. The next most lethal aspect which is seldom talked about, at least in this country until we had Irene a few years ago, is freshwater flooding caused by torrential rains. And the second deadliest natural catastrophe in the western hemisphere was hurricane Mitch of 1998 which killed 12,000 people, all from freshwater flooding. So those are the risks that we need to be concerned with.

This is a aerial photo of the horrible tsunami in Japan not very many years ago. And you've probably seen gruesome pictures like this. Hydrodynamically, the storm surge is identical to a tsunami. The only difference is the tsunami was caused by an earthquake, and the storms are just caused by the wind.

But they're the same physical phenomenon. They're just excited by different means. And so if you want to imagine what a storm surge is like, you can think of a picture like this, but think of it occurring in the middle of a typhoon, or a hurricane. That's why it kills so many people.

So it's a big deal to try to cope with surges. Now, how do we actually go about assessing the risk? Well, maybe a few people in the audience are lucky enough or unlucky enough, depending on your point of view, to own property near a coastline somewhere. And if you do, you are paying a property insurance rate which reflects the insurance companies, to a less extent than it should, but reflects the insurance company's analysis of risk.

And what is that based on? It's based almost entirely on the history of events in that particular region, not unreasonably, except when you actually look at the numbers. So if you look at hurricane damage in the United States since the middle of the 19th century, and Roger Pielke, Jr., at Colorado State has done a lot of work on this. It turns out that more than half of the damage is caused by, has been caused by eight events in our history, more than half the damage by only eight storms.

Now, you don't have to be an advanced statistician to understand that eight is not a very big number to work with when you're assessing risk, and that's for the whole U.S. More than 90% of the damage has been caused by these high-intensity, high-category storms, not that many of them. They're only 13% of the total. All the rest are weak storms.

And there have only been 30 of these high-category events since 1870. So the obvious conclusion from that is just the statistics of landfalling hurricanes aren't really up to the task of assessing hurricane risk, they just aren't. History hasn't been long enough, and in many ways, the historical record isn't good enough to do this. So, your property insurance rates and the City of New York's assessment of their risk for hurricanes, the City of Miami and so forth, they're all based on a really small number of statistics.

And we can do better, I would argue for that. So, what we do at MIT is we try to use physics to augment that. And what we do is we... The key to doing that is to, first of all, recognize that global climate models, or climate data sets, are too coarse really to have strong hurricanes in them.

So we embed in those climatological data sets or the climate models, really detailed models of hurricanes. But how do we decide when and where to embed them? So we invented a way of doing that. What we do is, is we go down to your local farmers market, and we buy some packages of hurricane seeds.

And we take those hurricane seeds, and we toss them into the climate state and watch what happens. That's pretty much what we do. We toss them randomly in space and time. The next task is once they're embedded in the climate, where do they go?

That turns out to be fairly easy, because hurricanes just move with the large-scale air flow in which they're embedded. So if the wind's blowing from east to west, the hurricanes will move toward the west. It's not that much more complicated than that. And we know what the large scale flow is, it's part of the climate state.

So we find out where they're gonna go. That's sort of the easy part. I put in a few details for the atmospheric scientists, there. And then the key thing is to run this very detailed, but simple, hurricane intensify model along each of the tracks of those seeds.

And I'm not gonna take the time to tell you all about that model. But, it's very simple, you can run it, a track in about 10 seconds on an ordinary laptop. So it's not expensive, but it's pretty accurate. And when you do that, you find that about 98% of the seeds you put down just die.

You put them down in the wrong environment, and they just go away. So, you throw them away. That's about my success rate as an amateur gardener, by the way, it's also about 2%. I don't know if that's a coincidence or not.

It's a process of random seeding and natural selection to use the biological analogy. Survival of the fittest. All the things that were put down in a bad environment go away. A few that have the right things going for them amplify, and then we regard that, the survivors, as constituting kind of the climatology of that.

Well, what we can do this way, very easily, is to generate for a given climate 10 or a 100,000 hurricanes very easily. And there you can start to do statistics. Now of course, this is only any good if this technique works. And so how do you know that it works?

There you really do need to compare to historical hurricanes. And, here, for example, is a bunch of tracks of these what we call "synthetic" storms that have been generated off the rendition of the current climate called the ERA40, color-coded by intensity. That's sort of an eyeball thing is it sort of, kind of, what we expect, but it's not nearly good enough. To be quantitative, you really have to subject it to a lot of tests, and we have over many years.

And it actually does surprisingly well. And what you can do with that is you can start to assess risk in specific places. So here, for example, dear to my heart is a risk analysis for the coast of New England and it, from hurricanes, we actually do have hurricanes in New England once in a while, not many. This is a measure of how strong the storm is when it crossed the coastline, and this is called the "return period." It's just a measure of how frequently those events are likely to occur.

So every thousand years, on average, you might have a storm that's 120 knots when it goes across the coastline. Not very often, in other words. The green dots are what you get just from history, from historical events. The blue dots are what you get from a much larger set of these artificial tracks.

And the red curve, you can forget about. It's just sort of a curve fit to the blue dots. We do lots and lots of tests like this to see how well it does against historical data. It also produces rain.

So we can begin to effect to assess freshwater flood risk. This is the accumulated rainfall of a storm that moved up along this black track. So lots of rain clearly affected by the mountains like the White Mountains of New Hampshire, by the coastline, and so forth. And we can couple these events to models of the storm surge.

And I'm gonna show you the results of some study we did for the City of New York some years ago, before Sandy, in which we tried to assess the risk of storm surges at the Battery, the southern tip of Manhattan where there happens to have been a tide gauge since 1923. And we run those hydrogenetic models. There's one that we run on a kind of regular mesh that you see here, a computational mesh. Another, called Ad Circ, we run on an irregular mesh that's, you see with these points out here.

So it's a fairly elaborate procedure, but we have been able to do it. This is largely the work of my former post-doc, Ning Lin, who's now at Princeton. And this was the result of what our assessment of surge risk in the current climate at Manhattan. So, this is, again, the return period.

This is the magnitude of the surge, the peak surge. And this includes the effect of astronomical tide, this black curve. Now Sandy's surge was about 2.8 meters, and that puts you at a return period of a few hundred years, maybe 500 years. So we would have said that Sandy's surge was a 1 in 500 year event.

Other groups, using completely different techniques have estimated that Sandy was a 1 in 700 year event. So we were witness to, fortunately we think, a very rare event, And we published this before Sandy. We have to be careful what we publish. If it comes true, we're in trouble.

Now, this technique you cannot just use it in the current climate. You can apply this seeding technique to global climate models and look at the future. We've done this for a bunch of different climate models that won't mean anything to you, but five different climate models, both for the historical period and for a future projection. If you look at projections with this technique globally this is hurricane power over the next hundred years.

This is a hindcast, basically this green part. It seems to go up according to this method. This is the results from all of the models it used, all five. And the scatter among the models is shown by this sort of shading in the background.

So a lot of scatter among the models, but a general upward trend. And if we couple that, to go back to New York City, couple those to surge models and allow for the fact that sea levels expected to increase by about a meter at the end of the century, we can get an assessment of risk. So this is four different climate models. This is the result of generating these synthetic tracks.

This is for the current climate, and this is for the climate toward the end of the century, of this century, with the one meter rise in sea level. And you can see, for this particular French model, the storm that was a 10,000-year event in the current climate becomes a storm that might occur every 200 years in the future climate. A storm that was every 20 years becomes every year or every 10 years. Sorry, every 200 years becomes every 10 years.

And then similar results for other models. Of course, the models are different, climate models. There's a lot of uncertainty because we're not all that good at climate modeling. And that just shows where Sandy occurs on this.

We have to add this after the fact, because we publish these projections also before Sandy, it turns out. So the Sandy event in the present climate would be every 4 or 500 hundred years. According to this model, it would happen every 40 years in the future climate. So that's the kind of information that, say you're a city planner in New York and you're thinking about whether to build a storm barrier or something, that's what you need to make decisions.

Of course, you have to do the rest of the problem how much it will cost to build barriers? How much will that save? Do the cost-benefit analysis. So, let me wrap things up.

On the sort of climate front, in general, the weight of existing evidence supports the conclusion that unmitigated climate warming does present significant risks to future generations. Nobody I talked to in my field doubts this. The scientific uncertainty about the nature and magnitude entails a low, but not tiny, risk of catastrophic outcome. This is something we do in our everyday life.

We bet the farm on low probability events. You pay more money, substantially more money for your cars to have air bags in them, which in all likelihood you will never use, I hope. But when they are there, they'll save your life. We put a lot of insurance and a lot of risk at stake at the sort of tail-end of the risk.

And there is a tail, there's a dangerous tail in projections of global warming, that we need to be careful of. We don't just bet all of our farm on the most probable outcome. We take into account the less probable, but much more dangerous risks, as well. Among the myriad risks, and I've only talked about one here tonight, are changes in extreme events like hurricanes.

There's a pretty strong consensus emerging in the community of people who work, scientists who work on this, that the frequency of high intensity events should go up. There's some evidence that it is. And that would produce a greater number of storms like Haiyan that exceed empirical tolerance levels, and let me explain that rather awkward phrase. Almost every culture, modern culture, particularly, every society is built to withstand roughly a generational event, say a hundred years.

It's in our experience base. We know about it, we plan for it. You start getting events that go beyond that experience base not very far, then you start getting a lot of destruction. This is the story of, this is the true story the press won't tell you about Haiyan, in the Philippines.

The Philippines get hammered practically every year by Category 5 hurricanes. They have a much worse hurricane problem than we do, but you don't read about it because by and large they weather those storms fine. Haiyan was a little bit above what they're used to, and that's what happened. A lot of the people who died in Haiyan were in evacuation shelters that could and did take your ordinary Cat 5 typhoon.

You put it up a little bit, things start to fail. It's a threshold phenomenon, practically. That's why the change is so dangerous. You know, the wind speed may go up 5 percent.

That might not sound like much, but it's a lot because it pushes you over these thresholds. That's the end of my talk. Thank you. (applause)