How is Climate Change Affecting The Weather Now? - Myles Allen

Well, thank you very much everybody, and welcome to the ninth of our Gresham lectures on net zero. Although this lecture will be a bit of a departure from most of, well, all of the lectures today, which have been focused on the challenge of how do we stop global warming? Uh, how do we stop the rise therefore in essentially one number, the global average surface temperature. Um, but of course that one number is, as I emphasized in the last lecturer, hard enough even to measure, nevermind feel or experience. And what really matters to people in terms of impacts of climate change is what's happening to our weather. So in this lecture, we'll be talking about the two ways in which climate affects weather. So there's mull over and say, what are the, why are the two ways we'll talk about this? Um, so building on that, we'll talk about how we set about quantifying the role of climate change, uh, in extreme weather events in particular.'cause it tends to be the extreme weather events that do the damage. Um, and why it is that getting this right is really starting to matter. The Gresham College lecture that you're listening to right now is giving you knowledge and insight from one of the world's leading academic experts, making it takes a lot of time. But because we want to encourage a love of learning, we think it's well worth it. We never make you pay for lectures, although donations are needed, or we ask in return is this, send a link to this lecture to someone you think would benefit. And if you haven't already, click the follow or subscribe button from wherever you are listening right now. Now, let's get back to the lecture. I mean, at the end of the day, weather, like this storm in lanch, um, in the west of Ireland, um, is, uh, it matters much more than global temperature. I mean, this is, this is what actually affects people's lives and livelihoods. But we have always had extreme weather. And this is a photograph I took, um, just down the coast, um, from Lynch, um, showing that same winter of storms that destroyed this pretty old wall here. So, you know, it was a one in a hundred year perhaps event as we assessed at the time, but as you can tell, there's some infrastructure there that had no problem with the storms been around for a bit longer than that. And that gives you a sense of the, the perspective we have to take that, you know, you, you get a one in a hundred year storm, but it's not a one in a thousand year storm. And, and, and that's the kind of thing we'll be talking about in this lecture. And the reason it matters, um, is less, one thing I do get teased about is our obsession with flooding in the Thames Valley. You're gonna hear a lot about flooding in the Thames Valley, um, in this, in this lecture. And, uh, for those of you who live in the Thames Valley, I'm sure you'll find this very interesting. Um, but I am occasionally, um, teased by colleagues who point out that, you know, the floods that actually matter are probably not the ones that happen in in ox one, but you know, that's, uh, that, that, that's been a focus of, of our, of our attention. But so, so this is, um, some, uh, very, very serious floods that happened in Bangladesh a couple of years ago. And it's this kind of event that was being much discussed at the recent COP 28 meeting in Dubai, where countries have agreed to contribute to a fund to pay for loss and damage due to climate change. Um, which of course raises the question of how do we tell what is loss and damage that's due to climate change? And that's already becoming a contentious issue. You know, how, how are you, how are you gonna decide what, what events, uh, should get payouts from this fund? But we're gonna talk about, you know, as I said, more parochial level, um, OX one. Um, and this is really what prompted this event, prompted my interest, uh, in this whole, um, topic, uh, back in, uh, 2003. So just over 21 years ago, um, I was looking out the back door of our house at the flood water of the Thames getting steadily closer. And this was the scene, uh, at the neighbor's house. And then just, just actually Dave Frame, who I mentioned, uh, in the, in the, um, trillionth ton lecture who I've worked with extensively. Um, that, that, that's his front door. Just, just, just there. Um, fortunately he was renting the house, but, um, you know, it's, uh, flooding is very traumatic for people. Um, we were pretty concerned at the time because we couldn't get sandbags for love, no money, and the water was getting closer. We were very lucky we didn't actually end up getting flooding flooded, but, um, plenty of neighbor neighbors did. And at the time, the sort of mantra, um, of the, uh, climate science community was, oh, well, weather's not climate. These are weather events. Uh, what, when you're talking about climate, you're talking about global averages, global average temperature, and so on. These weather events have nothing to do with that. And, and that, you know, the thing we kept on saying is you cannot attribute any individual weather event to climate change. And, and this frustrated me because you had a feeling that, you know, there was something to be done here, um, which prompted this article. Um, and the reason, of course, it was sort of getting, starting to this line, oh, you know, you, you can't associate a particular weather event in the climate change. Uh, we're starting to wear a bit thin because people were saying this is a one in a, you know, we, we had a whole lot of floods in Autumn 2000. We'll come back to those. And at the time everybody said, oh, there's a one in a hundred year event or so, and there was a lot of discussion about, and then three years later, we had the same thing. And so it's, well, okay. Um, and, and that's the point where you start thinking, okay, so, so what, what can we say about the role of climate change in these events? And, and this is kind of what we, uh, what it, what it boiled down to. And I was, this is a, a line, a line in this article I was particularly proud of, um, quoting, uh, ed Lorenz, who I think in turn was quoting Mark Twain. Uh, climate is what you expect. Weather is what you get. So, uh, we'll talk a bit about what we mean by expect and get in a minute. Um, and then the, the flip side of that is, well, climate is what you affect. When we put greenhouse gases in the atmosphere, we affect the climate, but it's weather is what gets you, uh, because it's the weather that actually, um, affects us. So if we talk about sort of what we mean by expect, you know what we mean by expectation, what, we'll come onto that in a minute. But the, the question of attribution of linking weather events to external drivers in some way has been around for some time. And I came across this preparing to this lecture. This was in the 1915 Coursely Journal of the Royal Meteorological Society. And they were talking about another wet winter that occurred in, uh, uh, 1415. Um, and they, this is a quote from the paper, an explanation which has been most readily accepted by the general public Gresham audiences of the time, for example, um, and even found favor among a few meteorologists, is the heavy artillery firing in France. Um, and Flanders is the primary cause of the unusual wetness of the winter. So people have looked for explanations for, um, weather events since time immemorial. And of course, if you go further back, you can find that, um, parole led with ii uh, basically lost the support of the population because of a famine, which was blamed on him for being an evil king, but it was just a famine. Anyway. So let's talk a little, let's talk a little bit about, um, the, the, uh, the, what we mean by climate is what you expect and weather is what you get, um, uh, with the help of a, a dice, very familiar dice. And I'm gonna ask an audience member, if you don't mind, um, helping out here.'cause it's always more convincing if I'm not the one who's throwing it. So if you could, uh, roll the dice for us, uh, somewhere around here. I think they wanted it. And call out what you get. What's a four? A four? Let's try again. Just keep, keep rolling it a few times. You gotta roll it a few times. A three, two. This demo doesn't always work. Okay, let's let, let me try. There's actually, okay, we got a six, we got a six, we got a six<laugh>, We got a six Loaded dice <laugh>. But it didn't work to start with. That's the thing about a loaded dice. And it's just the same with the climate. Yes, heat waves are getting more likely, but it doesn't mean you're gonna get a heat wave every year. Cold winters are getting less likely. Doesn't mean we're not gonna occasionally get a cold winter. So actually the demo kind of works 'cause it's obvious now. It took, but it took about two sixes before you started to think, wait a minute. And then, and by the way, I didn't switch the dice, it was the same dice, uh, in case you're wondering 'cause it, we sort of suddenly, um, but, um, uh, but, but, but thank you very much. You're fired. Um, <laugh>, it is, uh, actually, there's, I'm sure there's a whole PhD to be done on what actually happens when you roll the dice, because it is interesting how, um, once I got the knack of it, I was getting a six every time. But, but you can easily not get a six every time. So anyway, um, the, the, uh, but the point is, so expect we expect to get a six one in six times. That's what we mean by the climate of the dice, uh, of a, of a true dice. Yeah. And we, uh, and what we get is whatever the number turns up. So that's the analogy to have in your mind when you're thinking about climate and weather and climate change works like loading the dice slowly, increasing the loading on the dice. Um, so if I had switched the DI didn't honest, um, in that, that would've been like a climate change. I would've, I would've changed the climate of the dice by adding some loading to make it more likely it came up six. So that's an analogy to, to bear in mind, and it helps us just thinking about a dice helps us think about the whole, uh, challenge of understanding the link between weather and climate. So, um, here's a plot, um, which, uh, again, you could probably do in your head showing, telling you what the probability is in the horizontal of rolling a certain number of sixes. And obviously on the first roll, the probability is one six, if it's a trued dice, this is the red dots are a trued dice. Um, and then if you want to get two sixes in a row, it's, anybody wanna shout out 36 1 in 36? There you go. And then the, the mental arrhythmia gets a little bit trickier now with three roles. It's one in, well, okay, six cubed and so on. Okay? So that's what it looks like. The probabilities go down very rapidly as the number of sixes go up. So as soon as I rolled three sixes, you were all going, oh yeah, it's, it's obviously fixed. I mean, you know, because it would've been a, you know, it, well, I can tell you it would've been to get three sixes in a row would've been, uh, more than one in a hundred. I just, this is the same data, but I'm just plotting it. Um, because you'll see lots of graphs like this through this talk. I'm plotting it as with probability, one in 10, one in a hundred, one in a thousand, one in 10,000. And I'll probably start talking about return times and when in, in weather and, uh, floods and so on, when we talk about the return time of something, we're not really talking about a time at all. It's not the amount of time you have to wait to get the next one, it's the probability of that thing happening this year. Well, one, over the probability of that thing happening, that's what we mean by the return time of event. So the return time of a triple six, um, with just three roles, um, is, uh, you know, one in one 20, you know, if you did it every day, it'd be one in one, six cubed, whatever the, the number is, uh, 200 and something, um, uh, day, uh, days. You know, you, you, you, you'd expect to get that once, once that, that, that, that sort of frequency and so on. Um, now of course, if we just load the dice to make, um, getting a, a six a one in five rather than a one in six chance, so to load the dice a little bit, this particular dice is very heavily loaded. But you know, I suppose imagine sort of fairly subtle loading on the dice, um, that changes the chance of getting a one in a single six from one in six to one in five. But the chances of getting a double six go from one in 36 to one in 25, okay? And the chance of getting a treble six, um, go up even further and so on. So, so what the loading does, it has some impact on, uh, the, the, the, the, the normal events if you like, but a really extreme impact on the more extreme events, or a different impact on the extreme events than the normal events. And we have to understand that. Now this is just a loaded dice. Now it's not, it's not generic to everything, but it tells you that you've gotta understand how this slope, the way things get less, get less probable as they get bigger, which is a feature of many, many things, changes as a result of the loading is something we need to do a lot of research on. And I actually have this, this is, you know, what we've, we've jumped now from climate research you can do with, you know, fluid in pipes and, and, um, a handful of equations to the most sophisticated questions that confront climate resurgence today, which is understanding what is happening to extreme weather. And we, for which we have to use the most sophisticated models we have available, which we all come onto. And I'm actually very happy to say that a couple of people who are actually doing this work are actually here, uh, in the audience. Um, so just to re remind you what return times, uh, look like, and this is a boring process if you like, um, uh, just a completely random process. If I was to roll a dice, um, a true dice lots of times and add up the dots, I end up with something called a Gaussian distribution. A a normally distributed process, the sort of bell curve of probability you expect. And if I, uh, plot here number of standard deviations in the vertical of my bell curve, what's the probability of getting one if I just generate lots of them? And the red dots here are lots of possible, lot, lots of gen, you know, lots of weathers. Weather is what you get lots of rolls of this process, and you can see they go up. It's not this straight line, unlike the dice, it curves a little bit and it it flattens out as you get towards higher and higher. Um, uh, prob uh, lower and lower probabilities, higher and higher return times. The blue is, is the sort of what you get if you have a, like a million, if you try it millions of times so that you get sort of very exact estimate of what the, of what the, uh, uh, return times are. And the red shows you that when you're getting to, you know, a few, a few tries, it starts to jump around a bit because you're in the, you're in the noise at that point, right? Um, but let's think about a more interesting, that's, that's, this is a kind of boring process, one which is very familiar. This, this Gaussian process. And you could use, if you knew, you know, if you knew what you're studying was a Gaussian, you could work out, um, from its behavior down here, you could make a very accurate prediction of what the chances would be of getting a particularly high magnitude event. You could, you could use weight, you know, what you know about the process here to predict the probability of what happens out there. But that's the, the hallmark of a Gaussian process is that you're rolling the dice and each roll is independent of the previous one. But let's think about another process that is also has a lot of randomness in it, but watch carefully. Um, this is sand pouring out of a tube, okay? Onto a disc. And you can see how it, it evolves, it, it climbs up and then it slides down, okay? And now if you, and then you get a big slide, a big, a big landslide, so to speak. And what's different about this process is that one roll, one grain of sand rolling down bumps into another one. And actually it's not like rolling a dice it self, it's self reinforces. So it's, it's what we call self-organized criticality. But I mean, it, it's, it's a, it's a, it's a hallmark of chaotic systems, non-linear systems which feed on themselves and don't. Although actually, interestingly, yeah, this is where the demo goes wrong. And if you watch this video, because you see it wasn't quite in the middle anyway, sorry, I didn't mean to get onto that bit, but it just goes to show how difficult it is to simulate chaos in the laboratory anyway. Um, but, but, but you get the, you get the point about when you have a system that that is, um, almost as it were, self-aware, it's behavior tends to be much more, you tend to get much more, um, uh, large events than you would expect from the statistics of the small events. And a, a para, since we're having these lectures in the city of London, of course a paradigmatic self-aware, but rather random process is the stock market. And that's exactly why you don't see that kind of Gaussian behavior in the stock market. But you actually see, you know, big, um, big swings much more frequently than you would expect if you assume the stock market was just a, and various people nodding wisely, I suspect that's what they do. Um, so, so, so, so that's, that's an example of the kind of process. So we need to think about chaos in thinking about understanding the link between weather and climate. Um, this is the original chaotic attractor proposed by Ed Lorenz back in the 1960s to explain the lack of predictability, the, or the challenge of predicting, um, atmospheric weather and why it was that, that, uh, weather forecasts tend to, um, go wrong, uh, after only a few days. But I'm not gonna talk in, I'm not gonna use this to talk about predictability and prediction I'm gonna use, um, so that, that, that's Ed Lauren, I'm gonna use Ed Lauren's model to talk about the link between weather and climate to illustrate the link between weather and climate. So here's a view down over the Lauren butterfly and the dots are sort of what, what happened in the system in a particular hour of, of a Lauren Day. So it, you can imagine this is, it's a, it's a, it's a model of the atmosphere, but obviously it's very idealized stylized model. And the blue line at the bottom shows you the distribution. So how often you find dots in that value of the x variable of the Loren tractor. So you've got this attract this, this is the, the, the imagine the particle is moving around in space, okay? And then we're just recording where it is every hour, um, and then looking down at it. So we're looking if this is Z in the vertical X in one direction, Y in another, we're looking down at the XY plane and seeing a, a, a map of its behavior and building this distribution. And Tim Palmer suggested this again, I think around the sort of late nineties. Um, here's Tim Palmer admiring a, a, a chaotic pendulum. What happens if you blow on the Loren butterfly, so to speak, if you impose an external driver that, um, that, that pushes the rents model just in particular direction, which is shown by the arrow there. And you can see the behavior, the, the response to this external driver is to make it more likely that it's gonna sit down here in the red area. That, so the red ones are the forced, are the ones with the driver, the blue is the original one. So you can see I've just added the red on here, and you can see that with the forcing, it tends to sit around here. So imagine that's wet conditions, that's dry conditions in, in, in the atmosphere. The atmosphere tends to sort of hang about in the dry regime a bit more. And you can see that, you know, even though the, the, the, the forcing actually is pushing it in, in the positive direction, you actually end up with fewer high, um, high ex events. Fewer, fewer floods if you like. Um, but what we, what is interesting is if we actually look at the impact of the forcing, and this is what I meant by the two ways in which we need to think about climate change affecting weather, one way to think about it is to think about the trajectories of the weather, the evolution of the atmosphere that leads to a high impact event. So imagine a, a hurricane coming, making landfall in the us So there's a whole series of events that happen that, that hurricane's trajectory. And you can ask, well, what does, what's the role of the external driver like climate change in the evolution of that hurricane? And we can think about this in the context of the Loren system and say, take away the, uh, driver. So, so this is, this is where we've got that Tim Palmers, um, blowing on the butterfly, and now we remove it and we look at the same trajectories. And you'll notice they don't go as far in the X direction. Can you see that? The green, the green lines peak, um, slightly less far to the right than the red one. It's quite a subtle signal, but I'm afraid, and we dealing in subtle signals when we're talking about climate change and weather. Um, so, you know, the external driver increases the magnitude of these high X events. Yeah, if you just look at one event and assume it's gonna happen anyway, the external driver might make it bigger. So that's one way in which climate climate can affect well up. But at the same time, if I just color in high ex events just over a certain threshold, you'll see there's more blue ones than red ones. So the external driver is actually reducing the frequency of these high ex events. So these are the two ways in which climate and they needn't be the same. That's the point that the, there's two ways in which climate affects the, um, the, the weather. One is by affecting the actual individual behavior of individual events by making them bigger, making or, or making them smaller in, in, in, in some cases. And, and secondly, by affecting their probability of occurrence. And you need to understand both of these if you're gonna understand the link between climate and weather. One of the, um, things I'm actually quite proud of over the past, uh, uh, 20 years is sort of in doing this work, um, on, uh, understand the LinkedIn climate. Well, obviously it, it's, lots of journalists are very interested in this question. Whenever a weather event happens, they want, you know, they want you to, you know, answer the question. And, and back in the early two thousands, um, I I had a lot of conversations with people where they got quite exasperated with me is, can't you just answer the question? Is it climate change or not? And I would wander off saying rabbiting on about, well, it might be making it bigger. It might be, you know, um, and, and, you know, it might be affecting the probability of the event. And, and, and this, what, what are we, what we've seen over the past 20 years is now the journalists themselves are going off and explaining all this. And I think we've really made, you know, testament to, to the science journalists, um, surviving science journalists, uh, of, of the UK that we've, we've actually made a lot of progress in, in accepting that this is actually just quite complicated, that there isn't a black and white answer. Um, but that doesn't mean, though, there's no answer. It's just, it's an interesting complic slightly complicated answer. Um, so there we go. The external driver in this particular example won't of course happen generally, um, reduces the frequency of high. So, so this is so chaos. Chaos. And the, the fact that, um, events in our atmosphere are self-reinforcing makes life particularly interesting. It means that you can't just use the statistics of boring events, normal events to predict what's gonna happen, uh, at the extreme. And you can also have these two ways in which external drivers affect weather by through the, through their impact, impact on the magnitude and impact on probability. So, um, we proposed this sort of way of thinking about, um, the link between climate and weather in 2003 and the later that year, in fact, we had a chance to, to implement these ideas, um, in a much more, um, prominent event. The heat wave that occurred in the summer of 2003, uh, which was actually, it caught Europe very badly, um, by, by surprise. Um, and you can see it had very serious consequences. Um, notoriously in France there were a lot of deaths. But actually this is a figure from southern Germany where the health system was by no means overwhelmed. But you can see the two weeks of the heat wave, you can see the spike in mortality. These are, these are, uh, daily, daily death rates. Um, that was a flu epidemic earlier in the year. And you can see during the heat wave, um, it you, you was, it was the heat wave was, was killing people at a faster rate than the flu had been, uh, earlier in the year. And as I say, that was actually in southern Germany, which was not the most extreme, uh, temperatures. And so we, um, went into the, into this problem and working with Peter, start at the med office in Dhi Stone, who was a postdoc in Oxford at the time. Um, interestingly funded by the welcome trust, um, on an epidemiology, um, grant. So we, we were using the point of the, the point of this research as we started, we, we, we realized we needed to use techniques from epidemiology, the kind of the kind of techniques that we've all got gotten ever so familiar with through Covid, um, about how do we work out what the, what the, uh, role of climate change is in, in effecting the risk of these weather events. And, um, this shows, uh, apologies to the crudeness of the, the, the graph, but you know, that that was then, and this, this is the kind of graphics we we got away with in our papers. Um, and, uh, so, so the red line is, uh, if you just include human influence, um, uh, simulated summer temperatures, um, various simulations, um, so there's four of them plotted here. Green is the average if you don't include, um, human influence. So you can see there's a bit of a, an offset, um, in the climate. Um, but the individual spikes, uh, sort of as big as the offset in the climate. And the, the black line is what actually happened. And you know, we used this to argue that, um, human influence on climate had increased the risk of this event by around a factor of at least, at least a factor of two. Uh, and best estimate, probably more like a factor of 10. Um, but you can see immediately from this figure, um, there was sort of one event that looked like, yes, the model could do this, but the only way we, but you know, it's not that convincing. And oh, by the way, I should mention because Peter would love me to mention it. He's got a great book out. Um, and, uh, so it looked for Hot Air by Peter Stot, the Inside Story of the Battle Against Climate Change Denial. Um, sorry that, that's a complete digression, but, um, uh, if you're watching Peter, you owe me. Okay. Um, and, uh, the model, we, the, the, the, the, the challenge with, um, Scott Stone and Allen, I mean, it was a long time ago, but there were plenty of things wrong with that study. And the most obvious was that the model we were using, we, we, we were simulating summer temperatures to see how, because when you simulate the world, you can take human influence out and see what happens. So it's sort of, you can do experiments like that. Um, but um, but we were simulating the world with a model that we knew couldn't actually do what happened. It couldn't generate a two week heat wave like the one which actually occurred. So we were talking, what I was showing you, there was three month averages.'cause that was the kind of thing this model could actually cope with. So that was the one problem. We also had very small numbers of simulations, so we had to make lots of assumptions about the distributions and so on. And again, you are then using, um, statistics from a small number to try and predict what happens in an extreme event, which again, we've just tried to explain with the sand pile. That's, that's not what you want to, that's not ideally what you want to do with a non-linear system. So the next step, um, was a study actually led by party Paul, um, a grad student in the University of Oxford, um, who suggested, um, in 2000, uh, actually in 2004 when, when, uh, our paper came out, um, he's, he just started his doctorate and he, we were also running at the time the climate prediction net experiment, where we were getting members of the public around the world to run the simulations of the climate and send us the data back to Oxford. And, um, it, that was all that had all got going. And Pardeep suggested, um, that we could, we could apply the, this network of computers around the world to, to roll the weather dice lots and lots of times in order to try and address, in order to try and, um, uh, uh, address this question of, of the role of human influence in extreme weather. Um, I, I, I think it's important to acknowledge 'cause for, for, for grad students to remind us, not always listen to your supervisor, because I actually advised party strongly not to do this because I figured it would be very risky and difficult and slow for a PhD. And, and so it did take him a long time to be fair, but he got funding and so on. He, he was able to do it, and he did get the last laugh after all. Um, so that was his paper, which came out actually in 2011, which if you do the maths, it was a very long PhD, but he, it was a, it was a great result at the end of it. And I'm gonna show you an example, not not actually Pip's original paper, but one, we, we, we worked up a bit, um, more in more detail, um, applying party's approach to the UK winter of 20 13 14. Um, this was an extraordinary winter. It was, it broke the, um, the, the record. It was a record breaking wet winter, you probably remember it. Um, and, uh, here's, we, we actually maintained the world's longest daily weather record in Oxford at the, uh, Radcliffe Observatory. Um, and here's Ian Pol, uh, making history, pouring out the, uh, the, the, the, the, the flask, uh, to, to, to record the, the, the, to record the, the, the record breaking, um, uh, the record breaking, uh, uh, wet, um, rainfall. Um, so, um, and at the time, of course, uh, everybody was talking about the link with climate change, and here's David Cameron, um, trying to look scientific or something. Um, and, uh, and, and, and he came out saying he very much suspected this was something to do with, uh, uh, climate, uh, climate change, um, and, uh, with no particular evidence at the time. Uh, but it did sort of motivate, um, the scientific community to address the problem and say, okay, well, was he right? You know, was there any justification of this? So the, when we're talking about, um, winter weather, um, it helps to think about, um, the North Atlantic as a sort of four-sided dice, if you like. Um, and these curves show pressure, um, in different weather regimes in the North Atlantic. So here's one, the sort of northerly cold blast. Um, that's one weather regime. Um, here's a wet southwesterly, another regime, um, a very chilly easterly, but gentle, but cold winds. And finally nice warm, uh, southerly. So these are the sort of four sides of the dice. And the question we wanted to ask ourselves was, you know, obviously what had happened in uh, 20 13 14 was we had a lot of this one, the dice kept on coming up, um, in, on this side, um, again and again and again. And, and that's what built up the floods. Um, and so what we did was we, we used a model now which we, uh, were convinced ourselves a higher resolution model, even than the one that, um, pip was using, convinced ourselves. The model was able to simulate what happened. So this shows you what actually happened in reality and what happened in one of the wetter members, or the wettest 1% of our large numbers of simulations. So it doesn't happen all the time, but it can do what happened in reality. Um, so we, having convinced ourselves of that, we then rolled the dice with the model. So experiments in silico, if you like, um, simulations of in red the world that might, the world as it was in 20 13, 14, with all the drivers prescribed, um, as, as, as the world as they were, um, in terms of temperature, greenhouse gas concentrations, and the rest of it. And okay, this is back to these return time plots. So this is one over the probability of an event of rainfall. That high is one in a hundred, okay? Prob. So it's the same plot I showed you with the dice, but it's just now simulations, uh, of the weather of that winter. And each dot corresponds to one simulation, which is why it was, uh, tremendously helpful to us to have all these members of the public doing these simulations for us.'cause there's a lot of simulations that need to be done. This is only showing you a small fraction of them. So you can see that the red is simulations of the world as it was in 20, uh, 14, 15. And the blue is simulations of the world as it might've been without human influence. Uh, if you, if you sort of reverted to 19th century greenhouse gas concentrations and changed the surface temperatures of the model to be to an estimate of what the world would've been like, uh, back, um, uh, 150 years ago or, or back in the way it would've been if we hadn't raised greenhouse gas concentrations. And as you can see, there's a, a change in the risk. There's a shift from blue to red. It's not huge around a 40% increase in risk. So this was the bottom line of this study was that, yeah, human influence had increased the risk of this, um, flooding the floods that occurred in 20 13 14 by a fairly modest amount. Um, and, uh, so, you know, maybe Cameron got it just right by stroking his chin and looking sort of thoughtful because it's not a huge signal. There are other stronger ones, and it's not always the result you get. Um, this was the autumn 2000 case, which par, uh, did, and he found a doubling of the risk, a hundred percent increase in risk. Um, but interestingly, uh, the following year, um, Allison Kay at the Center for Ecology in hydrology in Wallingford, um, did a study where she considered how climate change had affected the risk of an event that didn't happen. So she asked, well, we didn't have floods in the spring of 2001. Um, how did, how would climate change have affected the possibility of us having floods in the spring of 2001? And what she found was, interestingly, there was a decrease in risk. And why, because the heaviest flood events in Britain in the spring occur because of rapid snow melt. And we don't get that anymore because of climate change. So risks of different events are changing, and they're all, they're, they're, they're all changing because of climate change, but they're changing in different directions. And that's really important if you know, because you still get, you could still get a spring flood, but should you prepare for more spring floods is as a result of climate change. And the answer is, well, you've gotta go and do some work to find out. And we shouldn't just assume because bad weather happens that it, it it's definitely going to become more frequent. Um, you may remember, uh, there was a big argument, um, in 2011, I think, about whether, whether we needed to invest in more snowblowers for Heathrow because of climate change. Um, we didn't. Um, and I think that was the correct answer. Of course, I may be proved wrong by subsequent events, but, but all the evidence is that actually heavy snow events such as we had in 2011, I think it was, are actually getting much less likely, um, because of climate change. So probabilities are changing, and it's really important to understand how, um, and one of the key things that Freddie Otto introduced into this was relating these different ways of quantifying the impact of climate change on an extreme weather event. There's the red lines which show you the, um, in increase in, uh, magnitude in the vertical, or change in probability in the horizontal. And they both suggest that, um, the, if we look at, um, the, uh, if we look at the, the changes we might reach the conclusion this event has been made much more likely because of climate change. So it's mostly because of climate change. But if we also remember what would've happened, um, in a normal year, we very often find that there's a big contribution from chance as well. And so you can have rather sterile arguments about whether an event is mainly due to human influence and whether it's mainly due to, um, uh, whether it's mainly due to, um, luck or, or bad luck. And the answer is always, well, it's a bit of both. Um, and the understanding how these different things contribute, um, is, is the essence of what we do in understanding the link in climate change and weather. Um, which brings us to a, a very different approach, which I just want to acknowledge. Although in the, the length of this lecture, I can't really go into this approach in, in detail, um, which has pioneered by Ted Shepherd at the University of Reading. Um, we're all very obsessed by the attribution problem in Thames Valley. Um, but, um, Ted, uh, considered the, if you remember the, the picture with the, the, the, the Lorenza tractor, the impact of the forcing on an individual trajectory that was already going in a particular direction. What, what difference did climate change make? And, and this is what, uh, Ted calls the storyline approach to attribution where you, you take out or put in the influence of human, you know, you, you, you simulate an event and then you take out or put in the, the role of human influence into that simulation to see what it does to the size of the event. And, and this is an example, uh, which he gave of Hurricane Sandy, um, a with a superstorm Superstorm Sandy, uh, which, uh, clobbered New York a few years ago. And, and this is with observed sea surface temperatures at the time, you can see the intensity of the storm. And this was if you modified the sea surface temperatures to make them more light, sort of what they would've been back in the seventies or so. And, and you can show there's a, you know, the storm is, was intensified by the sea surface temperature. So this is an example of how you can trace out the, the role of a particular driver. Now, did this is not thinking about ties at all. This is saying Superstorm Sandy happened, but what we're asking is how did climate change modify what actually happened when Superstorm Sandy went home? Um, I wanna get onto what we're doing now, um, and, uh, 'cause I, 'cause I, you know, this, this is bringing you right up to date to where, um, if attribution is going and, and, uh, uh, to, because all of those studies that I've talked about up until now have been using models that are essentially making a lot of compromises in their representation, uh, of the weather. And, and they tend to be at their weakest when talking about the most extreme weather events. But there are a class of models out there, they're in use every day that actually simulate extreme weather extremely well. And those are the models that are actually used for weather forecasting. And if I showed you these two, these are a syn, one of these is a synthetic satellite picture, the other one is a real one. And, and I asked you, you know, which is which, you know, you have to, you have to take a, take a moment to think about it. If you, if you're into this sort of thing, you can, you can tell what the, what the issues are with the model, but that's a where satellite observation, and this is the European Centers 18 kilometer ensemble forecasting system. That's the, the model that up until a couple of months ago when they increased the resolution to nine kilometers, the European Center, uh, for medium range weather forecasting in Redding was using to generate its weather forecasts twice a day. And the crucial thing about these models is they are scrutinized relentlessly because if this model fails to simulate an extreme event, well, it's not that somebody gets fired, but you know, there's a, there's a, there's a, there's a SWAT team comes in straight away to say, well, what went wrong? Because, you know, they need, they need to get the forecast right. Um, and if you, if you miss events in the forecast, that's the worst. If you miss an event e even if the event's not particularly predictable, if you, if you fail to say that it might happen, um, then that's a, that's a real, um, you know, well in advance then that, that's a real failure of the forecast. So, um, this is, uh, uh, the nice graphic produced by, by Nick Leach who's here showing and whose work I'm now moving on to talking about, um, showing the contrast between, you know, what you get out of, um, a weather forecast model. This is water vapor swirling around over the North Atlantic and what it would look like in a climate type climate, a climate type model here. So if you're thinking about an extreme event like a storm here, you can see immediately it's gonna be sort of smeared out and, and you're not gonna capture the, the real, uh, statistics that of, of the, of the events that actually do damage. So just to, uh, as an example, um, I'm gonna, um, tell you about the, the first application of this to a, a damaging weather event. Um, uh, Nick did start off by looking at a heat wave in February, which wasn't, didn't, didn't, you know, like a heat wave in the UK in February, in some parts of the world, a heat wave in February does lots of damage, but it was a UK heat wave, um, um, which was interesting meteorologically, but it wasn't particularly interesting to the public. This was an event, um, that was extremely interesting to the public. Um, the heat wave that occurred in the Pacific Northwest in June, uh, 2021 when, and, uh, work with, uh, between Nick and Anja Weiser, um, uh, where we saw, um, an excursion out of the distribution of peak summer temperatures that was just several standard deviations out. And you see something like that, that's the time to start thinking about sand piles because yes, you could have a, a a, a random process like that, which suddenly does something like that, but it's, you, you'd have to be, you know, you'd have to be pretty unlucky. It's sort of, you suddenly, you, you're trundling along and suddenly you roll a, a triple sex. Um, and, and, uh, so, so, um, and he caused, you know, the, the event caused a lot of damage. There was a lot of, um, uh, the tone, tone of litten in British Columbia set the all time Canadian, uh, heat weather record and then burn to the ground the following day, I think. Um, but, uh, uh, it was, um, and, and it was also, you know, as far as we could tell this in, in the observational record, this was a completely unprecedented event. But interestingly, it was extremely well forecast. This is the forecast of the event. Color is temperature. The lines show you pressure. The forecast of the event three days ahead basically nailed it. Um, even the forecast seven days ahead was showing it was gonna be pretty hard. Um, and at 11 days out, again, you're seeing the, the broad, broad, you know, this was a predict, this was a, a well predicted event, even a co almost a couple of weeks ahead. And if we, even if we look, um, two months ahead, this is not forecasting what was gonna happen on that particular day, but just looking at what could happen in that season in a, in a, in a hot day in, in, in, in the, in that season. And you can see it's, it's close to what actually occurred. So we're dealing here with a model that can, unlike many of the other models we're talking about, actually capture the physics of this kind of extreme event. Um, we can convince ourselves even more of that by looking at, say, individual simulations. So that's what happened here on the right. This is one of the simulations in the model, which Nick has carefully picked out.'cause it, it, it is the one which is closest to what actually happened. And you can see it, it, it matches not only in temperature and, uh, pressure, but also this lower plots show you the flow of water vapor in from the atmosphere, uh, from, from the tropical Pacific, uh, which carried at this, this atmospheric river that carried a huge amount of humidity into the Pacific Northwest, where because of this, um, anti-cyclone, it didn't rain, um, but you had very moist air sitting over. So you had moist air, but not raining, which was sort of ideal conditions to build up the, this really intense heat wave. So the experiment we did, or the experiment that Nick and Ancher did, um, uh, with Chris Roberts also at the, uh, European Center, uh, for medium range weather forecasting. Um, you start with the actual forecast. So you're starting with a model, you know, could do the event because it did it forecasted, and then they modified it, subtracting the pattern of warming that we believe has occurred as a result of human infants and climate. And also changing greenhouse gas concentrations to something like 1900 conditions, and then doing the same in the opposite direction and repeat the forecast to see what difference it makes. And here's the difference it makes. So you can see in both the future climate and the pre-industrial climate, we still see a heat wave, but the loading of the dice is pushing us towards a more intense heat wave in the red or less intense heat wave in the blue, or conversely making the odds of a heat wave as large as what actually occurred higher as a result of the loading and finally talking about work done last week. Um, and so, so this is, this is what our, our group does a lot in, in, in, in Oxford, Olivia Vashti who's joining us, um, I haven't mentioned who's funded different, different people here, but I will mention who funds Olivia because that's the man group, uh, who actually based right near to Gresham College, they're, they're funding her work and she's looking at precursors to what actually happened. So looking within, because you've got all of these simulations, I should have, each one of these is a simulation of what could have happened given the starting conditions and then all the modified starting conditions at the beginning of the forecast. So each one of these is a simulation with that complicated synthetic earth that I showed you. Um, and so a huge amount of computing time goes into this sort of thing. Um, and what Olivia's doing is using the fact that we've got all these simulations to ask what are the precursors to the event that happened, and can we link those precursors to human influence? So if you look carefully here, this is pressure, atmospheric pressure, uh, red is high pressure, blue is, um, uh, low pressure and day see day there, day minus eight. That means eight days before the event that happened. And as I approach it, you can see it's about day minus six, a coherent pattern appears in the Pacific. Maybe there's a hint of it on day minus seven, but it gets pretty strong in day minus six. And then as I go forward, minus five, you can see it moves in there strengthens, and there's the high pressure three days before the event when it was really predictable, it was gonna be an intense heat wave and it builds up. And when you get to the heat wave, actually it's the, the, uh, bang on the heat wave. This is, this is asking within that cluster of simulations what tells us it's gonna be a hot one. Yeah. So that's, this is, this is the, the, the, the indicate the indicators within the ensemble. So you can see that's clearly one of the triggers was a, a week beforehand you had this trip pattern appear, blue, red, blue, and it moved in. That was clearly part of it. And another part was water vapor, water vapor transport, this, this carrying water in from, uh, from the Pacific. This is a much noisier picture here. And you can see though, around day minus six do you see this bright, this orange little feature appearing here, which strengthens going straight into the Pacific Northwest and it's really getting strong four days before the event, but by the time the event happens, it's all over. So it's not really, it, it's done, its work, it's loaded the atmosphere with moisture and then the heat wave, and then we had the, the, the spectacular temperatures. So, um, to, to wrap up, um, there's a sort of, uh, there's a, there's an appeal in this talk as well as just providing you with information, because we can do this kind of thing, we can use state of the art weather forecast models to actually answer the question of what is human infant climate doing to extreme weather? But it's not done routinely. This was a huge amount of work for a bunch of grad students in Oxford. Well one, one or two grad students in Oxford to be precise. Um, and of course got done years after the event because that's how long it takes to get these experiments done. Um, you know, we're all being asked to do something about climate change and when somebody asks you to, and you should point out to them, you know, again, I I keep encouraging you to write your mp, one of the things you could ask your MP is why don't, why aren't you doing a better job of telling us what climate change is doing to us? Because we could be, it's, it's only a matter of resources. We could be applying exactly the same models that we use for the weather forecast to the question of quantifying, changing extreme weather risk. But we, we need to be doing this every day. We shouldn't just be doing it. You know, when, when one event happens in one part of the world and somebody happens to get interested. Um, so, you know, we need to understand extreme and understanding. Extreme weather actually is, I believe the strongest case we have for developing virtual earths for developing, um, computer models of our planet, which are, which behave in a way that's actually indistinguishable from our, our real weather. And we are close to that. We, we have the tools to do this. It's not a, an an invitation to an open-ended research program. It's a matter of putting the, putting, putting pressure on, on, on politicians to foot the bill for institutions like the med office, like the European Center for medium wage rather forecasting to start doing this on a routine basis so that we don't end up sort of speculating and arguing over these things. There's just an answer that comes out when a, when a big bad event happens within a few weeks, we should just hear from our forecasting centers, yep, human influence made that event less likely or maybe made that event more likely. And here's the number that that information could be available to you, but it's not. And then I think you should change. So wrapping up, um, we can quantify the impact of climate on weather by thinking about loading the weather dice. Um, attribution of the most extreme events requires realistic models. If you try and rely on, you know, crude simulation averaging up from some, from simpler systems because of the sand pile effect, you, you, you can easily get the wrong answer. Um, and I believe this should be made part of the job of weather forecasting centers to make this happen. Thank you, Miles. Fantastic. That was absolutely fascinating. Scary but fascinating. I'm sure we've got heaps of questions, so we're gonna jump straight in if that's okay. Now I've got a question from Paul who wants to know, we are talking about how climate change is affecting weather. Is there, are there any reasonable studies on weather modification that could aid the negative impacts of climate change? We're gonna be talking about that in the final lecture of this series, so I'm gonna pump that one Sneaky promo there. Okay? Okay. Alright then let's try this one from Andrew. Um, which is the greater insurance probability mega cities flooding or global food systems going to drought? Oh, well, now you've gotta ask that question for other people who work in this spa in, in this part of the world who work in the insurance space. Um, but both of them are, I mean, it, it it's depends on the mega city, depends on the event. Both of them are affected by climate change and they're the kind of things that we need to know how those probabilities are evolving and at the moment, because we need to know how to adapt to them and how to, how to, how to deal with changing risks in the future. And at the moment, in my view, our planners are being underserved by the climate research community. I mean, you know, I'm, I'm, I'm, I'm, I'm fessing up here that because when we do these climate simulations, we're using models that just can't do these events. So what we should be doing is, and we've got bet much better models available'cause we use them for forecasting. So, you know, we should be, we should be applying the best tools we have available to address this problem. And, you know, yes, it would cost more money 'cause it costs computing time and so on. But, um, you know, compared to the cost of preparing flood defenses or, you know, preparing for a global food crisis, um, it's trivial. So we, we need to understand what's happening and we need to make the investment to do so. And that touches on the question I was going to ask you, how prepared are we as a society, as a culture, as an economy for living with a state of sustained unpredictability? And is the answer to that just to do more science? So I've not talked about predictability in this lecture, and that would be a whole other lecture about whether climate change is making the weather less predictable. That's a, you hear that a lot, but actually the evidence for it is, is tougher. That's, that's work for the future. Um, so, but um, we are, we are having to cope with designing things to deal with different climates. So if you're building a building today, it's gonna be dealing with a, a, a, a range of weather risks in the early days of its life that will be different from the range of weather risks that that building will experience at the end of its life. So that makes things more expensive. It makes it harder for planners to do it. Um, and it makes it really important for people to understand how these risks are changing, which is what this science is all about. A simple question. Most of my questions are very simple. Um, you're talking about the need for global coordination, uh, so we can predict better what the future will be and whether it is due to climate change or natural phenomena. Is there good communication? Will a wa or do meteorologists work in isolation? Um, that, that's actually a really good question. Uh, at the moment we have, um, so climate research, um, is remarkably nationalistic. So we have, we have national climate simulation programs that produce a climate model that they've put together at the end. And, and that's a good thing in one sense 'cause it means you have a sort of variety of models and so on. So you get to sort of explore uncertainties in that way. But to do this kind of thing, um, that's probably not the best way of doing it. You're much better off pooling resources to get the best model you, you can. Um, and, and then all, because of course these models are global, you can't run a model of just one corner of the world. I mean, so, so you, you may as well pool resources and do it. You might aren on just one. But, you know, uh, uh, pooling resources and having just a, a, a handful of, uh, of, of these, uh, forecast models is the way to, which is the way the forecasting community has gone. I mean, if you'd like, you can sort of almost tell that where the forecasting really matters because it costs money when they get it wrong. So we've actually had a big winnowing out of forecasting, of, of forecasting centers over the past 20 years. So we now have really only a handful of of countries maintain, um, full global ensemble forecasting capabilities and other countries just buy in. So, so, and, and I think we need to move in that direction, or, or rather, one, one option is that the forecasting community just adopts this and just applies their approach, which is much more, you know, in, you know, global, uh, in approach, uh, than is, than, than the way we're approaching climate issues.'cause climate, climate research has come out of universities, which is all, it's all quite parochial and so on. And, and whereas this sort of stuff, it, it's, it's gotta be approached in the same way that we approach the weather forecast. Hi there. I asked the question about the insurance risks that are predictable. Yeah. Um, but really my question is, can you use AI to enable a valuation of halting deforestation and increasing forestation? Yes. So, um, there, there's a, there's an enormous potential role of AI in this because as Nick, uh, uh, is, is dealing with at the moment to, to tearing his hair out with the terabytes of data you generate from these simulations. Because, you know, imagine we're, we're simulating the whole world at 18 kilometer resolution. We're doing it twice a week and we're doing 50 me, you know, there's 50 50 member ensembles. You know, it all builds up to ridiculous amounts of data very rapidly. And I think, you know, we all recognize that actually analyzing this properly will will require ai and that would allow us then, if you've got a realistic in silico model, so in a computer-based model of the climate, then of course you can ask the kind of questions you are asking and you can then play around with it. You can deforest it and see what happens and see what happens to weather risk in other parts of the world. Um, and, and, but you'll need to do multiple runs in order to do that.'cause you know, one, one run is never gonna, you know, because of the butterfly effect, it's never gonna tell you everything you need to know. So that's where I think the first application of AI to this is just dealing with the big data problem. It's more a comment than a question.'cause I, I live in Miami and we have hurricanes and they all have on the pa on the news. They always have all the, the, the, the models. And I just wanna say that the, the European model is always the one that's most reliable rather than, um, all the other ones.<laugh>, we like to work with the best. I'll pass that on. We Like to, on a positive note there ladies and gentlemen, join me for, to, for thanking Miles Allen. Very, very much. Such a fascinating.