 dry in that decade. How can we understand this sort of variability in climate, the drought index in this particular region? What we do as climate scientists is try to identify relationships. So for example, if we consider this red and blue curve here, this is the so-called AMV index, which is a measure of how warm or cold the North Atlantic is at the surface. And you can see by eye that when the North Atlantic is warm, in other words, when it's red, there tends to be drier conditions in the Southwestern United States. And when it's colder than normal, there tends to be more precipitation or moisture in the Southwestern United States. So we say they're correlated. There's a relationship there. And we can look at that relationship also spatially when AMV is positive, this whole region tends to dry out. And here's another index that's also relevant. And this is the Nino 3-4. It's related to ENSO. It measures how warm or cold the tropical Pacific is. And you might also be able to see by eye that this black curve shows some wiggles that are in common with the green curve. So these are also correlated. And we can look at that spatially. And we can say that when the tropical Pacific is warmer than usual, the Southwestern United States tends to be drier than usual. So the point I'm making here is that as climate scientists, we're looking for these types of relationships to understand climate variability. And a key point is that climate over land, over places like Colorado, is intimately intertwined with what's going on elsewhere in the world. In particular, what the oceans are doing. So when the North Atlantic SST is warm, that impacts Colorado. When tropical Pacific SST is cold, that impacts Colorado. And we're trying to discern what are the key relationships in the Earth's system that can help us understand regional climate change. And for decadal time scales, a key player is the ocean. The ocean is this slow-moving fluid. When it changes from decade to decade, that is a key driver of changes in climate that last for decades. So what I just showed you there was an example of trying to understand regional climate change. But when we think about the bigger picture, the globe, the planet, and we think about what causes climate change on this huge scale, the planetary scale, then we're more interested in these forced components. So, for example, the concentration of atmospheric CO2 has been measured to rise at this observatory in Hawaii. This shows how it's risen since roughly in 1960 to present, clearly going up. But there's other stuff that's going on. There's volcanic eruptions that happen periodically that eject aerosols into the atmosphere, some big ones recently, Pinatubo and El Chichon. We've also got sulfur dioxide in the atmosphere that has risen over time with industry, plants spewing sulfur dioxide into the atmosphere. And so we see this increase in concentration in the atmosphere. Both these and all of these contribute to what's happening on a planetary scale, which is this, the global average surface temperature is rising. And this is the data that we have from 1850 to 2020. You can see it's been pretty steady up until the early 20th century and then a very dramatic rise since then. And so the key takeaway here is that changes in atmosphere composition, gases and aerosols in the atmosphere are important drivers of climate change on a global scale. Now a key point in this talk is that when we think about these different drivers of climate change, some of these are predictable in advance. So we can predict the greenhouse gas concentration, or at least attempt to. We can't predict volcanic eruptions, but once they occur, that gives us some means to predict the future. And we can predict solar cycles. And it turns out that we can also predict these key modes of variability in the ocean, like El Nino, La Nina, and changes in the Atlantic Meridian overturning circulation. We can predict those. And so those gives us some means to predict climate change. There's also things that we can't predict. And so there's going to be limits to how far into the future we can predict things because we can't predict weather beyond two weeks. And we can't predict these chaotic fluctuations in the ocean much more than a month in advance. And we call that unpredictable noise. So how do we tackle this problem? We use global climate models, which basically encapsulate all of our scientific understanding of how the atmosphere, ocean, sea, ice, land work in terms of physical equations. And we solve those equations on a grid that is very blocky. If you have sons or grandsons that play Minecraft, you'll know what a blocky world looks like in a computer simulation. This is similar, but we're doing it to simulate the entire earth system, the oceans, the land, and the atmosphere above it. And so we run these computer simulations, these virtual earths, to understand past and future climate change. And these are very complicated computer programs that we have to run on very large supercomputers like the one that we have in Carlyamang Supercomputer Center. So a picture is worth a thousand words. I want to show you an animation of what a computer simulation of earth looks like. What you're going to see is output from our climate model, CCSM 4, and what you're going to see is monthly sea surface temperature and monthly sea ice concentration in this video. So here it goes, and you can see the date up above. So these colors out here in the ocean are the ocean surface temperature, and this white stuff is how much sea ice there is over the Arctic. The red line shows what the observations were. So you can see that our model is doing a pretty good job of reproducing what was actually observed with satellites. You can see the seasonal cycle. In summertime the sea ice retreats, and in wintertime it expands. And if you notice as we get closer out in time, that extent of summer sea ice is reducing, because we are warming the earth and the sea ice extent is reducing. Here we're seeing the seasonal cycle of sea surface temperatures, warm during summer and cold during winter. And we're progressing down here to the southern hemisphere, where we can see what sea ice looks like around Antarctica. And you can see that our computer simulation is doing a pretty good job of getting the seasonal cycle of advance and retreat of Antarctic sea ice. So this is just a simulation. It's only showing two fields, and in reality when we run these models there's an atmosphere that has clouds and precipitation and wind and all of that. And I'm not showing you that, because when we do things with a model we can dissect it and we can only look at what we want to look at. So to study past and future climate change we run these model simulations with observed and projected changes in gas and aerosol concentrations, like I was mentioning. Here's the time series of atmospheric CO2 that's been rising up to around 2020. This has been observed. We're not making this up. And then when we go into the future, then we do have to make things up. We have to come up with different scenarios of how humans will act. Will they continue business as usual and just keep pumping things into the atmosphere? Or will they make changes that will make this curve taper off? And so we simulate this range of futures. And when we do that we get something that looks like this. We run our global climate model and what we see is that it's showing a very stable global temperature up to about 1950 and then this sharp rise in temperature up to 2100. And if you look closely you'll see a little red curve here that's the observation. So the observations have been showing this upward trend. And our model is doing an excellent job of tracking what was actually observed that gives us confidence that this future projection is realistic. Now an amazing thing that we can do with computer simulations of the Earth's system is we can run multiple Earths at once. So for example this gray curve is actually 30 different model simulations that were run in parallel. So we basically run our model over and over again many times. And this is what we call a large ensemble. These large ensemble simulations allow us to study the difference between forced and internal variability. Those two drivers of climate change that I was talking about earlier. If we zoom in here at 1920 where these gray curves start this is what it would actually look like. We started them all from the same point. And then they all started to do their own thing and diverge and they have different El Nino, La Nina cycles and this is what we call internal variability. And that's a natural artifact of having a chaotic system. We can tweak the initial conditions just slightly and then these different Earths follow different paths. And in our minds as climate scientists we treat the Earth, the observed Earth as just one of many Earths that could have occurred. So there's one of many possible realities, alternate realities is what we do at NCAR. And then we can do this amazing thing which is we can use our model to separate forced and internal variability. Forced variability is the variability that all of these different Earths share in common. So they all show this rise in global warming. We call that the forced change in climate. And then the spread across them is what we call internal variability. That's because they're all doing their own different thing. And so the key takeaway here is that when we project future global surface temperature the forced component that is coming from the increase in greenhouse gas emissions is much larger than the internal component. So that's on the global scale. What does all that mean regionally for people living in Colorado for example? Well here's what we can do with the model. We can run all these different versions of Earth and ask what was the 50-year temperature trend in wintertime over this period from 1963 to 2012. And what you can see is that these different virtual Earths give different answers. That's because they have different internal variabilities. So for example some simulations show that it cooled over this time period over Colorado. Others show that it warmed. What's the right answer? We don't know. We know what happened in observation. That's down here. But what we can conclude is that internal variability probably contributed a lot to what was observed over this time period. It wasn't just external forcing. There was a lot of other stuff going on. If we do the same thing into the future, now 50 years, 2010 to 2060, what's the winter temperature trend? Now these all look a lot more similar. They're all red or orange. They're all warming. We don't know what the observations are because we haven't observed it yet. But we can conclude based on this experiment that forced variability from anthropogenic CO2 emissions will dominate future temperature trends in the United States. So if you're kids and grandkids, it's going to be forced variability that's going to dominate climate over the continental U.S. in the 21st century. Here's kind of a schematic to think about that. In terms of spatial scale and time scale, which matters more for predicting climate? Internal variability or forced climate change? The answer is it depends on what scale you're talking about. Now we live down here. We live in cities and we go about our lives from day to day. What we mostly notice is weather. It's sunny today. It's rainy tomorrow. That's what we notice. That's internal variability of the climate system. We don't experience forced variability as humans because that occurs on long time scales, centuries, on large spatial scales, global. This is where forced variability dominates. But as we move into the future, this region of red is going to start to expand. It's going to start to be noticeable to humans who live, say, for a decade or two in the same place. You might have already noticed that October and Colorado is warmer than it used to be back in the day because you've lived in this city for several decades. Anthropogenic forced climate change is no longer a far off academic concern. It's starting to become noticeable even to humans who live in a very small spatial scale. This is the key. Future generations will be most affected because we're here on this curve right now and kids and grandkids are going to be living out here. What will that look like? That's the question that we're concerned about. But for decayal climate prediction we have a more focused concern, which is what's going to happen on this kind of time scale from years to a decade into the future, kind of from states out to global scale. We're interested in this part of the curve and so we need to take into consideration both forced change as well as internal variability. Before we get into decayal prediction I want to address this common question. If we can't predict weather beyond two weeks then how can we see anything about what will happen in the next few years to decades? So let's answer that question. The answer is this. We're not trying to predict weather ten years in advance. We're trying to predict climate ten years in advance. Climate is the statistics of weather. So you can think about it like a pinball game. You drop the ball in it bounces around and it's going to land in one of these slots. That's a weather event. That's one day's weather. Now if you drop that ball many times then you'll start to get an idea of what the statistics of climate are where you live. This is what's most common most likely to happen. This might be an extremely cold day. It can happen sometimes but not very often. That's what climate is. That statistical distribution of weather events. And the point I was trying to make before about the oceans is that changes in the ocean like this fluctuation in the surface temperature of the North Atlantic, cold, warm, cold that can tilt the climate game in a particular region so that you're more likely to get warm, dry events than before. And that's what we're trying to predict. These changes in the multi-year statistics of weather. Now greenhouse gas emissions are also tilting the climate game almost they are going to break the climate game. They're going to change the climate statistics beyond what we've observed to date into some realm that we don't know what it will look like. That's what we're trying to ascertain with our climate simulations. And so as climate scientists we're the frog on the left. We are in a warming environment but it's not pressing. Today was a beautiful day. But this is the conundrum. This change is happening on time scales that are hard for humans to notice and hard to motivate reaction to. And so we're going to end up like these frogs unless we do something. The water is starting to boil. So now I'm going to get into a little bit of technical stuff. I want to share with you what we're actually doing at NCAR. How can we predict climate in the next decade? This is what we do. We have some estimate of what the earth was like back in time. That's the black curve. We get that from observations. And then we're going to take our model and we're going to start it from those observations and we're going to run these ensembles, these multiple virtual earths and let them run forward in time. And we're going to do that many times in the past. And then what we do is we say how well did these red trajectories from our simulation track what actually happened? Did they do a good job of predicting the past? That's called a hind cast trying to predict the past. If they did a good job of predicting the past then that gives us confidence that these future forecasts that we can't verify are reliable. We can trust them. And that initialization that step of starting the model from what was actually observed, this helps us to predict the internal variability. And we run these and we force them with observed external forcings and that helps us predict the forced variability and change. We're getting both of these drivers of regional climate represented in our model. So here's kind of a summary of our climate prediction systems at NCAR. This uses the CESM model community earth system model. We have some predictions that are uninitialized. We don't start them from observations. We just start them and run them for a long time and then we impose greenhouse gas emissions and they start to warm. We call that the CESM large ensemble and the skill comes from correctly simulating the forced variability, the increase in greenhouse gases mainly. And we expect skill for long time scales and large spatial scales. This is how we know that the planet is going to warm out to 2100. And then we have another kind of prediction that's called an initialized prediction. Here the skill comes from simulating both the forced and internal components as I just described. And for this kind of system we expect better skill for short time scales and small spatial scales. And you might say well why don't you just do this all the time then if it's better. And the answer is because this is much more expensive. The experiment that I'm about to show required 26,000 simulation years of our virtual earth and that takes a long time to run even on a supercomputer. So here's a result. This is from our decal prediction system and what I'm showing you is our ability to predict ocean surface temperature. Red means we're doing well, blue means we're not doing well. And this shows how far in advance we can predict it. Seven years before, nine years before. You can see we can predict ocean surface temperature very well, nine years in advance. Over most of the globe, not everywhere. And a lot of that skill is because the ocean is warming. It's warming because we've got more CO2 in the atmosphere and we're able to predict that. But how much better is the initialized than the uninitialized? That's shown here. And what you see in this bottom row is that the red is really in the North Atlantic, maybe some parts of the Southern Ocean. And so synchronizing the model with observed reality helps us in some regions, not all regions, but in particular in the North Atlantic. So we think there's some potential here to make even better predictions of what's going to happen in 10 years. Now you might say, who cares about predicting ocean surface temperature? No one cares about that. Well, as I've mentioned, ocean surface temperatures impact things you do care about. For example, if you're living in Africa, in the Sahel region, you care about rainfall. And this is our skill of predicting 5-year average rainfall over this region. You can see it's quite high, it's quite red. And this is how much better we do when we initialize. Much better. Because we're able to predict what the ocean is doing and that influences how well we can predict what's going to happen in the African Sahel. So for example, here's the observed time series in black of precipitation over that Sahel region showing this mid-80s drought in the Sahel, negative values of rainfall. And the red shows how well we're able to do with our prediction system. And you can see how well the red tracks the black. That means we're able to predict this roughly 7 years in advance. So if we had developed this system back in the 70s, we could have predicted the Sahel drought that caused the famine in Africa. We would have known it was coming. Alright, I'm going to hand it over to Isla soon and we're going to have a quick audience participation segment here. How much do you think the planet is projected to warm in the next 50 years? Raise your hand if you think it's going to stay stable. How many people think it's going to warm just as much as it did in the last 50 years? How many people think it will warm twice or more than in the last 50 years? Everyone. Everyone's correct. This is the answer, at least according to our model. So from 1971 to 2020, the observed change in global temperature was one degree Celsius. 1970 here we are at 2022, but that was a one degree warming of the planet. In the next 50 years, based on our models, we're expecting a two degree warming. So the rate of warming is increasing. I'm going to hand it over to Isla and she'll give you some examples of regional climate predictions that we're making. Alright, thanks. It's a pleasure to be here. So I'm going to follow up on what Steve has just gone through and give you a couple of kind of real world examples of using the model and the simulations that Steve described to actually come up with some predictions for the forthcoming decade or decade. So I'm going to give you one example for European winter precipitation where the internal variability is really the important thing. And then we'll go through another example a little closer to home looking at the hydro climate of this region for the next two or three decades where really the forcing, the increase in greenhouse gases is going to be the big player. So we'll start with the first of these, which is looking at internal variability and the role of that ocean variability that Steve described is predictable on precipitation for western Europe. So the focus here is going to be on the North Atlantic jet stream. So the jet stream is kind of the strong westerly winds that encircle the planet and I'm just showing you a measure of the jet stream here. This is the strength, the average strength of the winds from the west to east direction at about two and a half kilometers above the surface and so we have these strong winds that are blowing from west to east here and they're kind of tilted toward the UK. They kind of guide storms across the Atlantic. The jet stream is very strongly coupled to the storm track activity so storms blow across the Atlantic. In addition you've got these kind of westerly winds that are flowing in towards the UK and they're bringing moisture from the ocean so really the jet stream here is one of the big reasons why the UK is a very rainy place to be. But it turns out that this jet stream has varied a lot on kind of decadal time scales over the last few decades particularly in March so we don't know why March but it seems like March in particular there's been a lot of kind of decadal variability and I'm just going to show you here this is a measure of the magnitude of that variability in the jet stream's strength basically what I've done here is just taken the strength of the winds from the west to east, the kind of average speed, taken 10 year chunks over the historical record and then calculated the variability across those 10 year chunks. And so you see you've kind of got like three kind of hot spots of variability one in particular here is to the west of the UK so if I just take an average of the speed of the westerly winds to the west of the UK here and look at that going back to 250 as best we can so there's a number of different lines here you don't need to worry about that they're just different kinds of observational data sets the important point is that they kind of overlap where they agree where they overlap so we kind of trust the observations for this and so this is how the strength of the west really wins in this region has varied in time and so you've got all these up and down fluctuations from year to year but you've also got these lower frequency fluctuations where for example over the course of like four decades the strength of the winds has increased from the 50s up to the 80s and 90s here and this has impacts on the weather and the climate in western Europe so I'm just going to show you two 31 year averages so this is the structure of our jet stream in the historical record here for the period centered on 1995 and then here for the period centered on 1950 and so you can see that the structure of our jet stream is really different between these two 30 year periods so in this one you had pretty strong westerly winds to the west of the UK but in this one the jet stream is kind of going more directly across the Atlantic and it's intersecting the continent more kind of at the Iberian peninsula and this has an impact on the precipitation of western Europe so here's the precipitation averaged over that 30 year period and centered on 95 and this kind of fits at least with my view of western Europe which is that Scotland is very rainy and Portugal is not as rainy as Scotland but if you go back to this 30 year period centered on 1950 you have a pretty different picture the precipitation for this 30 year chunk is actually similar in Scotland as it is in the Iberian peninsula and that's because in this time period our jet stream was kind of going directly across the Atlantic it was bringing the storms and the moisture more towards Portugal whereas in this period it's bringing the storms and the moisture more towards Scotland so I grew up in kind of this purple blob in the 80s and it's very surprising to me that there could have been a 30 year period where Portugal had just kind of the same amount of precipitation as Scotland that's not really the way we think of the climate of western Europe but that's what happened and we have this low frequency variability in our climate system now it turns out that this low frequency variability in the jet stream is very closely linked to the Atlantic sea surface temperature variability that Steve just described so here I'm showing you in red is our measure of the jet stream strength to the west of the UK that I already showed you the average in this box here and then in black is the measure of the sea surface temperatures in this box in the north Atlantic and I multiplied by minus one because they're kind of anti correlated with one another but what you see is like kind of if you look at yearly values you've got all sorts of ups and downs whether the sea surface temperatures in the jet stream are not really connected to one another but if you go to lower frequencies so here I've just taken 20 year averages and run them through the record to get rid of all the year to year fluctuations you can see that these two things go along very nicely together so in the 80s and 90s you had cold sea surface temperatures in the north Atlantic and your jet stream was strongly tilted and it was bringing all the moisture and storms towards Scotland but back in the 1950s the sea surface temperatures in the north Atlantic were warmer and you had less precipitation in Scotland there's some caveats here of course just showing that these two things are correlated does not mean that one is causing the other but we have some further evidence that I'm not going to go into here to argue that it really is the sea surface temperatures driving the jet stream variability that's kind of the driver of it all and of course we also only have a short observational record we have maybe three or four wiggles that we can really see these two things coming along together and that's all we have but we'll see as we move into the future how robust is this connection really but that's what we have if we go back to kind of 1850 so what this connection represents is that kind of when it gets warm in the sea surface temperatures in this subpolar region in the north Atlantic you tend to have the jet stream in the north Atlantic kind of displaced to the south and that makes it drier in the UK and it makes it wetter in the Iberian peninsula now as Steve mentioned the sea surface temperatures in this region can be predictable a decade at least ahead of time and so this presents with the potential of being able to predict for the next decade how is precipitation going to look in western Europe but our models are currently they're failing to capture this connection but we can predict the sea surface temperature so even though our models are failing at some aspect and that's kind of a current topic of research because we can predict the sea surface temperatures and we can see this connection and observations between the sea surface temperatures and precipitation we can have a go at trying to predict the future for western Europe so here's our just to show you that the and Steve kind of already went through this but to show you that the model really can predict these kind of low frequency wubbles sea surface temperatures in the north Atlantic the black line here is what actually happened in the observations and then the red line is our prediction and so you can see that we can predict kind of the main wubbles that happen so we can predict for the 10 years ahead of time that we would have had colder sea surface temperatures in the 80s and 90s when it was wet in Scotland and warmer sea surface temperatures that so even though we we kind of can't capture all of these connections in our model the fact that we can predict the sea surface temperatures in this region means maybe we can kind of bypass this issue and we can kind of try and use this connection between the sea surface temperatures and precipitation to predict the future so I'm going to show you our attempts at predictions for two regions the western UK and then the Iberian Peninsula so on the top you're seeing precipitation for the UK and on the bottom is precipitation for Portugal here these are 10 year averages over the course of the observational record and on the right you're going to see some measures of skill of our predictions you don't need to worry too much about the details but numbers closer to one are better higher skill so here's what our model does it's uninitialized so when we just give it the external forcing so increasing greenhouse gases and aerosols and volcanoes like Steve went through the model doesn't really predict any of these low frequency fluctuations this is what we get when we initialize it we start the model from observations and try to predict the decade ahead and still that doesn't really give us anything because the model's failing to capture the connection between the sea surface temperatures and precipitation but instead if we kind of use our prediction for sea surface temperature and then what we know from observations of how sea surface temperature connects to precipitation then this is what we can predict for the forthcoming decade in our simulations and now we have some skills so we could have predicted starting here that it was going to get wetter in Scotland for the kind of 80s and 90s and then it was going to get drier after that so I can ask well what's going to happen for the next decade actually this is you know a couple of years ago that we did this now so here we're predicting what's going to happen for the decade average from 2018 to 2020 and this is what our predictions suggest so they suggest that the UK will be a little wetter than it was in the early 2000s but not as wet as in the 80s and 90s and Portugal will be a little drier than it was in the early 2000s but not as dry as it was in the 80s and 90s so in sort of six years time we're going to take a look and see whether our predictions were correct and whether this connection has carried on so now I want to finish up with an example of our predictions for here for this region and this is going to be predictions based on what external force things are going to do so what rising greenhouse gases are going to do so I'm going to place this in the context of what happened in 2020 so 2020 was a very extreme year for the hydro climate of the US Southwest now we had a lot of fire activity this is the Cowwood fire just along the road there of course we've had more destructive fires since then in 2021 and 2022 but 2020 from a hydro climate perspective was the most extreme year and so I want to kind of go through what role did climate change play in what we saw in 2020 and what should we expect as we move forward and for this we're going to use these uninitialized earth system model predictions and as Steve mentioned as we move into the future we need to make some choices about what's the likelihood of what our behavior is going to be and we account for that by using different scenarios I'm going to show you the results from the most extreme pessimistic scenario where we keep on emitting our greenhouse gases at a similar rate but actually it doesn't make much difference over the time scales that I'm going to show you so we're going to look at annual averages averaged over the 6th state region of the US Southwest and here I'm showing you time series of adaptation on the left and near surface air temperature on the right these are anomalies from the 1950 to 2000 average and so what you see is that 2020 was very dry it was actually the driest year on record even if you go back to the 1890s and it was also very hot it wasn't the hottest but it was up there in the hottest years on record now we can look at our uninitialized model simulations and so I'm showing you here all of these grey lines are all different model simulations like Steve described you get large ensembles where you run multiple earths all at once and you can see they're all kind of wobbling about with their own internal variability that's not predictable and it's not aligned with what we've seen in observations and we don't expect it to be because some aspects are not predictable at all but we can average over all of these grey lines and pick out what's the kind of common signal among them what's being forced by the forcings that we're giving the model the rising greenhouse gases in particular and that's what you see in the black line so our models don't suggest kind of historically that we have had a forest trend in precipitation over the US Southwest but we have a very clear trend in near surface air temperature that's very aligned with observations and you can even pick out these kind of little cold blips here these are volcanic eruptions so Benetubo in 91 and Alchichon in the 80s you can see their impact on the average in the temperature in the US Southwest so I want to use these model simulations to kind of give you some perspective of the year 2020 from using two variables precipitation and then vapor pressure deficit which might not be quite so familiar so it's a pretty important quantity it's basically a measure of the atmosphere's thirst for water it's the difference between how much water vapor the atmosphere is capable of holding and how much it actually has and how much is capable of holding that depends on temperature so we might expect to see a pretty important force change in this quantity and the reason we care about it is that if you have a higher vapor pressure deficit you'd expect more evaporation so you'd expect more drying out of the vegetation and the soils and there are very clear links between this quantity and wildfire like burned area over the US Southwest now these two things interact with one another if you have a year with high precipitation you'd probably have low vapor pressure deficit because it's going to be a cooler year and there'll be more water vapor in the atmosphere both can be affected by the internal variability which may or may not be predictable and both may also be affected by anthropogenic forcing so just to finish up here I'm going to kind of run you through this diagram which is going to show horizontally here on the horizontal axis precipitation and on the vertical axis this vpdu vapor pressure deficit quantity which is very closely linked with wildfire now these are going to all be anomalies compared to the 1950 to 2000 average and all you see at the moment are observed years again averaged over the US Southwest now these are color coded according to the years so we've got the the 1950s in the dark blue going up to the 1990s in the whitish blue so you see the link between these two quantities like I mentioned if you have a higher precipitation you probably have lower vapor pressure deficit and vice versa now here's an estimate of kind of the probability of a given year sitting at a given location in this space based on our models so we've taken thousands of years of simulation and calculated what's the probability of having a year with high vpd high precipitation low vpd low precipitation basically the darker colors means it's more likely so you're very likely to be kind of near zero anomalies you're somewhat likely to have high precipitation be accompanied by low vapor pressure deficit but you're really unlikely to have high precipitation and high vapor pressure deficit what you see is that the observations they sit very nicely within our model estimates of what the climate should look like but this is all for 1950 to 2000 and this is what has happened in the last 90 years in our observations we've gone into the yellows and the oranges here and we've clearly shifted out of the 1950 to 2000 climate this is an example where we can see climate change within our lifetime and we've shifted to higher vapor pressure deficits and then 2020 is very extreme kind of out all on its own so we can use our models to kind of place 2020 within context and say well what are the chances that we could have a year with as low precipitation as 2020 or with as high vapor pressure deficit as 2020 and if we do that based on our model simulations of 1950 to 2000 it's kind of like a one in 50 year kind of event that you'd have precipitation as low as 2020 which is kind of consistent with the fact we've seen it once but if we ask that same question for the vapor pressure deficit it just doesn't happen in the 1950 to 2000 climate but of course we've had climate change and so we can instead say well what do our models look like now so here's our model simulations now centered on 2020 and you can see that this distribution has shifted to higher vapor pressure deficits very aligned with what we've seen in observations now if we ask that same question what are the chances of having a year with as low precipitation as 2020 and it hasn't really changed but for vapor pressure deficit it's become possible it's still very extreme it's kind of a one in 200 year kind of event but it didn't happen in the climate of 1950 to 2000 and it's possible now and then we can ask well what's going to happen in the next few decades so here's our predictions for the models and so this is for 2030 to 2050 so our climate has kept on shifting up to higher vapor pressure deficits and now if we ask what's the chances of having a year like 2020 there's a slight increase for precipitation but it's not very systematic but for vapor pressure deficit it's increased a lot it's become kind of a one in eight year kind of event and then of course you wonder well what is an extreme in this future climate look like given the way 2020 was so our model suggests there's been kind of no systematic change in the likelihood of having a low precipitation year like what we saw in 2020 but for this vapor pressure deficit which is very closely linked to wildfire anthropogenic climate change has basically made a year like 2020 possible although it's still very extreme and it's our model suggests that we're going to see that happening much more in the future in a 200 year kind of event now they suggest that in the future it'd be more like a one in eight year kind of event so here's kind of the take home messages from both mine and Steve's parts of the top so there are really two factors that contribute to changes in regional climate from decade to decade we have internal variability some of that's predictable some of it's not and then we have forest variability and change that we can simulate with our models to some extent so these kind of initialized decadal predictions they show promise where you can kind of start from observations and you can ask well what's going to happen in the next decade and there the North Atlantic Ocean is a really big player and I showed one example of where that might be useful a case where you could try and predict precipitation in western Europe and work is kind of ongoing to find other sources of predictability on these decadal timescales and then we have our forest climate projections where we're predicting what a greenhouse gas is going to do and how's that going to impact the climate and that's becoming an increasingly big part of our kind of forthcoming climate but there are uncertainties one is the four things we don't know what we're going to do one is the model response not all models respond the same way over the south west I mean they pretty much all predict that it's going to turn but they don't necessarily agree on how much and then of course there's going to be internal variability as well that's unpredictable so over the course of the next decade or two that internal variability could change things and lead to things that we don't necessarily expect but I think here in the south west we have some confidence that it's going to get warmer which will very likely lead to more eridification more dried and more wildfires but it can make a difference what we do so like Steve mentioned before we use different scenarios when we're predicting the future and what I showed you was the most pessimistic scenario here I'm showing you how vapor pressure deficit is expected to change for two of the other scenarios the yellow one and the blue one on the time scales that I was talking about it actually doesn't make much difference because we're pretty much committed to warming this much kind of out for the next two or three decades but it makes a really big difference by 2100 what we do so this change here is much bigger than what I showed you out to 2050 but if we reduce our emissions and we follow this blue scenario would be halving the increase in vapor pressure deficit that would be seen by children and children so that's all we have I guess we'll leave it there and happy to take any questions any questions I had a quick question on your whole gesturing with precipitation and that whole section I can't even remember for some reason when you talk about the warming of the sea like do we know like at what temperature like water is absorbed up into the atmosphere and like at what temperature the greatest absorption is up into the atmosphere and then like does that seem like it would be consistent with the gesturing shipping that moisture as long as the air temperature is not too hot to evaporate that moisture in the air to get it to a certain place like I would imagine you guys probably thought of every last scenario and I would imagine that would be in a scenario in the computer as well but does that have any correlation yeah so actually why the jet stream varies with the sea surface temperatures in the way that it does is not clear at all because it's consistent we'd expect stronger westerly winds to the south of colder temperatures which is what we see but it's not as basic as that because our models don't do it and there are people who would argue against me and say the ocean doesn't influence the atmosphere in these latitudes in that way because our models don't do it but there are various things we might not be representing correctly in our models one is like Steve mentioned these kind of fine scale ocean features are small enough scale that we don't capture them necessarily correctly plus their impact on the atmosphere might not be represented correctly either so we're doing one thing at the moment which is so the simulations I showed you here had a one degree longitude and one degree latitude grid boxes and we're running a simulation now where we over the North Atlantic we refine that to one eighth degree to try and capture these high resolution processes so I think it's not at all obvious exactly why the jet stream is shifting in the way it is with the sea surface temperatures because something as complex as our Earth system model is not capable of capturing that entirely I'm not sure if that answers your question totally Steve do you want to add anything? Maybe 12 years ago I did my first several trips to Ireland and never going to the west side of Ireland for the first time and seeing a palm tree that reflects on your data and how they've had that period of time in which would support that type of vegetation with what's going on and I'll just generalize here with the Arctic ice melting the Greenland ice sheet melting of course we read about these things that I'm going to assume some of that at least is dumping into the North Atlantic do we have yet a scenario whether or not how significant that is or will be and cold water tends to go down right but it also tends to do we know that that's factoring into your surface temperature model predictions now or is that something still to be determined or studied more? That's a great question I mean Greenland ice sheet melt Antarctic ice sheet melt those are huge outstanding questions that still remain difficult to model because you have to model you know land fast ice and all the dynamics of melting and motion of ice on land so we still are kind of on the cusp of getting those processes represented in the model but I would say our current understanding is that Greenland ice sheet melt will have an influence on say ocean circulation in the Atlantic it will probably tend to slow it down and when you slow down the ocean circulation in the Atlantic you bring less warm water up into high latitudes that's part of why Ireland is habitable is because it's being warmed by the ocean so that slow down is going to tend to offset global warming at least for Western Europe you're going to have a slow down of the ocean not compensating the global warming of surface temperature and so it's one reason why we have in our simulations we have a warming hole up in the North Atlantic is because the ocean circulation is weakening there. I was curious whether your models take into consideration methane emissions and concentrations that kind of thing knowing that much more of a contributor than CO2 yeah we do have methane in the model it's prescribed so we tell the model how much methane there has been and like we do with CO2 we come up with scenarios of what we think methane is going to do in the future but it means that kind of these tipping points or feedbacks that people talk about melting of permafrost and that impact on methane emissions that's not fully interactive in our model at this point that's something that people are researching how to represent that so it's there to some extent but we may be missing some other emissions yeah thanks for the nice talk. Regarding the simulations what typical ensemble sizes do you use and how do you determine that yeah there are papers written about that what is big enough ensemble size it kind of depends on what you're looking at if you want to start looking at the tails of the distribution like extreme heat waves then you need a bigger ensemble size and if you would just want to know like what's the annual mean temperature change going to be I guess our ensemble sizes have grown as our computing capabilities have our previous ensemble of our kind of regular climate projections was 40 and our latest one is 100 say for most things 15 is kind of okay except if you want to start looking at extremes that's a great question I wish I knew the answer to it I would say more is always better like we always want more ensembles and at the end of the day it's a pragmatic decision how many can we afford to run and how much data can we afford to store that tends to answer the question for us with his last question and with your response being said have you done like a system model to where if all of the ice caps were to melt or if all the ice were to melt like what that scenario would be I would imagine you have done that one kind of curiosity so our current climate model doesn't have dynamic ice sheets so that's a new capability that's going to be in soon that would allow that kind of simulation to be done I think people have done idealized simulations they call them hosing experiments where you just dump a load of fresh water in the North Atlantic to mimic what would happen and look at the ocean circulation and things but it's only going to be in our next generations where we're going to be able to simulate things like sea level rise our current ocean doesn't but it will soon and then have ice sheets that are actually like Antarctic ice sheet that can actually melt or a Greenland ice sheet that can actually melt thank you my impression from the media is that the ice sheets are melting more quickly than expected so my question is is that correct and if indeed that is correct what does that say about what's left out of the models? Yeah so neither Isla nor I is an ice sheet expert but I don't know if it's true that the consensus is that observed rates are faster than expected that may be the case I'm not sure it certainly is the case as Isla said that we're not representing these physical processes in our current models this is kind of cutting edge stuff that's going to go into next generation models I think the processes that you have to represent are so complicated that's one of the reasons why they're not in right now because they involve very fine scale interactions between the ocean and the sea ice where the land ice kind of is against the bedrock and then projects out over into the ocean so you've got these ocean currents that come in and melt the ice shelf from below this is a very fine scale process that's very difficult to represent in a model and so one way forward is to concentrate resolution in those areas that's a topic of research but when you do that your model gets a lot more expensive and we're talking about long time scale simulation so it's a very challenging question to answer but one that many people are working on right now so in the first part of the presentation you talked about how the advanced the models are now you would be able to predict something like the the drought in Zahal right so is that have you all seen anything like that currently or is that something that you have to focus on a certain area or run models on certain areas in the globe to be able to see something like that yeah it's a good question I mean our models are always run globally so we always run the whole planet at least other people run more regional models but ours are all done globally yeah and this is the health thing like Steve said the model didn't exist back then so that wasn't possible but the southwest drought for example that's going right now we ran simulations where we could see what is the role of we had a La Nina state that year we're still having another La Nina probably and so we were able to run simulations where we could see like what was the role of La Nina in the drought it was a slight player but it was really kind of unpredictable at least yeah unpredictable atmospheric variability on top of this long term warming trend that was predictable by the from the forcing so that's an example of where you know we can see have an event right now and try and understand it using our simulations yeah there's other things going on with like Australian wildfires there was the big Australian wildfire year in late 2019 and there are some indications from our model if we give it the emissions from those fires that it increased the likelihood of having a La Nina so yeah we definitely when big things happen we kind of look and see what their impacts are I'm not sure if Steve's ever noticed anything in our like projections for the next decade that are kind of unusual or no I mean it's a good question like would we have the foresight to kind of know this is a significant anomaly that could be coming down the pike I'd say you know we like our CESM model and we think it's a good model compared to other models but we have to go through a pretty rigorous process to identify you know where are we confident that the model is correct and so on the maps that I showed you you know there's some places where we're doing well and other places where we're not so what we'd have to do is we'd have to have a prediction of a future event that was in a region that we had confidence that the model was able to simulate correctly before we would publicly announce kind of a prediction for the decade that people ought to take note of and at this point in time you know Eila showed a prediction for Western Europe that you know that's a prediction but it's basically a prediction back to normal conditions which isn't a very exciting prediction but yeah stay tuned for future predictions of big events Any more questions? Oh we have one last one okay. So this is kind of a two part question so you just kind of mentioned that the CESM is a pretty good model relative to the other climate models that you've been looking at but is there kind of a consensus within the community of what the best performing climate model actually is and then the second part of that question is so you know using those different forcing scenarios what is typically the spread of the temperature predictions for each of those scenarios? So I think for that yeah what's the best model really depends on what you're looking at in some cases there are some models that are just bad at everything but like ours is generally quite good it ranks highly on many things but there are some things that we don't do right like our westerly winds in the southern hemisphere are quite a bit too strong kind of on the extreme end of other models are better in that regard. Another thing is the UK Met Office has these kind of remarkable predictions for the North Atlantic Oscillation so they start in November and they're able to predict it very skillfully for the average from December to February of that year and no one really I don't know it's not clear how they do that our model doesn't do that so yeah it depends on what you're looking at whereas our model has an excellent prediction of North Atlantic sea surface temperatures do you want to add anything more on that first part? Well you know it's a great question all models are flawed but some are useful right that's always true and certainly true of CESM as Isla said it ranks highly internationally but models there's certainly recognition that there's always room for improvement and the main thrust for improving the model is to increase the resolution throw more compute power at it and we want a high def climate model not a low def climate model that's where things are headed but to do that we really need quantum computers we're reaching the limits of what silicon chips can deliver in terms of these kinds of simulations they're extremely expensive so that's the future for advancing climate modeling is getting to where you can actually resolve convection clouds which we can't do in current models and as I mentioned we can't resolve ocean weather which we think is important so there's definitely going to be kind of a paradigm shift in the quality of our simulations when we can get down to that really high definition and I forget the other part of the question yeah the second part was yeah the spread and temperatures and you can kind of see it on this I mean the units are not temperature but basically it would look pretty similar so this is between this one and this one so the spread is not very big over the next kind of out to 2050 we're already committed to some warming but after that it can get pretty big I feel like we're talking about the difference between maybe 5 Kelvin or 5 Celsius global average temperature rise with the high emission scenario to what 1.5 or 2 with the low depends on the model too like some models warm a lot more than others I really feel like I want to ask a question so I'm going to ask it do you feel like we're just headed towards catastrophe are you I'm wondering if you're sort of bummed out by your work at all you know what I mean or are you hopeful or are you just so focused on these models and these scenarios that you know who cares about optimism or pessimism for me it's changed a bit over time depending on what I was working on I don't think we're heading to catastrophe it's like what I want to look at the signal is too small compared to the noise and then I get frustrated that climate change isn't big enough but then I started working on hydroclimate in the US southwest and there it's not a small signal at all and it's a little frightening so it's seeing the fires that have happened here over the last three years and knowing that that's going to be more normal in the future and not really being able to comprehend how an extreme year in the future would look like in terms of wildfire I think is pretty frightening for us locally here so I guess I'm not catastrophe but I definitely think there are things we need to worry about hesitate whether I should say this but you know scientists are pretty dispassionate and to some extent this is like a very interesting experiment going on you know what's going to happen this is interesting that's part of my own personal reaction but then you know on the human side yeah it's incredibly scary and what we're doing is we're basically shifting responsibility to next generation to solve this problem and I'm an optimist humans are innovative and we've come up with ways to either engineer ourselves out of this problem or you know take the measures that are necessary but we're kind of we're trusting that our kids and grandkids are going to be able to solve this problem that's the truth let's thank NCAR and Dr. Simpson and Dr. Jaeger for their important work and thank you all for joining us