 for coming back from the break. It's my pleasure to invite the next speaker, Samantha Stevenson, who's a professor at the University of California in Santa Barbara. And Samantha is interested in studying climate change impacts on the tropical Pacific and drought in the Southwest US using a combination of ocean and climate models, field observations and paleo climate proxy. And the list of people she acknowledges reflects those different groups. And I'm looking very forward to your talks, Samantha. Great, thanks for the introduction. And yeah, for the invitation. Excited to talk to you all, even if it's virtual. So yeah, I'm going to present some work today from a paper that's currently in review at PNIS. We'll see what happens with that. But it is in collaboration with some people both at the university and at NCAR, continuing collaboration with NCAR. So yeah, I'm going to talk about how seasonal precipitation extremes respond to climate change, what that means in a situation like we're in right now where the background climate is actually continuously changing and how that affects how we interpret what should be considered an extreme in a continuously changing baseline case. So I don't probably need a huge amount of motivation for why this is important, but just a couple of recent examples of how precipitation or hydroclimate extremes, I should say more broadly, have really significant societal impacts. So just a couple of quick photos. So in California, where I am, so I think about a lot, we've had several recent extreme precipitation events that have had significant impacts. This one is from the Orville Dam in Northern California, where an extreme rainfall event caused a overflow of the dam spillway, which almost led to catastrophic flooding in the communities downstream. Luckily that they were able to mitigate the worst impacts before that happened, but flooding is always an issue here in California. More recently in Germany, I may have seen in the news there's been some substantial flooding events throughout Germany and other parts of Europe. Here's just a picture from, I think this is North Rhine with failure, or just some failure of riverbeds and significant impacts on people's houses and property. And on the dry side too, so it's not just rainfalls that hydroclimate extremes which have significant impacts also include drought. This has been a factor in various places around the world. Over the past few years, it's been a significant problem in parts of Southern Africa, where a lot of farmers have faced impacts to their livelihoods and impacts survival in certain communities as well. Places like India gets in California as well, but it impacts, sorry, experience impacts from both too much water and too little. So here's just a couple of examples from, I think, within the same few year period where there was significant flooding throughout parts of India, and then there was a monsoon failure a couple of years later. And so we would want to understand both how changes to precipitation and drought extremes might be expected to respond to climate change because you want to be able to plan for both. So the other thing that I wanted to mention is that of course we are not in a stable climate at the moment, we're in one which is continuously changing. And so we are already in a situation where these extremes occur naturally, but as a result of climate change, the background trends are expected to be substantially different as well. So here's just an example of some climate projections of soil moisture both in the surface, like the upper 10 centimeters of the soil layer, and then within the full soil column from some of the more recent CMIP6 projections. So this is a paper that came out last year. And so I guess the point I wanted to make here is that the trends are large and they are not the same everywhere. So regionally, you can see significant trends either towards drying, for example, in the Western US, in Europe, in the Amazon basin, so in Africa, or towards wetting. This is the case in central, the Eastern Africa, and in India. And these trends, they're slightly different depending on the season. I'm not going to focus on the details of this too much, but I just wanted to point out that climate change is significantly altering background trends in ways which have substantial regional expression. The climate change is also anticipated to alter these statistics of future precipitation extremes. So here's just one example of the global distribution of changes in the, this is the annual maximum precipitation, maximum daily precipitation, I should say. In units of percentage change per degree of global warming. And once again, the tendency is for extremes to become more extreme, but there is some structure to that. So we see significant changes in the equatorial Pacific and to a lesser extent over the majority of the land surface. So there's a background trend towards either drying or wetting. There are changes in the statistics of precipitation extremes. So, and the changes in the spacing of extremes is also projected to be altered, right? So there's one other example from the literature where they, this is from the CESM-1 lens where they computed the change in the frequency of what the lead author calls weather whiplash events. So basically alternating between dry and wet winters in California, the paper was focused on California, but just spatially the relative change in percentage is shown on the left. And then time series of the regional change in the frequency of these whiplash events is shown on the right. So in both Southern and Northern California the frequency of alternating between very dry and very wet winters is increasing substantially, right? So not only are we increasing the absolute frequency of precipitation extremes, but we're also, because of this whole dryness situation there's also an increase in the projected dry extremes which means that there's going to be potential for compounding impacts because you're alternating between dry and wet. So this brings me to my research question. So what I'd like to understand is how is more globally these changes to seasonal precipitation extremes expected to change in the future? So how does that differ from place to place around the world? And particularly what will the effective impact be? So people are pretty good at adapting, right? And so you might anticipate that on a long enough time scale we could be like, okay, well extremes are becoming more extreme but we're better at dealing with them. We can build higher dams, we can make the reservoir capacity different and get used to just managing with less water for a certain amount of time, right? So what will the effective impact be relative to that change in the new normal which is going to alter continuous? And the tool that you need to do this is not just a single model large ensemble but actually multiple large ensembles. And I'm not sure if anyone's talked about this during the workshop, maybe someone's mentioned it already but there's this amazing new archive of large ensembles which is maintained at NCAR but it contains output from large ensembles run with multiple models from multiple model of centers around the world. I believe the last time I looked there was six or seven large ensembles here and that doesn't include the seem of six large ensembles which came online after this but of the model in this archive, four of them had enough information on soil moisture for my purposes. And so these are the ensembles that are used in this study. So we've got the CESM-1, oh, I should change that to CESM-1 and then three other large ensembles from other models that have, they're all run under the RCP 8.5 scenario and have from 30 to 50 ensemble members. So they're pretty big. Okay. And within these ensembles, what I wanted to talk about today is wet and dry extremes. So here I'm using some definitions which are fairly similar to that Swain et al weather-replaced paper I just showed you but here I'm considering a wet extreme to be an event where the 90 day accumulated precipitation exceeds the 99th percentile and the 99th percentile is defined according to the distribution over 1960 to 1990, right? Because that distribution's gonna change. So why this definition? Well, it's approximately analogous to the extreme wet season that happened, say in 1976 to 77 in California that had significant impacts or in 2010 to 2011 in Australia, right? So it's kind of like a seasonal wet extreme. On the dry side, the timescale is somewhat larger because it takes a little bit longer for a lack of precipitation to manifest in terms of the societal impacts. So now I'm considering a three year accumulated precipitation below the first percentile of the distribution again over that same 1960 to 1990 reference period. And so again, this is roughly analogous to historical events. In particular, there was a 2012 to 2016 droughts in California, it might be somewhat weaker than the drought we're now in in California, but anyway, that drought was approximately this intensity and the 2018 to 2020 drought in Southern Africa was also on par with that kind of three year one percentile definition. Okay, so that's what I mean when I say wet and dry extremes for these purposes. If you then look at the changes in wet extremes by that definition, then this is what they look like. So this is an average over those four large ensembles, so this is not just one anymore, all four of them. And the green values are more frequent wet extremes, brown is less frequent, but there's not very much brown because over much of the global land surface, the ensemble mean tendency is for wet extremes to become more frequent. And that change is about a maximum of like 10% or so, might not sound like a lot, but when you consider that's compounded over the course of several decades, then it's something that you'd need to be aware of for planning purposes. All right, so I wanted to look a little bit at the regional structure of these extremes. So I guess I should mention that the region I'm about to show you are from these boxes that are shown on this map here. So the Europe box goes roughly from like Spain, Portugal over to Italy, so basically Western Europe. And when you consider the time series of wet and dry extremes, those are gonna be the top and the bottom boxes respectively, this is what it looks like. And the advantage of looking at this in the time series sense now is that you can look at the behavior of each of the model ensembles individually. So as a function of, oh yeah, I'm doing it over a moving 30 year window. So the function of time, most models project that wet extremes are going to become more frequent. So that's where that kind of 10% increase is coming from. And there is some intermodel diversity, which is interesting. And there's more appearance in other regions actually, but I'll get to that in a minute. But yeah, three out of the four models are pretty consistent saying that we're gonna be increasing from about one event every 30 years to two and a half or so in Europe. But interestingly, the dry extremes are also increasing but the models disagree more strongly on the behavior of those dry extremes. So we've got, what is that, GFDL CM3, I believe is showing a really strong increase in the number of dry extremes. The other models are maybe more conservative or the project to smaller increase. So Europe, for example, would be in maybe one of these whiplash situations potentially where you're gonna have both an increase in wet and dry extremes. Then if you look at other regions, so for some regions, this same thing is true, right? You have an increase in wet extremes and an increase in dry extremes simultaneously. So this is true for the Southwest US, Mexico region, the Australia, which I think I'm using a box that covers the majority of the continent of Australia and the Western Amazon basin. But I will note, again, that there are some differences between models. And in particular, CSI or RO, Mark 3.6, is different over Australia, which is kind of interesting given that that's an Australian model. I don't know if there's any particular reason for that. And in many regions, the CAN ESM also tends to show a tendency towards wetting. But it's also showing regional differences there as well. So yeah, just the point is that we're seeing increases in extremes of both signs, but the magnitude of those increases still appears to be somewhat uncertain, which highlights the importance of using multiple large ensembles to do these kinds of assessments. And then I also wanted to show you a couple of different regions where what I just said isn't true, right? So in particular, South Africa or Southern Africa, I should say, this is not the country of South Africa, is the only region that we examined where the change in wet extremes is actually not an increase, but a decrease. So got a reduction from about 1.2 events every 30 years down to 0.8, but the number of dry extremes is still increased. So maybe that's a situation where one would want to plan more on the be ready for drought end of things as opposed to the worry a lot about floods, right? In India, the opposite seems to be true. You get a large increase in the number of wet extremes. Actually, it's one of the largest increases that we see anywhere. So we go from like below two events to upwards of eight, nearly 10 in some models. That's also true in Eastern Africa. So maybe that increases even larger. Anyway, those two regions are the ones where the magnitude of the increase in wet extremes is the largest. But in India, interestingly, none of the models really projects a large change in the frequency of dry extremes. So again, the management actions that you might take in response to this are going to potentially be different. So I wanted to also mention to the paper that I'm drawing this from talks a lot about the behavior of a multi-decadal drought, such as not to focus so much on that today because it's an S2S workshop, but we do consider the possibility that the long-term trends in soil moisture are related to the statistics of extremes. So here's what the multi-ensemble mean change in the column soil moisture looks like. This is just a difference between the end of the 21st century and the end of the 20th. So this is kind of consistent with the figure from that cook paper that I showed you at the beginning of the talk where we see drying in places like Southern Africa, the Western US in Europe, and wetting in Eastern Africa in India. So the models are telling a somewhat consistent story, but if we kind of flip back through what the extremes are telling us, so these India and Eastern Africa regions, again are the ones where we get this super strong increase in wet extremes. And we also see that this home soil moisture is tending towards wetting as well. So I haven't done a detailed attribution of this, but it seems plausible that this change in precipitation statistics is tied to the overall change in the background trend as well. So I just wanted to highlight that briefly. And then the other thing that I wanted to talk about today that you can do when you start looking at multiple large ensembles is we can think about things relative to the changing baseline. We don't have to be tied to just the absolute behavior of these extremes. So here is what I mean by a changing baseline. This particular example is a time series of surface soil moisture, but the story is gonna be the same no matter what variable you're thinking about. So just do some representative time series. And the black ones are the actual soil moisture time series. I mean, they've been standardized, but that's the total moisture time series which includes the trend. And then the red dashed line is what happens when you remove the background trend. And why is this relevant for ensembles? Well, in order to get a good estimate of the background trend, the most robust way to do that is to build a large ensemble which kind of samples the, hopefully the full extent of internal climate variability and then compute the time-varying ensemble mean. And that's a statistically good approximation of what the real background trend is. So it works a little bit better than just trying to compute a linear trend through your time series. So yeah, so then you can say, all right, well, you know, in this particular case, I was thinking about drought, right? And it's like, well, you're in a drought whenever you're below that dashed line. So in the black line, you're just going to be in a drought for 50, 60 plus years because you've got a background trend. But maybe that's not what you care about. If you're trying to plan on like a 10 or 20 year timeframe, what you really care about is like the variability around that local mean climatology. So by having a large ensemble taking at your background trend, now you have a handy estimate of what the kind of true internal variability is in that particular time series. So you can do this for drought, but you can also do this for precipitation extremes is what I was trying to get at here. So I'm considering the behavior of precipitation after the removal of that background trend to be a representation of the effective impacts of precipitation extremes. So, you know, and again, like if you think about like what that means on the ground, it's like, well, you're managing and you've managed to figure out how to deal with the, you know, but after your new normal, if you want. And then what are the extremes relative to that local mean? If you look at this in the drought context, actually the majority of the land surface is going to have either drought or pluvial conditions. You know, what we would now consider drought or pluvial conditions becoming the new normal. I'm not going to talk about the details of this, but if you do a time of emergence calculation to figure out when the force trend emerges from the noise, you see that either drought in regions where you've got a drying trend or pluvial in regions where you've got a wetting trend emerges from the noise, essentially nowish. It already has in many places and it will very soon in others. And again, this depends on the details of which ensemble you're thinking about, but this particular case is only plotted regions where I think three of the models agree on the sign of change. So anyway, the baseline is going to change. It has changed in fact in many places. So this is not really an academic exercise. I mean, it is, but it's also has real world implications. Right? So if you then, so okay, let's go back to precipitation for a second. So if you now have detrended your precipitation, recompute the extremes by those definitions that I was talking about, this is the behavior that you get, right? So it's a lot of information here. I could not figure out how to get this to be any more clear given the amount of information on the graph, but each of the little sections of the plot is showing you the behavior of detrended extremes in a different region. So we've got those same seven places that I've been talking about for this entire talk. Each of the different colors represents a different ensemble. And we've got sections for the 20th century and for the 21st. So the 20th they're outlined in black, the 21st are outlined in red. And then the bolded ones are the ones where the difference between the 20th and 21st century is significant. All right, so the upshot is that most of the models are, you know, not that consistent in the changes to dry extremes. So, you know, here and there you can say, well, some of the models project an increase in Australia, for example, in South West US and Mexico, right, but there's not a huge amount of agreement. The exceptions are, again, India and Eastern Africa. Right, so this is saying that change in extremes, like the just extreme part, those time series I was showing you before is substantial, but a lot of that is coming from the background trend. But even relative to that new, that background change in climate, in certain places, you're gonna have a significant change in the effect of the extremes and effective impact, particularly in this case in India and Eastern Africa. So these are the dry ones and here are the wet ones. All right, so again, the effective changes in these extremes is very regional, right? So now we start seeing more places where the changes are significant in multiple ensembles. So in Western Europe now, all of the ensemble changes are significant. In the South West US and Mexico, we've got two out of four. And I think in Australia, all the changes are significant. But again, India and Eastern Africa pop out as the places where you see the largest change. So these impacts are not gonna be felt equally everywhere. Okay, so that's all that I had. I hope I haven't gone over time, but because the main conclusions are that's like, the zero order conclusion that I wanted to tell you is that large ensembles are good. They really give you some unique tools that you can do things with that you can't otherwise. In this case, removing the background trend accurately and identifying the role of model structural differences. Precipitation extremes are going to change substantially in both the dry and wet sense and the regional differences are really significant. And then the difference between the results with and without the background trend are substantial, right? So I interpret that to mean that maybe we can adapt to some of these changes. So if we just change our ideas of what is normal. But in some places that might not be good enough because we're seeing changes in the behavior of these extremes even relative to the background trend. So by understanding them in more detail, then we can hopefully make more informed choices about what to do about it. So, great. Thanks for listening. Thank you very much. Yeah, please post questions or raise your hand. I was wondering if you would, I had a question. I was wondering if you could talk a little about the processes that lead to these, to the increase in extreme on. Processes, what do you mean exactly? Oh, which physical conditions? Well, why does it get drier in the different region and why does it get better? Yeah, so I mean, I haven't looked at that in a huge amount of detail for this particular paper. We were trying to characterize the statistics primarily. I think that a lot of it has to do with the reorganization of the general circulation. There's been a lot of work on that, right? So you've got like expansion of the Hadley circulation and so on. So I think like that's what's going on in a lot of the subtropics. Physically, why the precipitation extremes are increasing? I know there's been a large literature on that as well. And I think that it has to do with changes in the convection in the tropics. And, oh God, I'm going to have to go back and recall exactly what is driving the change in the extreme statistics actually. But yeah, that's a good question. No, thank you. I mean, obviously, I wasn't sure how much work you had done in that area. Yeah, thank you. Yeah, next question. Go ahead. Sorry, I have to unmute myself first. Thank you very much for this very interesting talk. It was a great overview. I was wondering more specific question about Europe because I'm from Germany. And I noticed when you showed the map with the wet extremes and the overall increase that you showed for Europe, I also noticed like a difference between more Germany, Belgium region compared to Spain and France. So I was wondering kind of you, the regions you chose, how you chose them, like what your, did you think about what regions chose? What was your motivation to choose certain regions and how do you think the variations within the regions matter? Yeah, that's a really good question. Is this the figure you're talking about? Yeah, yeah, that makes a lot of sense. Yeah, it's a good question. They were somewhat arbitrary, but yeah, I primarily chose them because they coincided with features in, oh, I don't know if I have it, the column soil moisture trend plot. So yeah, it might be worth rethinking whether it makes sense to center different regions on where the changes in extremes are maximized. So yeah, it's a good point. But yeah, they match better onto the trends in soil moisture, is the short answer. Yeah. Thank you. Thank you. Anish had a question? Yeah, thanks, Judith. So Sam, my question was somewhat related to Judith's, but not specifically process related, but more like climate phenomena or like climate mode variability related, right? Like some of these regions, like for instance, Western US, Southwest US, Spain or Southern Europe are impacted by like atmospheric rivers and atmospheric river variability on like seasonal timescale. And then like for instance, India or to some extent, like Western, Eastern Africa and also Australia has impacted the monsoons on seasonal timescales. How much do the changes in these large scale variability and extremes of changes in these large scale climate drivers impact these changes in the extreme precipitation or extreme dry conditions in these regions? Like are the drivers more responsible for this swing in extremes or it's more local impacts, I guess? Yeah, I think that's a really good question. I'm super interested in that. And I haven't had a chance to figure it out in a lot of detail. Yeah, because I think about ENSO a lot, right? And so for example, like the amplitude of ENSO changes in the future is really different across these models. Probably the same thing is true for atmospheric river statistics as well. I haven't looked, but I'd be surprised if that wasn't true. So yeah, I bet that that has a lot to do with it, but and that's like the next thing that like my postdoc is actually starting to think about. There are some suggestions that in the Western US, the occurrence frequency of extremes is a strong function of the magnitude of ENSO. For example, so I think it's gonna be a big deal, but I don't know exactly how much that contributes right now. If I can just a follow on question from that and there's something I will Chapman and I had discussed as well. So in CSM, at least CSM one, there's like periodic ENSO, right? And for the teleconnection pattern is not as real or like not great compared to like the nature of teleconnection. How much does the periodic nature of ENSO influence these periodic swings in the dry and the wet conditions over Western US and CSM? And do you see that less so in like other models which have a more realistic ENSO with a periodic oscillations in ENSO? Yeah, I'm trying to remember what our results looked like. I mean, CSM, like the spectrum peak isn't that bad. It's much too strong in CSM one, but it's not as bad as like it was in say, CSM three or something. But yeah, I think like generally, if you do have a two periodic ENSO, then that's probably going to change your statistics and extremes for the Western US in particular, and so isn't the only game in town. There's a lot of internal atmosphere stuff that happens. So I don't think it's going to, well, I don't know. I think it will affect the statistics, but whether that's the dominant driver or not is probably a little more unclear, but yeah, I suspect that like, yeah, the model behavior of both like the amplitude and the spatial structure of the teleconnections is going to matter quite a bit for the statistics of these extremes. Yeah, okay, thanks. Is that your question or not? It does, yeah. Thank you. Thank you. Kelsey, go ahead. Hi, great talk. I apologize, I can't turn on my camera at the moment. So yeah, great talk. And I have one question about what you think the representation of clouds, what that's like in the different models and how that might affect your results, especially because towards the beginning of your presentation, I think you were showing precipitation trends over the globe, and it seemed to be dependent on maybe representation of strato-cumulus clouds like off of South America. And so I just was curious like what your thoughts are on that. And if you think that it depends on like the representation of clouds in these different models. Yeah, I guess a short, oh, sorry, I'm gonna try to put up the figure. I thought you were talking about, is it this one? No, the next one looks like slide seven maybe. Okay, cool. Yeah, my computer is super slow. Yeah, so I guess again, like I think that it probably matters a lot, but I don't know exactly how. So this is the other thing that I have my little soapbox about, right? Is that we know that there are all these physical differences between models, but exactly what those are is actually very hard to figure out because you have to go into the description paper for each of the models and be like, okay, well this type of aerosol process is here or this type of convective scheme exists, right? But there's not like a place you can go to just look up what type of atmosphere physics, for example, is in different models. And if I'm wrong about that, please tell me because I would read that all the time. But I guess it's not non-trivial to figure that out. So I suspect that's, yeah, like the behavior of strato-cumulus decks in different places or, for example, like the strength of the North American monsoon which is not super well represented in models, like all of those things are going to have a big role to play, but it's kind of hard to decompose exactly which process is responsible for which difference because, you know, you're also changing like what, several hundred things as each generation of model gets updated. So yeah, it's just really hard to know, but I think it matters a lot. Yeah, I'm sure it's hard to isolate that. Definitely. Thank you. That's a really important question, I think. Thank you. So if there's no more, let's, first of all, let's thank Samantha. Thank you so much for this talk. It was nice to sort of have a kind of change component on it and Mike DiFlorio is scheduled as the next speaker. However, he fell ill and is not able to make it. So I suggest we take a 20 minute break for a minute break, yes? Yes. Yeah, I just saw that Jacqueline had a question in the chat. It was just posted. Okay, yeah, so maybe if you stay on for a moment, Samantha and answer Jacqueline's question and then we'll have a break and then we convene at 11.50 which is in 10 minutes, 13 minutes and then Andy will be the next speaker. But before we turn off our cameras we're going to break Jacqueline. Why don't you ask your question to Samantha? All right. Can you hear me? Yeah. Thank you. Your talk was very interesting. I'm not that familiar with soul moisture and like extremes of real land but I was just wondering if the average soul moisture changed that you show from 90, 50 all the way to like the next century. Is it taking into account changes in land use? For instance, if you only look at the Amazon forest deforestation is going to change soil moisture. So I was just wondering if you have thoughts about this and like how can you include these land use changes in your results? Yeah, that's another really good question that's hard to answer. So land use is included in these scenarios. It's part of the RCP, external forcing that's imposed actually. I don't remember exactly how it changed. I think Amazon deforestation is probably in there. So yeah, the way they do it is like they just prescribe like the types of lands allocated to different purposes. Is it crop? Is it urban? Is it forest? What have you? But then how that gets translated into changes in soil moisture depends a lot on the details of the land model, which I'm not, you know, I don't know all the ins and outs of land models, but I do know that it's quite complicated. Like the number of like vertical levels is different. Like the types of plants that are simulated in each model is different and the rooting depths and like all of that. So it is in there and it probably matters a lot. Again, I feel like I'm just saying the same thing to all of these questions. But I think that it's actually really important to maybe try to do things like single forcing experiments where you don't include land use, but you do include greenhouse gas emissions to really try to partition out exactly like what is the land use driven component of that. But right now we just kind of a hard time telling, but. All right, thanks. Yeah, thank you. I think that really shows that it's so important to look at things piece by piece, but then put it together. You have all the non-linear interactions and so that the answer can be quite different or sometimes you understand the answer from breaking it down. Yeah. Totally. Thanks very much. Let's all have a 10 minute break and see you at 11.50. Back. Thanks.