 So, our first speaker is Enrique Kurtchesser, who is a well-known, well-recognized and one of the top numerical modeling experts in the world, and he's going to talk about regional and global ramifications of boundary client upwelling. Let's give a warm welcome to Enrique. Enrique. My honor. Welcome. Thank you for the invitation to come to this meeting and to come to Boulder. It's always a real treat to come here, and for those of you that know me, I tend to spend my summers here for the last 11 or 12 years, so I usually have a nice slide and call it except for the summer. There was no plan to come. We're going to do something different. And so this invitation came out in a row. It's great I get to come and visit and get a little taste of Boulder. And I like it so much here that what I'm going to talk about today is really what I want to talk about. I'm going to talk about my summer hobby. I was living in New York City, and I had to get out for the summers, and I thought, oh, there would be ice. And I came here to do a little bit of work with the people up in Angkor for a couple weeks at a time. And one of those lunchtime conversations led to what has become a significant part of the world that I've been on doing. And what you're looking at here is two model simulations. So I tend to work on regional scales in the ocean. We all know that the global climate system, but there's significant problems in these models and how they represent coastal regions. And there's some good reasons why we want to do better in them. And these are the kinds of scales that I work in. Here's a California current model that we do. And here's an alternative model. And what I'm going to talk today is mostly about up-buildings systems, which are the entity of the Eastern Bundy Current. Not only mostly in Western Bundy Currents, but the Eastern Bundy Currents, that's the part that really dominates our understanding of the dynamics and the physics of these models. So the results I'm going to show you are the results of the work that we've been doing over the years with people up in Angkor and some collaborators, and there's not just no one here. And I will probably get at the University of Alaska. So I know what that is. I'll stand here. So these are the Eastern Bundy Currents. Two of the most prominent ones. There's other ones that are kind of recurrent. And there's other systems as well. But here we have the California, the Peru system, and the Benguela system of the coastal of Africa. I'm going to talk a little bit of motivation why we're interested in this and why the particular approach that I'm taking in trying to model these currents within the global climate system. And I'm going to talk about implementations, two particular implementations in the California and the Benguela system, and some summary remarks. And I understand this is a diverse audience. I wasn't quite sure what to expect. But I think I have enough motivation in there. If you don't quite get all the details, no matter where you stop me at any point, you need to hopefully understand why we're doing this. So why is it that we are interested in doing regional models or looking at regional impacts? So you can go back about 40 years to the first time simulations that Jim Hansen did, and he might not be at GFGL and Hansen at NASA. And they did it back at the end of the calculation. Let's say if you double see or two, you're going to get 240 degrees. 40 years later or 50 years later, almost a model like PCC, what are we, AR5, and movie one and on and on. The answer is still the same. It's not that we haven't learned anything about it, right? We've learned a lot. We're reducing the uncertainty in these models. We have much better understanding of how the climate system is working, right? But basically, we know what the answer is, right? The solution is not going to be scientific solution, but this is going to be political social solution, right? But there's still a lot of room for understanding. These models were designed to look at the global climate. The next phase for us is to understand the regional impacts and what we need to, how the climate is going to affect particular areas. What is what are the adaptations that different regions will have to take in terms of responding to the forcing that we're adding. So there's a question of regional impacts, as we just said, and we need to improve our understanding of how weather and climate is connected. Is the next hurricane that's going to hit us? Or is it really because of climate? Is the fire up in Canada because of climate change or not? Or what is the contribution of climate to that kind of shorter scale, time scale phenomenon that we observe? So climate model biases. I'm going to talk a little bit about that. Another really interesting problem is that if you look at the global climate models, they do generally fairly well, except as I will show you in regions coastal regions. So trying to understand what is missing from those models and how we might be able to improve on them is part of the motivation here. And also ecosystems. Not only are these coastal regions important in terms of physical climate, it's also where your fisheries are, significant amount of protein in the world comes from Eastern Hungary, so the international populations or other small projects that they're in the ocean. So trying to understand how these models are doing and linking it to ecosystems ultimately is absolutely important. So here's a picture that if you were, if you were in another country, you would have seen many, many variations of this before. It shows you here you have temporal scales, so it's a phase of diagram. Temporal and spatial scales, and you look from the climate system from the planetary scale and for their 100 years or more down through turbulent patches, molecular processes. And all of these scales are there in the ocean, and this is the big challenge that we face. How do we model turbulent patches in terms of the climate? So we have many, many scales that we have to probe. If you look at most of our models, this is the phase that they live in. Even our high-resolution, you know, eye-popping, beautiful results that I showed you before, at best they're beginning to resolve some eddies and synchrons. So maybe over 10 kilometers and maybe a month or so, right? We still have a lot of phenomena here on turbulence, the plenipers cells, biological processes that we are not resolving at all. So, oops, not a button. So there's still a lot to explore in terms of how these things are. So one of the things that we talk about is downscaling. So what we do is we take information from the climate system, large-scale climate global models, and we try to bring it down to maybe a regional model that can then start resolving, even if we're short periods of time, some of the processes that we're interested in. And what I'm going to talk about today is to try to convince you that there's a two-way feedback between these. So it's not just that we need to understand how the climate system is driving these smaller-scale processes, but it's the feedback, right? The smaller-scale processes that eventually ultimately form the dynamics that contribute to climate. Or at least we've got to convince yourself that they're not important. So continuing on in terms of motivation, here is looking at the biases of these climate models. You're looking at a long simulation that was done here at NCAR some years ago now. This is a century point five. And this is the anomaly relative to our observational data. So our climatology. You see that you run this for a long periods of time. You see that the models generally do pretty well. This is a global ocean model, except, right, in the regions of eastern boundary and western boundary currents, right? This is, of course, where most of the people are on Earth live. This is where the ecosystems live. So the global climate models are great, except if you really care what's happening locally in those areas, right? This is like a significant motivation that we are aware of it, right? It's been for a while. It's not been sold. It turns out that cranking up the resolution doesn't always help you. So there's a lot of work to still be done here. So too warm, too cold. If you look at, I think this was from AR4, where we read off the Randall chapter one. You know, there's an admission that there's significant errors in these models. And there's always this assumption that as you go to higher resolution, you're going to get rid of them. So our computers will become bigger, we'll run at higher resolution, and you don't have to worry about it so much. I'm going to put it to you that there's some truth to that, but there's a lot more going on. Okay, ecosystems. I always like to show a fish story in there. This is, you're looking here at fluctuations of Sardinian anchovies over decades. This is going back there about 1920, and here you have the California system in Japan, which is a west among the current. You have Peru, you have the manguella system. And it's this amazing low frequency variability of this fluctuation, of these populations in Sardinian anchovies. And you also see this kind of tantalizing synchrony between the systems, manguella and the Atlantic being out of phase, but still there's this incredible synchrony between all these systems, right? So you have these long time scales over which this population is going up and down. And on the right, what you're looking at here is the distribution of eggs in the California current for Sardinian. And what you're beginning to see here is sort of the interplay between the biology and the physics, because you're looking at turtleneck actresses, you're looking at eddies in the ocean. And the question is what role do this kind of mesoscale, this is two different years, so you've seen this significant inter-annual variability. And the question is what role does this scale of spatial processes ultimately be played in the long-term population dynamics if you will about this particular species? This problem has been with us for a long time. If you ever read the Stein-McCanary role, you know, it was describing this collapse of the Sardinian populations here at the time that thought it was open because of overfishing. Now we know that, well, probably that wasn't the only cause of it. But trying to understand this spatial and temporal links is we have not cleared it out yet. Okay, so how do I think about these things? We're thinking about small time scales and small spatial scales and much longer periods. So the way that we've been approaching things is we started a global climate model that gives us the background. And then we try to maybe downscale it and two-way downscaling to a regional model that can start to resolve these turtleneck processes that we think may be ultimately important. And then we can link these models to perhaps a concentration-based nutrient cycling model, so your nitrate and your carbon and whatever else you need. And then we can even link to these ultimately fish models. So we actually have fish and boats swimming in our models and trying to understand how all these things are linked together and what are the feedback between all those potential systems. A little bit of an aside. So I'm going to jump now to some results looking at these bias problems. This is a same picture. I'm going to show you once where we're enhancing the resolution of the ocean, but there's other ways to approach it. One of them is, for example, increasing the atmospheric resolution. So here's your classical two-degree global model and what happens when you bring it up to a half-degree. And when you can see that in some places, for example, especially in the purutula, you get a significant improvement. A few years ago now, basically, if you can resolve holographic effects, the Andes, you will improve that representation of that particular system. However, there are still problems that remain in other places. So it's not the only issue. You can improve how the earthy fluxes are computed, and then there's the work that we've been carrying out, which is trying to increase the ocean resolution and try to understand what is the role of the ocean and the dynamic of these two wonder systems. So the way we go about it here is a schematic of what the NCARC or, for that matter, most climate models look like. You have a flux stopper that sits at the middle of all these different components. So you have the atmospheric model, the land model, the sea ice model, and you have an ocean model. And then each of these models computes its own state about a day or so roughly, and then you pass it to the coupler and the coupler computes all the different fluxes that go back and forth between these different systems. What we have done some years ago is we took up the ocean, we put in a new layer that allows you to have both the global model and the regional model around, so you can see that level. And so both in this new driver that communicates between the two. Okay, programming nightmare, but we do this. We do this regularly now. A picture maybe is a better way to do this. You have here your global ocean model, and we put down a region where we want to run a higher resolution regional model that we've worked with before. This extracts boundary conditions. We compute then this and every day that it gets together and we form a new surface, global surface temperature field that then gets passed back to the flux stopper and the atmospheric never knew where it came from. So this is what's happening. We are running this. We started with variability type simulations. So we have the basement simulation. We run this for 150 years. We branch it from an 1870 control run. So this is, if you're not familiar with the climate tolerance, this is a typical basement simulation. And then we run the composite. So we're running another 150 years and we have this composite problem and ground solution and all the other standard components of the climate model. And then we do two tests. And there's a very important point here. You can think of this high resolution model as introduced in the perturbation to your global model. So all of a sudden we're resolving up well. You have colder water that wasn't there before. How do we separate the impact of that particular cold patch from the natural wearability? These are nonlinear models. If you run these models for about 1,000 years or 1,500 years, like in Pakistan, and you, let's say, break it up into 10-year periods and you start taking differences, you will see a lot of variability in there, right? The model is nonlinear. Now you're going to introduce a perturbation. You kind of take that a little bit and you're going to get a response. And how do you separate that response that's due to that from the natural wearability of the system? That's a very, very subtle and important point to carry out. And by the way, that's why we run for like over 100 years. It takes us about that long to get statistical significance. So we do that, and wherever you see these little dots here, that's where we have 95% statistical significance that what we are looking at is actually a difference due to the perturbation we introduced as opposed to natural wearability of the system. So what you're looking at here is now temperatures average over the last 140 years of the simulation for June, July, August and October, November, winter months, spring months, and this is the difference. So we average them for those 140 years and then we sort of pick them up in continued periods and we do this t-test to separate out the natural wearability from the force of response. And this is what we get. So let me see that we can indeed start addressing. We can get in the summer months, these are the upwelling months in the California current, we can get something at a speed of three degrees. I think I neglected to say if you look at these bias pictures I'll show you before, the biases can be 8 to 10 degrees too warm, right? These models are really, really off when they are off. So we're not recovering the whole signal here but we're recovering some of it. Basically, I'm going to go quickly through this. We then can diagnose what's happening. Let me see how we've done the time. Okay, about 10 minutes or five minutes. Here is the two-degree too cool and then we look at the fluxes and basically what that does is it leads to an increase in the low clouds and then you go show irradiation. Increase is the only irradiation. And the final result is that it actually leads to an increase in the total heat flux into the ocean. So it will be counterintuitive, right? But if you make the water cooler what that's going to cause is it's going to cause more heat to go into the ocean, right? So what this is basically telling you is that you have a couple of response. If you're interested in what the upwelling system is going to be in the future, you cannot simply take some future scenario and ignore the couple of feedbacks. So the couple of feedbacks are very, very, very important in this system, right? You will overestimate what the upwelling would be if you ignore the damping response from this increase in heat into the ocean. Okay? Upstanding. So I've told you before that you have to look at what the feedback is to the global system, not just water, right? So you're looking here now at the air surface temperature. So this is the 2-meter temperature coming out of the model. Again, we have about the statistical significance down there, some here. And we see, again, it's not a very big number, but we are actually cooling over North America by about, you know, a quarter to a half degree Celsius over 104 years. So there is a feedback to a larger-scale system. Okay, so we know we learned a lot. I'm going to go quickly now because I think I only have about three or four minutes. So we said let's try this in the Benguela system, right? It's another system that the models have a really, really hard time doing well. And you can see here right, these are upwelling cells of the West Coast of Africa. And you see these are very small-scale features and a very complicated current structure with polar undercurrents and surface currents and temperature funds here, the Benguela fund right there. This is from the paper. We just did a kind of, you know, last December. So we said, oh, you know what we're doing, right? We're going to plot it, put down a really high-resolution ocean model in there. We're going to run it again and we're going to do better, right? We're going to get to between degrees. Well, guess what happened? Here's the bias. We made it worse. We made it warmer. We actually put in higher resolution. We resolved some of the things we do to warm it. This is a head statuette. Again, you cannot blindly just throw a resolution that means unexpected things to happen. Very important lesson. So we went through this whole exercise. Here, I'm going to skip this in terms of time. So here is that warming. Then we said, oh, let's increase the atmospheric resolution. So we went from a one-degree atmosphere to a half-degree atmosphere. We made it a little bit better but still to warm. So then we started looking at how actually this interpolation is happening between the winds, the global winds, and the ocean. And it turns out that atmospheric scientists and ocean scientists don't always talk. We need a flux coupler between the scientists as much as we can in the field, right? They're grids. You can have an atmospheric cell, right? That will be partly over the ocean and partly over the land, right? And you have two different boundary layers that can happen, right? So when you're actually interpolating and if this algorithm is very, very sensitive to the windstress right at the coast, right? You have no idea when you're applying it whether it's coming from mostly land or mostly ocean, right? This is something there's a lot yet to be done even at the low resolutions. So what we did is we said, okay, we're going to make sure that whatever wind we are applying here, we're actually going to be dragging it from the cell and that atmospheric cell sits purely over the ocean. So we should be doing that a little bit, right? So we... I like to say we got very in our interpolations. And lo and behold, you see what happens. This is not stronger wind. It's not higher resolution. It's this resolution, right? It's just the way that interpolation got done. So we do that and we can get a much nicer response. And here's two cross sections. So you see what the ocean... This is what it was before. This is the half-degree atmosphere here where the temperature structure looks like. And then once we shift to the winds you'll see that we can start getting upwelling. And here is your upwelling. The deep and clear water is reaching the surface in a very, very narrow band. And the only difference between these accelerations is the way the atmosphere will be interpolated onto it. Of course, you also need the ocean resolution to be able to resolve this problem, right? So it's not that you didn't need the high ocean resolution. You needed it, that there was only part of the story. I'm going to go quickly, basically at the minute or two. I just want to talk quickly about ecosystems and here is... This is our point of current. And we did the simulations with a carbon cycle in it. And here's what we learned. Here's an ocean model of 30 km resolution and 20 km resolution and 3 km resolution. And we're looking at the outgassing to the CO2 exchange, right? So this is, again, important for climate. The red line here is the CO2 outgassing limit. So if you're... if you're in any line of latitude and you go west, right, so you're outgassing CO2 to the atmosphere and then the blue line here is the equilibrium line. So when you get to that point and you integrate across, you end up with a net zero CO2 exchange between the atmosphere. So it's not a way to look at this. And here's what you see. If you're running it at 30 km resolution, which is still better than most climate models nowadays, which are running at one degree for about 100 km resolution, you see that the carbon account to higher resolution, that kind of goes away. Here are some of the durations. Some are in the California Current and these little dots. And here is, you see that the 30 km model, right, this red line grossly over us makes it source of CO2 coming out of this upwelling region in the California Current. And once we go down to 10 km resolution, we do much better relative to the data that we have. And then going from 10 to 3, maybe we're getting some of the details better, but the basic characteristics remains the same. Okay, so when you bring in an ecosystem consideration, it may give you a different answer to raise the resolution that you need or what are the features that you need to resolve. So depending on the question that you're asking, you might approach the problem a little bit different. That's perhaps not surprising. So some final remarks to close this. Upwelling is a couple of phenomena. And especially if you're interested in making future projections, you have to consider all the feedbacks in the system. The ocean atmosphere feedbacks are very, very significant. And if you ignore them, you do it at the wrong risk. There's prior feedbacks of all the things you can make yourself. One of the things that we learned is that the dynamics upwelling is a very generic term. In terms of, like, you blow the wind, you know, the air for rotating, you want to move to the right, you bring water up. It turns out that the dynamics comes to start looking at the details of some of the different systems, California too, and there's different things that matter. I mean, yes, the basics are at the right, but the strength of the upwelling relative to the horizontal objection would make a very big difference in what the final product that you're looking at looks like. Ocean dynamics is important and resolution is not the only fix but you do need that. And the upwelling is a good example of the atmosphere and ocean resolution to address any problems. And then we're doing similar things in the West Indonic kind, and I'll talk about that. And one final thought I'll finish with is we can't wait for the high resolution upwelling models. The computers, to discuss this with computers are getting more powerful, they're not getting faster. And for all kinds of problems, it's a real big challenge. They're actually getting slower. I mean, I just bought a small cluster and each core is a little bit slower than the ones that I built three years ago because they had to slow, they couldn't come in more into them, they had to slow down the clock speed down. If you have the latitude, you cannot put more than 300 system processors in that time.