Okay. Well, thank you all for coming and for attending this VIP seminar. I have to make this sort of formal announcement because it's being recorded. So, this entire talk is being recorded, so any questions after the talk, please use the mike because as I said, it's being recorded. And because it's being recorded, I have to introduce myself. I'm Graeme Stephens.
I'm the co-director of the Climate Center here at JPL with João Teixeira, and this is one of the Climate Center's VIP seminars from Jen Kay from the University of Colorado in Boulder. And I've known Jen for a quite a time. In fact, she actually ... I think she ... I might not get this quite straight, but she came to see João just at the heels of graduating with a PhD, and I can't even remember exactly where the PhD was from.
I should know, but let's just go. Okay. And she came to see João because she was looking for postdoc position, I think. And that was probably right, right? And her research with that time was in clouds, in ice clouds, and so forth. And I suggested to her, "Well, the polar regions with the new A-train," that was sort of just being pulled together in its pieces at that time was a goldmine for looking at polar, polar processes and polar clouds, and subsequently she did a postdoc at NCAR and then moved into ultimately into a faculty position at CU where she's now basically the superstar in the cryoclimate community. So, we're really, really pleased to introduce her today and have her as a seminar speaker.
She's recognized now as a superstar. She's getting all sorts of awards from societies. She's just received this year the AMS Houghton Award for the young, I guess it's a young or early career scientist I suppose. I should know because actually I was the first recipient of the Houghton Award, but that was back in the days of the dinosaur. And I can't remember what it was actually for.
But Jen is now this is a 2019 recipient I suppose. 2017. 2017. Oh, okay, okay.
Thanks. She was also the Turco Lecturer at AGU last fall, and part of this talk, not all of it, but parts of it would be as taken from the Turco Lecture. And so, please welcome her. She's, as I said, a rising star and someone at JPL we want to keep close to the JPL family because who knows? She could be PI of something major in the future.
So, anyway. So, welcome, Jen, and we look forward to your talk. Thanks.
It's a pleasure to be here. Just to continue along Graeme's story a little bit, I did arrive as a postdoc, and Graeme did suggest looking at polar clouds. But then, that next summer was a very exciting summer in the Arctic.
It was the summer of 2007 when the Arctic sea ice plummeted, and it was dramatic. People who had been in the field for 30 years were saying things like, "The Arctic is screaming," and we've only seen more sea ice loss since then. And so, it's been a very interesting and rich place to think about the climate system and how it's changing and the physical processes, and we're observing right before our eyes processes that are relevant for climate change. This is not something to be waiting for. This is happening right now. The background for this is actually I went to sea ice camp.
I think you think after you're done being a kid, you don't go to camp anymore. But I went to camp, sea ice camp, and we did some fieldwork off the coast of Alaska, which is where I took this picture. And in this picture, I'm on a snowmobile, and I'm behind somebody else. And we're going over some very thin ice. It's first-year sea ice.
It's about 40 centimeters thick. And the person driving, Chris Polashenski, who's a sea ice scientist, said to me, "Don't worry, Jen. If we hydroplane, we won't sink through."
So, the ice was exceptionally thin. It was thinner than even the sea ice expert had really realized, and I think you'll see that as a theme through my talk. There's a lot of change in the Arctic, and it's really surprising a lot of us. So, I didn't fall through, and I'm here today.
Yay. But, yeah. Sea ice camp was still a lot of fun.
So, we're all a bunch of scientists mostly I think, so I think looking at the data is always a nice place to start. So, the observed Arctic sea ice loss. So, from 1979 through last September, September is the time during the year when there's the minimum amount of Arctic sea ice, and you can see the dramatic 2007 loss there, right? We went from here to here.
That was first year of my postdoc when I started studying clouds and how they interact with sea ice. But we've had even more dramatic minimums. There're a lot of ups and downs to this curve. There's a lot of variability, and I think a very basic question.
Okay. I can compute the statistical significance of this trendline. Believe me. It's highly statistically significant, but that doesn't really answer the questions that I think we have about the changes in the Arctic.
We want to know why is this happening? How much of this is due to the fact that we've increased greenhouse gases in the atmosphere, and how much of this is just due to variability intrinsic in the climate system? You can get a statistically significant trend, and it can have nothing to do with climate change. And so, we need to try to use tools to try to understand what's going on, and so here I've introduced one of those tools, just observations, seeing what's going on. But I've also since my postdoc time with Graeme done a lot with climate modeling as well, and I think those two tools are really complementary for helping you think about these why questions, the really hard questions. Why is this happening? It's also important to know that this ice loss was predicted a long time ago, greater-than-global Arctic warming, so the fact that the Arctic is warming three times faster than the global mean was seen in the earliest climate models we had. So, here's from Manabe and Stouffer, 1980, "The CO2-induced warming of the surface air is particularly large in the high latitudes." So, one of the pioneers of climate modeling.
Believe me. Climate modeling now is a lot different than it was in the 1980s. If you talk to folks at the NCAR or GFDL, they thought very carefully about how to get one plot. These days, I can make a million plots in one day, and the question is what does any of it mean, right? But they made some very useful and interesting model runs and some very useful and interesting plots, and they did so with the supercomputers of their generation, which had about a half a megabyte of memory, which I'm pretty sure this slide and the images on it are more than a half a megabyte of memory. So, they had to think very carefully about what to include in these models, right? Which physics dominates? What's really important here? And they had to make a lot of sacrifices in terms of resolution, in terms of the representation of processes. I would add I'll talk a lot about clouds in this talk, and I want to remind you of what clouds looked like in the 1980s in the state-of-the-art climate models.
They were zonal monthly means, year after year, the same zonal monthly mean. So, you had more cloud in the tropics and subtropics, but there was no Hadley circulation. There was no mid-latitude storm track.
There was certainly no interaction with sea ice loss. They were just they were set. They were fixed. So, we'll ask, "Was that a good approximation? And how? What did we lose by having clouds be so simple?" Well, some things that we see in modern day climate models look shockingly similar to the plots of Manabe and Stouffer, albeit in color, which is pretty fun. So, similar to Manabe and Stouffer, 1980, modern climate models show that during the summer, shortwave radiation is absorbed into the surface ocean fueling a strong, surface albedo feedback. And that during fall, that heat that accumulated in the ocean in the summer is fluxed back into the lower atmosphere warming it, warming the near surface especially, and reducing the atmospheric stability.
And all of this was present in these early climate models, this kind of seasonally delayed Arctic warming, surface-based, resulting from sea ice loss, and the figure on the right here is just from a paper by Ariel Morrison, who just graduated with her PhD actually last week, so congratulations to her. She was a student in my group, and she looked at things like the ocean heat content, the total heat flux, and the stability as a function of month of the year and then also different time periods. And one thing you can see in these simulations which go into the future from kind of a present day in the green all the way through the yellow, as you see there's more and more heat being accumulated in the surface ocean during the summer months as you go in the future, that the heat flux from the ocean to the atmosphere is still pretty small in the summer, even way into the future, and that as you move into the future, the stability really reduces, especially in the fall when that heat gets fluxed into the lower atmosphere.
So, all of this is consistent with Manabe and Stouffer, 1980. And this is the CESM1, so modern climate model. So, kind of going back to the observations, we also have observed increases in absorbed shortwave radiation associated with summer sea ice loss. So, this is a figure from a review paper I wrote in 2016 just showing the trends. This is based on series on top of atmosphere shortwave observations showing the large amounts of increases in shortwave radiation over the Southern Ocean, and then here in map form, also, we have the trends in sea ice fraction, so again, from satellite.
I will note there's not a one-to-one correspondence between the two, but there's certainly places such as here off the coast of Alaska, Northern Canada where there's large increases in absorbed shortwave radiation associated with the sea ice loss. There's reasons that make a lot of sense for this, these not matching perfectly. One is, for example, that the absorbed shortwave radiation is a function of the incoming shortwave radiation which is maximized in June, and as I mentioned, the sea ice minima occurs in September.
August is when there's less sea ice, so there's some seasonality issues there. And then, also, the absorbed shortwave depends also on the surface albedo of the sea ice which varies quite strongly depending on if you have melt ponds on the surface or snow. Snow-covered sea ice has an albedo .8, whereas melt ponds on the sea ice, you can get down
to .4 or .3. So, the absorbed shortwave radiation is not just a signature of the binary, is there ice there or not? There's other things going on. When I show these graphs to folks who don't deal with clouds, they're like, "Yeah, yeah, yeah. Surface albedo feedback, that's like the classic feedback. Everyone talks about it. Why are you so surprised?" and I want to say, "This is kind of surprising."
Most of the world's ocean is covered with cloud, so if you remove the sea ice, which is sort of a cap on the ocean, you might expect to see more low-level cloud associated with that sea ice loss. But this discovery or this observation of increased absorbed shortwave radiation associated with summer sea ice loss is consistent with the lack of observational evidence for a cloud response to summer sea ice loss. And so, I think that's really important to say the summer, there's a very active surface albedo feedback, and it is not limited by a cloud response, i.e., an increase in low-level clouds associated with a newly open ocean in the summer.
So, Ariel looked at this in a very sophisticated way, looking at cloud-sea ice feedbacks, and she looked within something she calls "the intermittent mask". So, this is the region of the Arctic where there's over one patch of ocean, some years, there's ice. Some years, there's water, but you're not mixing, for example, the Beaufort High part of the Arctic where there's sort of semi-permanent, high-pressure circulation to the North Atlantic where there's just never any sea ice.
If you just take the 70 to 90 north, you really can average together a lot of things that don't make sense physically. So, she's just looking at one patch of ocean. Is it ice? Is it water? What do the cloud profiles look like? And these are cloud profiles from CALIPSO, and you can see on the left, we have the summer within the intermittent mask. We have the cyan and the blue, and they basically lie on top of each other within the interannual variability. Oh, thank you.
Graeme Stephens: [crosstalk 00:13:59] Yeah. So, on the left here, these two are the same. So, the cyan again is over sea ice, the blue over open water. Again, no cloud response to sea ice loss. In the fall, we see actually an increase in the low-level cloud over the open water. This is, again, consistent with the physics that was in Manabe and Stouffer, right? We're in the fall now.
That heat and also moisture is fluxing from the open ocean into the lower atmosphere, and so actually, there are increases in low-level cloud associated with the summer sea ice in the non-summer months. So, there's some important implications here for the influence of clouds on the observed sea ice loss that we've seen so far. There's no evidence for a summer cloud-sea ices feedback, so far that's been not something that's been important. And then, in the fall, again, these increases in low-level clouds, the fall is a complicated time where you have both shortwave and longwave radiation, and the effects compete with each other.
So, the shortwave cooling and the longwave warming, so it's sort of unclear what the sign of this feedback would be. But it is clear that these things compete, so it's probably relatively weak. So, in addition to the cloud-sea ice feedback, which we've talked about already, I do want to talk about a way that clouds have influenced the summer sea ice loss, and that's through cloud masking. So, again, if you don't think about clouds ever, you probably think about the surface albedo feedback something like this, right? You have the sun, a bunch of shortwave radiation coming in. It gets absorbed in the surface ocean. Seasonally delayed, that heat comes back into the lower atmosphere in the fall, and then you have the surface albedo feedback contributing to Arctic amplification, large warming in the lower troposphere.
But you can imagine an alternate universe where you have infinitely optically thick clouds, and none of that shortwave radiation makes it to the surface, right? So, in that case there would be no surface albedo feedback. There's no interaction between the two. And so, if you think of these two as end-members, you know probably the real world is somewhere in-between these two, but this cloud masking or this cloud preventing shortwave radiation from even interacting with the surface is a really important way that clouds in the summer, even if they don't respond to the sea ice loss, they control how strong the surface albedo feedback is.
And this is not a small effect. So, here, we have basically the planetary or top of atmosphere albedo, and we have the total. Then, the atmospheric contribution to that albedo, so the atmospheric contribution is primarily dominated by clouds. And then, the surface contribution, and you can see at high, northern latitudes, so here, 90 degrees north, that the atmosphere is contributing a lot to the top-of-atmosphere albedo.
So, there is a lot of cloud masking that happens, and this is courtesy of Aaron Donohoe at the University of Washington, which is where I did my PhD. Graeme Stephens: That's right. You had to dissolve with that line.
So, there's some really interesting things that can happen with this cloud masking term which is driven in large part interannually, variations in it, by the large-scale atmospheric circulation. So, here, what we have actually are the sea surface temperature anomalies in the late summer, so it's an idea of how much absorbed shortwave radiation went in in an anomalous sense. So, remember that 2007 was this dramatic low in Arctic sea ice extent, and you can see in addition to that, it was also a year where there was a lot of shortwave radiation that got accumulated into the surface ocean. So, here, the SST anomaly is here, and the Pacific sector of the Arctic are approaching 5 degrees Celsius. But then, interesting, I said 2012 was a more record-breaking year in terms of the minimum in sea ice in September, and yet you look, in terms of absorb shortwave, 2012 was not a very impressive year at all. In fact, 2007 is a much more, much larger anomaly than in 2012 in terms of the absorbed shortwave radiation and the SSTs.
And this is in large part due to the clouds, so 2007, the sea ice loss was associated with some very large reductions in cloud cover, which we viewed very well with satellites that had just been launched, CloudSat and CALIPSO. And in 2012, it was not as sort of "clear a year", right? There were many more clouds, and that inhibited the amount of warming that was happening at the sea surface. So, I think there's a lot of rich questions here, and actually, this is Figure 1 from CloudSat and CALIPSO's science team proposal, so we're going to be able to dig into these a little bit more upcoming. So, I think there's a lot of interesting aspects of this story, some surprises, no cloud response to sea ice loss in the summer. Some things that if you were in the 1980s, you might have thought, "Oh, my gosh.
This is so boring. It's exactly what our models told us in 1980. It's just the shortwave albedo feedback and then warming in the lower troposphere delayed." So, communicating these results and studying engagements in them is something that I'm actually funded to do through the National Science Foundation, and I figure this is a perfect time in the talk for a little break. How many people here like polar bears? Anyone? Audience: We would like to have bears.
All right. Polar bear movie maybe? A little bit of that? Audience: Hell, yeah. Okay. So, yeah. We have this project, the Polar Bear Project. We went to Churchill, Manitoba, which is on the shores of Hudson Bay.
It's at about 58 degrees north, so it's not quite in the Arctic, but there's quite a lot of sea ice there. Hudson Bay is very cold, and there are a lot of polar bears near Churchill. There are about as many people as polar bears in the fall, 900, so it's a pretty interesting place to go.
You see all kinds of things. This is a picture. I didn't take this. I'm not a professional photographer, but I was there when this picture was taken. And I saw this.
I was like as close as I am now. Don't worry. I was protected. I was in one of these Tundra Buggy Adventures. [inaudible 00:20:49] But pretty exciting, and so I'm going to show you guys just a short. It's just four or five minutes here.
A short video that's about the cloud response to sea ice loss, and it includes some polar bears. And I want to say we also made a version of this movie that does not include polar bears, and we're doing research right now to see what is engaging in these videos without really telling people. So, I'm just going to show you the polar bear version, but anyways, that's another story, and I'm happy to chat with you another time. But I figured this might be a kind of a fun interlude here. So, here we go. Oh, not without polars.
You guys want with polar bears, right? My name is Jen Kay, and I'm a professor at the University of Colorado. In this video, we introduce feedbacks that are important in the Arctic climate system. Ariel Morrison: Hi, I'm Ariel, and I'm a graduate student working with Jen. I study Arctic clouds and sea ice.
I'm here in Churchill, Manitoba, which is a really cool little town on the coast of the Hudson Bay. Polar bears spend time nearby while they wait for sea ice to refreeze each year in the fall. In fact, during late fall the 900 human Churchill residents share the tundra with about 900 polar bears. Now, talk about neighbors you'd want to get along with. The nearby Hudson Bay has sea ice on it during winter.
Sea ice is bright, white, and reflective. The high reflectivity of sea ice means that it has a high albedo, which is simply a measure of the reflectivity of a surface. For example, if a surface reflects 50% of the sunlight that hits it, then that surface has an albedo of .5. Albedo ranges from zero, which is reflecting no incoming sunlight, to one, reflecting all incoming sunlight. The albedo of sea ice is about .6 on average. Snow on top of the sea ice increases the albedo to .9.
A lot of sunlight can get reflected from fresh snow which is why you wear sunscreen and goggles when you ski or snowboard in Colorado. Snow and sea ice maintain a pretty high albedo for the Arctic surface, but that albedo changes as snow and sea ice cover change. Every year in the Arctic, snow and sea ice start melting in the spring and continue melting through the early fall. When sea ice melts completely, the dark ocean is exposed. Since the ocean underneath is very dark, it has a really low albedo, .06.
The ocean only reflects away about 6% of incoming sunlight, so it absorbs a lot of energy from the sun. When the surface ocean absorbs energy, it warms, and more sea ice melts. As more sea ice melts, more dark water is exposed. More open water in turn leads to more sunlight being absorbed by the ocean.
As the surface ocean absorbs more sunlight, more sea ice melts, and you get the picture. This whole process is called the surface albedo feedback. Feedbacks are a key concept to understand in the climate system. Feedbacks can be positive and negative. Let's say the climate system is warming.
A positive feedback occurs when factors further increase that warming. A negative feedback, on the other hand, occurs when factors reduce that warming. We are particularly interested in Arctic feedbacks because over the last three decades, the Arctic has warmed three times more than the global mean. The evolution of the Arctic climate system is influenced by many positive and negative feedbacks. The surface albedo feedback is a positive feedback that further increases Arctic warming and sea ice loss. What are the negative feedbacks in the Arctic? Specifically, is there some sort of negative feedback that could slow down sea ice loss? To talk about a negative feedback in the Arctic, we're going to move up into the atmosphere and talk about clouds.
Because clouds are bright, they have a similar albedo to snow-covered sea ice, about .8 for clouds, and remember, about .9 for snow-covered ice. In the summer, when the surface albedo feedback is strongest because the days are long and there's lots of sunlight, forming clouds over newly open ocean could replace one lost bright surface, the sea ice, with another, clouds, thereby maintaining a fairly high albedo for the Arctic surface and reflecting sunlight away from the ocean. However, in my research, I found no evidence that Arctic clouds form in response to sea ice loss during the summer. Summer clouds can't be relied upon to reduce the strength of the surface albedo feedback and slow the rate of Arctic sea ice loss. The Arctic climate system is a complicated place with many positive and negative feedbacks.
We learned about the importance of feedbacks, the positive surface albedo feedback that we have evidence for, and a potential negative feedback we aren't seeing evidence for. The Arctic has warmed and is projected to warm a lot more because of its particularly strong positive feedbacks. Yeah.
Thank you. So, that's one of the videos. We actually made three videos, and I think it's really interesting. I don't know.
I've seen these videos many times, but I still notice so many visualizations we do of the Arctic sea ice albedo feedback do not include clouds, so that's my plea to think about clouds if you're making some sort of visualization associated with the surface albedo feedback or talking about it because I think people forget that there's not only the cloud-sea ice feedback but also the cloud masking that's really important. So, I want to kind of continue the theme of observations and models and talking about clouds, so the representation of clouds and climate models has become more sophisticated, right? We don't live with zonal monthly mean clouds anymore. In fact, there's a lot of complexity that's gone in, and yet interestingly, the vertical and seasonal fingerprints of Arctic greenhouse warming have not changed. So, by vertical, I mean it's mostly at the surface, and by seasonal, I mean it's mostly in the fall, the non-summer months. It really sort of fingerprints this process, the surface albedo feedback and the heat accumulation in the surface ocean.
So, that really led me to a sort of existential question like, "Do the clouds matter?" Right? I mean they got the right answer 30 years ago. I mean we did all this complexity, and it hasn't changed. I've already given you one example of how they matter, the masking, and so let's explore that little bit more. Oh, this is just this figure from Manabe and Stouffer. I just want to show you, here, they have zonal vertical mean.
So, this is pressure, and then we have the latitude on the bottom. The largest warming happening here in the high northern polar region, so that's that fingerprint. We see the same type of thing now in models. And models, actually, some of them do a really reasonable job of reproducing this lack of observed sea ice-cloud feedback during the summer and an increase in the cloud cover during the non-summer months associated with sea ice loss. So, here, this is Ariel's second paper from her PhD where she reproduced the methodology of the intermittent mask within CESM1, which is a climate model, and she showed that the CESM1 matches the observations.
There's no change in the summer over open water and sea ice. These two kind of lie on top of each other, and then over and during the fall and the non-summer months, we see this increase in low-level cloud cover associated with the sea ice loss. So, that's kind of nice.
I mean there's a lot of things that that one can say models don't do well, but this basic physics was being reproduced and being reproduced for the right reasons. So, that was great to see. There are of course a number of issues with Arctic clouds in climate models. One of the issues identified in many models, and CESM1 is just an example of it, is insufficient opaque cloud cover.
And the opaque cloud cover here is defined very specifically. This is opaque according to CALIPSO, so we're looking at a comparison between the opaque cloud and CESM1 and the CALIPSO observations using a satellite simulator, which lets us make apples to apples comparison between the observations and the model. And what we can see is that basically we're just not seeing enough opaque cloud in both the summer and in the fall, and this is important because this comes back to this cloud masking effect, right? If we don't have enough opaque cloud, we probably don't have optically thick enough cloud, and we're probably not getting enough cloud masking in this model. But at least the cloud response to sea ice loss looks reasonable, and the physical mechanisms dominating it also look reasonable.
So, yeah. The implication here CESM1 has insufficient cloud masking, and it's not alone. There's many models that really struggle with the representation of mixed-phase cloud processes at high latitudes. So, one can get all into, "Okay.
The model's clouds aren't right. What does this mean? What does this really mean?" and you can get all upset that it doesn't work. But then, sometimes, I really like to take a big step backwards because I get asked at a barbecue or talking to my grandma, "Well, how does this contribute to the uncertainty that we have in future climate projections?" And so, I'm going to take a huge step back here and talk about what the sources of uncertainty are for future climate projections. And I'll bring it back to the Arctic, but let's talk really, really big picture here. And I want to be really clear, too, that when I talk to somebody at a barbecue, I always specify that, "Uncertainty is not ignorance." This is a quote that's something Kerry Emanuel said when he was visiting the University of Colorado.
And I think that's important because as scientists, we like to talk about uncertainties, and sometimes, that means, "Oh, you don't know what you're talking about." Well, some of these sources of uncertainty we know really well, and we know that they're fundamental barriers to prediction. And so, it's important I think to communicate that. It's not ignorance. So, the first source of uncertainty for future climate projections is scenario uncertainty.
There's a lot of uncertainty in what is going to happen in the future, and a lot of that revolves around human decisions, not decisions that just you and I will make, but the whole planet will make. So, what will be the level of greenhouse gases in the future and aerosols? And I would say this is a very large source of uncertainty, especially on the long timescales. Like 50 years out, we don't know what technologies we're going to have. We don't know what infrastructure or how we're going to get energy.
All these things I think on a 50-year timescale are pretty open, and I think there's a large uncertainty there that has nothing to do with science. I mean some to do with science, but nothing to do necessarily with physical climate science. The second source of uncertainty is internal climate variability uncertainty, and this is a source of uncertainty that is irreducible. So, the climate system is chaotic, and just like we know that we can't predict the weather out maybe more than a month, six weeks because of the inherent chaos in the system and that's in the best-case scenario, there's also uncertainties associated with climate projections that are just part of the internal chaos of the system. We're never going to get around them. So, this is uncertainty from chaos.
It's uncertainty we cannot reduce, and we can try to quantify it using models to try to get a sense of how big it is. I think that's really important to do, but it's always going to be noise that we yet struggle with when we try to make future projections. And there's some hints that this is really important on short timescales and small spatial scales, and you can see that. And when you think about going from global warming, canonical warming over a hundred years averaged over the globe to what is the weather forecast for tomorrow in L.A.? Those are kind of the end-members to be thinking about.
And then, finally, physical process uncertainty, and this is I think why most of us have a job, right? We're really interested in the physics of the system and how it's going to evolve, and then we certainly have some uncertainties associated with this. For example, the representation of cloud processes is really important for driving some of the intermodel spread in metrics such as the equilibrium climate sensitivity. And so, scientists do really have a strong role in this source of uncertainty because we can reduce it, presumably by understanding cloud processes and other contributions to uncertainties better, we can reduce the uncertainty in future climate projections. But I think these three sources of uncertainty are really important to keep in mind, and I think we've made some really important advances in understanding their relative importance. And in particular, I want to tell you about a project that I co-led with Clara Deser at NCAR, and this is an initial condition large ensemble. So, what you're looking at here is a graph of the global mean surface temperature anomaly, and this project started, as you might for any sort of big climate modeling project, contributing to international assessment reports.
We ran what is called an 1850 control run, so a pre-industrial control run. So, you see some variability in the temperature, up and down, but it's perpetual 1850. It's always 1850 greenhouse gases, always 1850 solar.
There's no change in the forcing over time. It's just some years there's an El Nino. Some years, there's not, but that's not driven by changes in the forcing. It's just driven by the physics of the model.
Then, we ran Member 1, so we start here 1850. We go through 2005 when the historical forcings ended, and then we went into the future under an RCP8.5 scenario. This is basically a business as usual scenario where we get to a thousand parts per million by the end of the century, and you can see we get to almost five degrees of global mean warming, which is quite a lot of warming.
Then, the novelty of this project comes. So, we went back to January 1st, 1920, and we redid this experiment. We redid it many times. This figure is from the BAMS paper on it, so we have 30 members. But we've actually ran an additional 10 after this.
And these members differ only in their initial condition, so their initial condition on January 1st, 1920, is different at the round-off level, so the precision of the computer, so a round-off level in air temperature fields globally. But if you look at a graph of January 1st, the first time steps of the model, you wouldn't even see it. It's not perceptible, right? The temperature pattern looks exactly the same because it's at the 10 to the -14th level, right? Over time, those small differences in air temperature start to grow, and basically, you start to sample, by looking at all of these different ensemble members, the spread due to the chaos in the climate system, so the uncertainty due to internal climate variability. And you can see here for the global mean temperature that all the different ensemble members here are represented in gray, and you can see the uncertainty is actually pretty small, right? If you increase CO2 to a thousand parts per million by 2100 in your model, you're going to warm by a lot.
And the uncertainty from the internal climate variability is maybe if you're going to warm 4.8 degrees or 5 degrees, but practically speaking, that's still a different planet that's warmed a lot. But this paper, which has been designated an ISI hot paper cited, I got some email from the science citations people. They said, "Your paper is the hottest paper in all of geoscience."
I was like, "Oh, my God. This is crazy." This paper has been cited over 600 times. It was published in 2015.
And not very many people are looking at the global mean temperature, right? They're looking at the Earth's system. They're looking at all the different variables in this. They're trying to understand the contribution of internal climate variability and the contribution of forced climate change. They're running downscaling experiments because we have the ability to run, for example, a mesoscale or a weather model embedded into the conditions that are a part of this ensemble, so it's a thrilling project to have been a part of.
There was a huge group of us that put it together, and it is super satisfying to see it so well used to understand this really basic question like, "What is the signal, and what is the noise in the context of this model?" A number of modeling centers are now running large initial condition ensembles, and I'm actually now part of a CCLIVAR Working Group that's comparing all of these different initial condition ensembles. And I think there's a lot of work to try to understand how different models represent this signal and noise going on now, but it's been a really interesting ride. If you had asked me when I wrote this paper, which was during the spring break of my first semester of teaching at CU, if I would be standing here talking to you about it now like this, that would not have been my answer. It would've been like, "I hope somebody uses this." So, yeah. So, in the context, I'm going to bring us back to the Arctic and ask this question about signal and noise in the Arctic looking at September sea ice trends.
So, here, what I've plotted is just the frequency of occurrence for different September Arctic sea ice trends, and these are all trends that are equivalent in length from this period 1917 to 2013, so a relatively long period, 30-plus years. And what you can see is I have the 1850 control run, so you can see that the trends there are some of them are positive. Some of them are negative. Some trends there's more sea ice.
Some, there's less, but that it's centered on zero, right? Because the climate is stationary, right? You're not changing the forcing. And so, there's just some ups and downs, and that's just the part of the system. Now, what we can look at here is the distribution of trends from the large ensemble, so from the members.
This is the 30 members in the BAM's paper from '79 to 2013. And what you can see is that basically all of them show sea ice loss over this time period, but that there's a huge range in the amount of sea ice loss that they predict. Some predict almost no sea ice loss, and some predict a lot of sea ice loss. And remember, that this range is due entirely to the internally generated variability in the model. It's the same model physics.
It's the same forcing. And I think that this is really sobering because I read a lot of proposals on a lot of papers that motivate their study based on Arctic sea ice loss, and many are not appreciating how large the source of internal climate variability might be on them. They think, "Oh, I can improve this physical process, and then maybe I'll better match observations." But the observations, at least in a statistical sense, are an ensemble, and but they have some uncertainty associated with them that is due to the internal climate variability. And yet, we only get to observe that once.
So, that once is here, so this is the observed trend. So, you can see that the model is reproducing the observed Arctic sea ice loss. The observed trend is within the spread of trends that are produced by these different ensemble members.
So, some important implications from this. One, given how far this observed trend is from the distribution of trends that we saw in the perpetual 1850 case, so where we haven't had any basically changes in greenhouse gases or other things due to human activities, we can see that the observed Arctic sea ice loss cannot be explained by natural variability alone. But we also see that individual climate model simulations can reproduce the observed ice loss, but that the ensemble spread is large. And if you just assume that the average of all of these ensemble members is beating down the noise and gives you the signal, you can say that about half of the observed Arctic sea ice loss is due to internally driven climate variability and about half of it is due to increases in greenhouse gases. So, to make this a little more concrete, let's just look at some time series and see what individual members actually look like. So, here's the September Arctic sea ice extent.
This is Member 38 of the CESM large ensemble, and we can see it starts here in 1920, goes to 2100. It goes ice free below about one million square kilometers maybe sometime in the middle part of this century, and we see the observations on top of it, a lot of variability. We don't expect the individual ups and downs of eddy here to match any ensemble member because these are freely evolving models, but in terms of the trend or the slope here, I mean it's doing a reasonable job at capturing the rate of change.
What about Member 13? This is quite a different story. So, what we have is going along, going along. We see this dramatic loss. Maybe the observations are a little bit faster than what the model is predicting for this ensemble number 13, but then there's like this 10-year period where the sea ice actually increases, and then this huge drop.
Like here, I think we're going from seven million or almost seven million square kilometers to almost ice-free, one million square kilometers in the course of five years, which is insane. And I want to remind you that this outcome is equally likely as this outcome in terms of the contribution from internal climate variability. We are just as likely to observe this as we are this, and yet I am pretty sure that if this happens, if over the next 10 years, we see an increase in Arctic sea ice extent, we're going to have a lot of explaining to do about what's going on in the Arctic. And then, we're going to have even more explaining to do when over five years, we go from seven million square kilometers to one million square kilometers. So, the internal climate variability is still pretty large in terms of its influence on the evolution of the Arctic climate system, and I think that should make us very humble about how we approach this and describe this. When I get asked, "When do you think the Arctic will be sea ice-free if we continue to increase greenhouse gases?" I say, "The middle part of this century," and I say that very purposefully.
I don't say, "2050," because if I say, "2050" they're going to think it means 2050, and there's a range of uncertainty that has to do with the internal climate variability that's irreducible. So, I say, "20. In the middle part of this century," not 2050.
So, if you look at the large ensemble in the Arctic, here, I just have some histograms of the monthly mean Arctic surface air temperatures. Here, from 1850 in blue and kind of maybe a more representative. This period was picked actually because it was the period the first 10 years of CloudSat and CALIPSO, and you can see that basically we're at a point now where the Arctic is really starting to emerge above this variability. We're starting to see summer, winter, spring temperatures we've never observed in the preindustrial, but there's still some overlap.
But that emergence is happening ... Well, according to this, it's already happening. It's happening right now. So, with that idea that the internal climate variability is really important, I do want to emphasize that the forced response is still very large in the Arctic.
So, here, if we just look at time series of the global surface air temperature out through 2100, this is the same figure as the BAM's, but now I've added all 40 members. And now, I've also added a panel down below with the Arctic surface air temperature anomaly, and you can see these are on two different scales. This one goes to 15 degrees Celsius, so it's about three times more by the end of this century. And you could ask, "Which feedbacks? Why is the Arctic warming more than the global mean?" And if we look at the ensemble mean, the ensemble mean is not warming more in the Arctic because of internal climate variability. It's warming more because there are positive feedbacks or less negative feedbacks that are enabling the Arctic system to warm more. So, won't go into the details here, but a lot of work, including work that I've done, has really shown that it's local feedbacks that are driving this increase in warming.
And here, from a figure by Felix Pithan et al., looking across all of the different CMIP5 models, what they found, "Our analysis shows that intermodel spread in Arctic warming is dominated by the spread in local feedback mechanisms, not heat transports." So, not atmospheric heat transport, not ocean heat transport, but local feedback mechanisms like the surface albedo feedback, shown here in red, and the lapse rate feedback, shown in green. The lapse rate feedback, for those who are familiar at lower latitudes, it's often a negative feedback, but in the Arctic, it's actually surface.
The surface is warming more than aloft, so it's actually a positive feedback. And these two are actually interrelated with each other, right? Because when you absorb more shortwave radiation into the surface ocean, that's what's dominating this albedo feedback, but then that absorption of shortwave radiation when it comes back into the lower atmosphere, that's what's warming the lower atmosphere. So, that's the lapse rate feedback. So, these interconnected albedo and lapse rate feedbacks are really important for explaining large Arctic warming, and the cloud feedbacks seemed to be a rather small source. There's some uncertainty on the sign, but it's a pretty small contribution compared to these larger albedo and lapse rate feedbacks.
And that's consistent with looking at CESM1, so we also see relatively small net cloud feedbacks here, shown in gray, and relatively large, for example, surface albedo feedbacks. We do see some changes in clouds in the future in the summer. It's kind of amazing. If you look at the cloud profiles in different future years, they actually really lie on top of each other. There's some evidence potentially for fewer clouds sort of here and sort of the mid troposphere of the Arctic, but again, no evidence for some sort of emergent cloud feedback in the Arctic.
And then, the low clouds in the fall are increasing, and here, again, because we're looking through the lens of a LiDAR simulator, in the climate model, we start to see basically more and more attenuation near the surface. But we do see that kind of what we learn from today's interannual variability in observation seems to be what happens into the future. There's not some big surprise there. So, the cloud-sea ice feedbacks, we have some understanding of them, but it's not just those that matter. It's the cloud influence on these other large feedbacks, and in particular, the influence vis-à-vis the cloud masking. So, cloud masking, again, is regulating the strength of the albedo feedback in the summer, and the strength of that albedo feedback is intimately tied to the strength of the lapse rate feedback in the early fall.
And so, clouds play an important role here, and I admit to have a very cloud-centric view of the world, seeing that clouds are what's driving everything here. But I think it's important to keep that in mind when evaluating clouds in the Arctic that the masking is really an important part of how clouds are controlling how much Arctic warming we get. Yeah.
So, this is just the cartoon of the masking again to remind you. So, really, it really comes back in large part to the physics that we understood many, many years ago to understand why the Arctic is warming more than the rest of the globe. So, here's the CESM1 projected Arctic Ocean absorbed shortwave radiation increases.
So, we can see as we move into the future, more and more absorbed shortwave into the surface ocean. This is due to this lack of a cloud response to sea ice loss. And then, as we go into the future, that summer ocean heating, which is resulting from positive surface albedo feedbacks, leads to this fall atmospheric warming and especially at the surface, which is a stability reduction or a positive lapse rate feedback. And these really large feedbacks, the surface albedo feedback and the lapse rate feedback, are interconnected, regulated by summer cloud masking, and are really what is driving greater-than-global Arctic warming. So, I just want to summarize here, and then I'm happy to take any questions. So, the observations show us some pretty fantastical things, some pretty crazy things, Arctic sea ice loss, increase absorbed shortwave radiation in the surface ocean.
We're not seeing changes in summer clouds associated with summer sea ice loss, but we are seeing more fall clouds associated with fall sea ice loss and a weak cloud-sea ice feedback. There's an implication that there's a weak cloud-sea ice feedback on the observed Arctic sea ice loss. Second, I think this forced response that the Arctic warms more than the rest of the globe, as long as you look at things that can beat down the noise of the internal climate variability, what you'll see are two interconnected local feedbacks, the positive surface albedo feedback and the positive lapse rate feedback, that really explain why the Arctic warms more than the rest of the globe. And it's really important to point out that clouds play a role here as well by masking and regulating the strength of these two very positive feedbacks. So, thank you very much, and I'm happy to take any questions. Graeme Stephens: One, two, three.
We're pretty much done. Oh, hang on. We're off mike. Know what I mean? Oh, no. I need a mike. I'll bring one.
Okay. No, no, not that mike. Yeah, this one. Graeme Stephens: That mike. Speaker 5: Okay. In the original Manabe-Stouffer paper you mentioned- Yes.
Speaker 5: And he talked about intensive warming in high latitudes. That's both south and north. Yes. Speaker 5: Okay. Now, in your 2015 hot paper.
Yes. Speaker 5: Okay. You have the model simulations.
Oh, okay. What happened to Antarctic in that paper, in your paper? That's a very good question. The Manabe and Stouffer, 1980 paper, was run with what's called a slab ocean. So, it's basically like treating the ocean like a bathtub, right? So, it can absorb heat. It can lose that heat, but there's no ocean circulation.
There's no thermohaline circulation. There's no heat. And so, there's none of these things.
So, in our simulations that are comparable to the Manabe and Stouffer paper, so where you instantaneously quadruple CO2, we see a lot less warming through time, and we also see that in the large ensemble where we're running a more realistic historical and future scenario. And that's because there's a lot of heat uptake at the ocean, and the Southern Ocean just doesn't warm fast enough. So, you're not seeing the equilibrium warming being realized on timescales that are like 50 years or whatever in a fully-coupled model. Yeah. Speaker 6: You were talking about thermodynamic radiation.
Yes. Speaker 6: What is the role of dynamics, like polar vortex and stuff? Oh, gosh. The polar vortex. That's a complicated question. I will say when we look in things like the large ensemble, there are changes to the large-scale atmospheric dynamics, but they are small when you compare them to things like the surface albedo feedback and the lapse rate feedback in driving greater-than-global Arctic warming. There are changes, but they are not what's controlling- Graeme Stephens: That's interesting.
... greater-than-global Arctic warming. I also want to bring it back because I know there's a lot of discussion in the literature at meetings about the influence of Arctic sea ice loss on mid-latitude weather, and I want to connect these things because you probably haven't. Maybe some of you have, but I want to do it really explicitly. In the fall, when that heat from the absorbed shortwave radiation in the summer comes back into the atmosphere, that is the reason that we have any hope of seeing the Arctic sea ice influence mid-latitude weather. So, during the summer, for example, when there's no heat coming out of the surface ocean to affect the atmosphere, there's no reason to expect that you would see an influence of the Arctic sea ice loss on weather at all, right? Graeme Stephens: Mm-hmm (affirmative).
The heat's going in. The heat's not coming out. So- Graeme Stephens: It was interesting, and just before I pass on the next question.
Yeah. Graeme Stephens: It was interesting that you made the statement from the model. Yes.
Graeme Stephens: That it's the local processes in the Arctic and not the heat transport from south, which I found was surprising. But, anyway, that's just- Yes. Graeme Stephens: This piece speaks back to the question about large scale dynamics versus local processes. Yes.
Graeme Stephens: Anyway. Yes. Graeme Stephens: No response. Just an observation.
Interesting, for sure. Speaker 7: I guess this is kind of related. Why is there no change in the summer clouds? I mean is this basically just the stability is dictating that it's too low? Good question.
Yes. So, in the summer, all the heat's going into the ocean, right? The sea ice is melting. If you look at like a profile of temperature, it basically looks isothermal, so you don't have some sort of instability. If anything, the atmosphere is actually warming more than the surface because the surface is constrained near the melting point and the atmosphere can get heated, so that's actually ... And we see this in these simulations that that's getting more stable in the summer, so it's actually even inhibiting some sort of cloud-sea ice interaction in the summer. But that's a really good question. And the opposite is true, of course, in the non-summer months, where you have a very warm ocean underlying a rapidly cooling or cold atmosphere.
And just like when you boil water on your stove, that's a recipe for convection, for stirring it up, and for interactions between the surface and the lower atmosphere. Graeme Stephens: Last question. Yes. Speaker 8: Okay. First of all, you've given a convincing argument that natural variability, chaotic variability is large compared to the differences between the ensemble and the observed. Yes.
Speaker 8: If you ask the question, "Is there a misrepresentation of the parameterizations of the physics?" can you sample your datasets in such a way to rule out that hypothesis? Yeah. I think that's a really important question, and one where there's a rule for things like data assimilation, very strongly, right? Because if you can control the large-scale atmospheric circulation, for example, by nudging to observed winds and temperatures, you'll reduce that uncertainty, and you can really focus, for example, on physical process uncertainty associated with, for example, clouds. Speaker 8: But is this an error of the physics that you'll barely predict the future? Hopefully, [inaudible 00:58:51] Certainly, areas that the physics can contribute to spread, and I'm not saying that- Speaker 8: [crosstalk 00:58:58] Yeah? Speaker 8: I'm talking about I mean an error in the trend which is significantly different. There can be an error in the trend, and- Yes. In the forced trend. Speaker 8: In the forced trend, of course.
Yes. Speaker 8: And there's so much variability to the chaos that it would- Right. Speaker 8: It's very hard to determine if there's an error in the physics. Yes. Speaker 8: But you might be able to if you sample the data that you've generated in a way that allows you to do that. I think sampling and data simulation.
So, for examples, [crosstalk 00:59:34] clever sampling, such as the intermittent mask, where you're looking at instantaneous profiles over cloud, over sea ice, and over open water, instantaneous profiles of clouds. That is trying to sample what is the surface influence on the atmosphere, and so that's an example of sort of a strategic sampling that enables you to kind of get around the chaos or average over the chaos so that you isolate the response. And we use that, for example, to try to identify if the physics of cloud-sea ice feedbacks was working well in this climate model. But, yeah. I agree.
It's a really hard problem because we just live in one reality, and that's ... We really need to think carefully about the contributions of the chaos of the system to any comparisons that we're making. Graeme Stephens: Yeah.
Very good. Yeah. Graeme Stephens: So, we're going to finish here, and I'd like to thank Jen, again, for inspiring this forum. It was so inspiring that the NASA panel even took a break so they could listen to the seminar.
Oh, wow, thanks. Graeme Stephens: And they're going to have to work overtime tonight to make up for it, so there you go. Thank you.
Graeme Stephens: Thank you. Audience: Get on back. [crosstalk 01:00:45] Graeme Stephens: Jen is very good. Audience: [crosstalk 01:00:52] This transcript was exported on Dec 18, 2020 - view latest version here.
JPL-20190513-NARRPTf-0001-CCS Distinguished Climate Lecture Series, Un... Transcript by Rev.com Page 1 of 2
2021-01-05