 Okay, good morning, exciting session. And I'd like to acknowledge my collaborators in this work. It's led by Adeline Bichet, who you hear from in, I guess, tomorrow. And other collaborators include Lawrence Moudric, who is working with me at the University of Toronto, Leontanet and John Fife. So this is work that I think dovetails nicely with Matt's talk to some extent. And I'd like to kind of go over what is really kind of a simulation protocol, and we're just, we're developing it now. This is work in progress. And so over the course of the talk, I will talk about the use of prescribed SSTs, AMIP type simulations, to predict continental hydroclimate trends, and a pattern scaling-based method that we've developed that we're still working on, that we'd like feedback on to extract the force component of climate change, the long-term global warming. And I'm gonna show you some applications on past recent climate trends, and Adeline will share with you some results related to projected climate. Okay, whoops. That seems, that seems rather, rather abrupt. Yeah, all right. All right, so to start with, I just wanted to take us back to a basic point that if you feed an atmospheric general circulation model, observed oceanic boundary conditions, so observed SSTs, and also throw in sea ice as well, that you can do a very good job at reproducing trends that have been observed over the continental region. So this is a figure taken from a paper by Shun and Sardish Mook showing in the top panel observed temperature trends over land in the second half of the 20th century. And then the ensemble mean from several AMIP simulations where all SSTs are prescribed globally from the CMIP3 generation of models, but these are atmospheric GCMs. So if you, and you can also get the same kind of result or at least part of it if you prescribe just tropical SSTs. So a way to quantify that is using a Taylor diagram, which we're introducing in the last talk, but so a Taylor diagram shows the, on the polar angle here indicates the degree of correlation with some target. So the target here is observed land surface temperature trends, the pattern of that and this is, these are three observational estimates. So the target here is the mean of these three observational estimates. And over here in blue are a bunch of results from AMIP simulations. So this is the CMIP3 generation of models. In green as well, there's ones in which only the tropical SSTs have been prescribed. But if you look at coupled models, you can see that a couple of models over the same period have had a really difficult time at reproducing the pattern of observed trends. And so there's no, this is kind of a ruling out internal variability to some extent as a cause of this disagreement because none of the coupled models from that period were really able to capture it to the extent that any of the ones with prescribed global SSTs. So because of that connection, we know that there's a lot of useful information in the ocean for continental climate. And the dynamical causes as Matt was saying, I'd say it's unclear to me, but I think if you take a long enough time history of ocean SSTs, you can get kind of good agreement with OBS over land. So exploiting that connection is an interesting idea, I think, and so what we're trying to understand is whether we can get additional insight from the observed history of SSTs. Okay, so one of the challenges in this and then one of the interesting points is to try to separate out internal and short-term variability that's driven by anthropogenic forcing from a long-term trend. So what I'm gonna explore here today are two questions about whether we can estimate the long-term variability associated with the global warming, so just taking the global radiative forcing as kind of a target, try to extract that from observations, and then whether we can use that information to gain insight into past trends and maybe use it to predict the force component of climate change in the future. So our method here, it's a somewhat complicated slide, but basically the idea is pattern scaling. So in the top equation here, what we do is we take an SST field or a sea ice field and we decompose it notionally into a long-term, slowly varying part which we call GW for global warming and a short-term part which could include internal variability and also any response to short-term forcing such as anthropogenic aerosols, volcanoes, and ozone, for example, ozone-depleting substances. And the very simple idea here, the simplest idea is that there's a stationary pattern which we call H as a function of X and that stationary pattern is scaled by the global mean temperature, an estimate which we call GFT. And you can make various choices in how to carry out these estimates and I'll show some evidence that we're on the right track, but basically we've done two things and in a publication that's out in Journal of Climate, we estimated GFT from observations, from a long-term observation, SST observations, so that's the global mean surface temperature and in the recent work we're actually using the estimate from CMIT five because we're going into the future. And H of X is just a simple regression, basically a trend pattern that's taken from your observational data set of choice. So in the previous work, we looked at the sensitivity of our results to various estimates of this global mean surface temperature, but basically we hit on a simple, slowly varying, slightly non-linear trend that goes through the global mean SSTs and that's what we're calling GFT. So in this sense, H of X, which is the pattern of SST and CIS is the regression of this curve on observations, its projection. So I claim that this is actually a method that we can quantify to some extent how well we're doing. And what we do is we turn to a large initial condition ensembles. We have a few of those at our disposal to test the idea of whether we are actually extracting the long-term force component of climate change. So we have the ensemble mean of a large initial condition ensemble to use as a kind of filtering of internal variability. And from that, we can compare the patterns of H of X from that ensemble mean to the individual H of X that we get from this method. And this shows the pattern correlation. It's one test we've applied is the pattern correlation between individual H of X is extracted in this model setting to the ensemble mean. And this is testing the sensitivity to model types. So we do it here over a 45 year period with CCSM-4, with CES-M1, the next generation model. Then we extend the period here and you can see that as you increase the period, not surprisingly, you get a better agreement between the estimate and the final result. And we've done a bit more of this in a more recent work where we've extended the period further to 2012. We've done some other adjustments to the method. So we have a sense now that we can capture about 70% of the observed pattern if we believe that there is some correspondence between the coupled model test bed and observations. However, we're not using the coupled model for those patterns. I just wanna emphasize that, that the pattern is coming from observations. Is that clear, by the way? I don't wanna, because this is important. If you don't, it's important to kind of understand it at this stage, all right. So what we actually get then is shown, the most important result I think is shown on the right. So on the left, I'm showing 30 year trends of sea surface temperatures from the recent period, 1980 to 2010, okay? And on the right, I'm showing the, what we're calling SGW, okay? It's a global warming pattern. And this is the long-term trend, basically, in SSTs, predominantly warming, smoother. There's no AMO, there's no PDO signal. It's just taking the 70, the long-term record, this is from the Hurl SSTs. So you might ask what happens when you vary SST data sets, but this is what we're using for convenience for now. And a lot of the structure that we associate with internal modes of variability or with short-term responses to short-lived forces is smoothed out, okay? So we think that this is an interesting pattern. It's a much broader pattern. There's much less structure in the tropical Pacific. And we wanna use this to drive AGCMs and see what happens. So before I show you that, this is the similar pattern from sea ice, okay? So we take the sea ice product that's contained in the Hurl data set, which is conveniently in our model. We're using the NCARA-CAM model, and so this is easy to do. So this is the sea ice that's coherent with that pattern-scaled, so with that scaling factor G of T. And you see that the recent, large sea ice loss that's seen in different seasons, particularly in the summer, Arctic summer, is reduced if you just draw basically a long-term trend line through that rapid sea ice loss. So that's what we're using to force the sea ice in this model. Okay, so now we're gonna apply and practice this pattern of warming and this pattern of sea ice loss. In CAM-5, we run a course resolution, two-degree model. We've done 10 ensembles of two types, okay? So these are 30-year runs. We have an AMIP run, which is observed SSTs, the time-varying history of the SSTs in sea ice. And then we set of runs we call GW, which is our estimate of the forced SST change. And we're using that to force the full AGCM. And we're looking out over the continents at what's happening there. Okay, and so what I'm gonna show you is a few results from these simulations. And we're looking at, we're cherry picking here. We're looking for variables and seasons where there's agreement between the AMIP runs and the observations. And so AMIP runs are not guaranteed to reproduce observations in all cases. So we're gonna focus on those regions. And also, none of the trends we find in these short runs, the 30 years, most of them tend to be insignificant because we tend to focus on mid-latitudes, although I'll show you one tropical result. But if we do these runs over and over again, we find that we often get consistent responses. So in regions that are not shaded in gray and the figures that follow, that means that we have 10 runs and the signal-to-noise in those runs represent it, which is the ratio of the response to the standard deviation of the 10 members is greater than one. Okay, so I'll start with surface temperature. Always a nice reliable field to start with. So this is wintertime surface temperature trends from CRU, from these AMIP runs that we've done and from the global warming pattern we've imposed. So what's interesting to us here is the correspondence between AMIP and CRU. And it looks like the AMIP runs are getting a pretty good match in several of the features to what we've seen in observations. This is not too observationally dependent or observation set dependent. And what you see, a pattern of warming, cooling over Northwestern North America, warming here in the Southeast in the winter. And in this GW run, we're seeing basically that this cooling spot is removed. So this has been reasonably attributed to the PDO trend that I showed in a previous slide. And this GW run helps quantify that. And then in this warming run, you see the warming has shifted into the central Eurasia from where it's been observed more or less. And so we can attribute some of this to internal variability, to the PDO variations. And then of course, cause and effect is hard to get out of these runs, but it is a kind of, I think it does give us insight into a separate part. What you can attribute to short-term variability versus to the long-term inexorable kind of warming that's been going on. Okay, now we're focusing on winter precipitation. This project has actually been largely funded by a Canadian network called CanSICE, where we focus on snow and sea ice variability and modeling in the Canadian setting. And so I'm gonna show you a couple of figures related to that. This is wintertime precipitation, okay? And I think the AMIP runs actually here do a pretty nice job of capturing what we're seeing in observations. And so there, you can get a sense that we can attribute the recent trends to the SSTs, SST variability. And the global warming run that shows very, very weak trends over this period. And there's a little bit of drying that's seen in the west coast of North America. And so the difference here is really highlighting that. So much of what we're saying here is that much of these trends are represent the result of short-term forcing and internal variability. This is a picture of a snow cover fraction. As I said, CanSICE is really focused on that. And so up here, I'm showing a couple of observational estimates of recent trends in snow cover, which is highly observationally uncertain. This is a satellite data simulation-based estimate called globe snow. This is the mirror reanalysis. And you can see there's quite a difference in the pattern. So it's worth showing this before I show you any results from our simulations. So I know there's a lot more gray shading here indicating that there's less agreement among the models as to exactly how the snow cover fraction will respond to these changes. However, what you see is that there's this reduction of snow retreat over Scandinavia, Northern Europe, over this marginal area here. And some of that is also found in the same global warming run. So we can attribute much of this pattern to what is an externally forced signal. And the residual shown here is kind of a very noisy pattern. Now I'm gonna switch, this is a sort of final results slide. This is a global, these are global simulations, and I thought it would be worth sharing this results as well. So if we look over Africa, which is an region, so this is summertime Africa in precipitation, okay? So July, August, September. This recent recovery of precipitation that's been observed over the Sahel is really captured well in the same upfront, which is certainly something that's been seen before. And I think just for, as a point of information, when we put in this global warming pattern in which there's no AMV signal, there's no PDO signal, this is the response over that 30 year period to just that broad warming of SSTs. And we see a drawing in West Africa, okay? And the difference between these patterns is showing you the result of how internal variability and short-term forcing might be compensating what's going on over the very long term with global warming. Okay, so I'm gonna summarize, I'm gonna show you one more slide, and then I'll conclude. So we've tested and extended it, I didn't give credit here to where it was due to, it was an idea that was proposed by Hurling and Hurl and Terrain and others to actually do this as a method of climate prediction in the near term to predict the force component of climate change using this kind of pattern scaling method. And so we thought to test this method first, and we have a couple of papers that we're working on with that. And so what I've tried to show you is that we have a reasonable confidence that we can actually estimate this pattern from observations given a 70 year record of SSTs, we can do a reasonable job. It's a very broad pattern, so we might not need a lot of spatial detail to get the general idea. And it seems like there's a good evidence that what we're seeing in North America, there's a lot of the precipitation changes are related to PDO variability, and that we can also apply this to other regions such as Sub-Saharan Africa, where we're seeing a global warming signal that's counter to the recent trends. A question that you might ask is, well, why didn't we use coupled models to do this? Okay, and it goes back to a point I showed earlier indirectly. Okay, if we try to estimate H of X from coupled models, this is kind of an unsolvable mean of what we tend to find, but we tend to find that, this is H of X estimated directly from Seaman 5, and we tend to get a lot more tropical warming and also a lot of North Atlantic warming that we don't see when we do the same estimate and observation. So the difference between these patterns suggests that we should at least try to do this with observations. And so the idea of near term projection with this is to persist this pattern into the future, and that's what Adeline will be talking about tomorrow. Okay, so the key points are that we can reliably estimate the observed pattern of long-term sea surface temperature response to radiative forcing from global warming. This pattern allows us to attribute the related component of past hydric climate trends, and we're able to use this for future projections, and we're still working on those results. And it's really different from what you'd get if you did the same exercise in current coupled models. So I would just argue that there's a lot to be gained on understanding forced decadal climate variability from looking at existing SST observations. Thank you.