 So last talk of the session, so I'll try and keep it snappy. So probably gonna go back a few talks to what I'm gonna be sort of basing my stuff on. So what we've been trying to do at IPSL is take a systematic approach and a perfect model framework to pull about the processes contributing to the PDO. We're actually in a broader sense, Pacific Decadal variability, but this talk's gonna focus more on the PDO at this stage, which is what we've been looking at. That's the job. Well, there you go. So in a broader context, Pacific Decadal variability can really manifest itself in two regions. You can have tropical Decadal variability, where you have low frequency variability in the tropics, kind of similar to what Ricardo was talking about earlier, or EDV. As we've heard from Matt earlier, you can have your low frequency variability over the North Pacific, or often referred to as the Pacific Decadal oscillation, or the base and white signature, the IPO. With obviously the PDO, which is what I'm gonna talk about, focus more on my talk on here, pertaining to the leading EOF of the monthly SSTN only. So this is basically what Matt was talking about earlier this morning. Matt left off with this figure, which was where I kind of want to start off, because this has been a little bit of a nice little model that we wanted to tease out as far as the processes contributing to the PDO. So you can see from the figure, we have the oninho influencing the allusion low through the atmospheric bridge, that then has an impact on the PDO as my mechanism. You also have internal stochastic variability within the mid-latitude atmosphere, which has an impact also on the PDO. Then you also have the Rosby wave, potentially feeding back on the PDO, and the reemergence mechanism. Matt sort of went into detail on all of these earlier this morning, so I won't go into on too much. So I guess, and as Ricardo was pointing out, your low frequency modulation in both regions can come from a combination of both local processes and remote forcing. So what we're trying to aim to do at the moment, and then what I'm gonna present today is to determine how much of this low frequency variability that we see in the extra tropics or the PDO can be attributed to tropical forcing. So how much of this is coming from the tropics at this stage? As a follow-on that we're planning to do, we also want to look into how much of this extra tropical variability in the North Pacific is potentially having an impact on low frequency variability in the tropics, but this is sort of a second stage of the project. I should outline at the start, this is what I'm presenting is largely a work in progress as well, so. And obviously, like anything, one of the limitations with studying decade or variability is the short observational record, which makes it really hard to really pull out and understand the interplay between these processes. So one thing that's been a nice development in recent years is the efficiency of a couple of global climate models. So now we can do these multi-centennial integrations with relatively little computational expense. So they provide an invaluable tool to be able to really understand potentially what the physical processes are that are contributing on decade or multi-decade or even centennial timescales. So that's the essence of what we're doing here. And so what we're aiming to do with the model is similar to what other people have been doing previously, doing a system of partial coupling, partial forcing experiments where we force regions to shut off certain processes pertaining to the PDO. So isolating the processes to highlight what it is we want to see. Or their relative contribution, as my slide points out. So what the model that I'm using here is the IPSL CGCM, the fully-coupled CGCM. I'm using the latest version, which has been developed for the new CMIP incarnation. So it's IPSL CM6-beta VLR, it's a low-resolution model. It's essentially the same physics as the 5A, so for people that want to pull apart into model differences, it follows those lines. The added benefit that we have at the moment is it's gone from being able to simulate six years in one day to 40 years in one day. So it's actually very, very efficient for its flaws. It provides a really nice tool to be able to do long integrations. So all the simulations that I'm going to be showing today, or the result of the simulation, comes from a pre-industrial control forcing. So at this stage, all we're interested is the internal variability of the model. There's no external forcing, so we're not considering external forcing. And we're working in a perfect model framework, so we're working to the model's own climatology, and we're not trying to recreate historical periods or 20th century records or any of this sort of stuff. We're purely looking at the processes inherent within the model. So we allow the model to spin up for 50 years, it seems to be, to get rid of that initial drift. And then we focus on a 250-year period or what I'm going to present today is concerning a 250-year period. One thing that needs to be reassessed, and I'll show a caveat at the end, is that at present the climatology that I'll be showing in the next few slides is based on a period that's actually 200 to 1,200 a year in the model simulation, which this needs some amendment, but I'll go into that as we go through it. So just to give a broad scope idea on how the model performs to begin with in the fully coupled simulation. So this, on the left-hand side, you have the HADIS. This is the figure that we've seen continually for the period 1901 to 2010, just your first EOF, regressed onto the regression between SST onto your first EOF of North Pacific. Then on the right-hand side, we have basically the equivalent plot for the 250-year period for the IPSLCM6 model. I mean, one obvious thing that we can see there is it has a slightly stronger variability in the North Pacific. The Wikipedia poll is also, the maximum video cooling is also shifted slightly west in the simulation. But overall, as far as the patterns and the variants explained, it's not doing too bad of a job. It seems like a, it's looking at that, it seems like a realistic tool to potentially be using. So the next question that came up from supervisors as well was how does the teleconnection patterns look in the model? So if we look here, this is a monthly stratified but seasonal values. So we start off with DJF at the top. On the left-hand side, we have the NCEP's SLP, regressed onto the HADIS ST. And as you go down, you see the maximum, the maximum teleconnection occurring in DJF. And then slowly reducing as you come into March, April, May. One thing with the model is it seems that the peak in this teleconnection pattern occurs more towards, probably in JFM, so late March. So there is a slight delay in the teleconnection pattern compared to the observations. But there's still a reasonably strong teleconnection there during the winter season. And so why might this be, it would be the obvious question, why is it that you have a delayed teleconnection pattern? Well, it can be attributed to seasonality events so in the model is the main factor, I mean. So at the top there, we have the seasonal breakdown of the ENSO, so the Tropical EOF-1, regressed onto SST. So you can see a peak SON, DJF, which we all know. With the model, this peak tends to occur more towards the springtime, which is one of, this is one of the caveats of the model that the seasonal phase locking is a little off. But insofar as looking at the teleconnection patterns, through DJF to MAM, there is still a reasonably strong tropical signal during the winter season. It appears more that the model has quite a strong tropical signal throughout the year rather than it just being isolated to DJF. And as Matt pointed out this morning, this signal has definitely shifted to the west as well. So even though the ENSO peaks in MAM, there's still a reasonably strong tropical signature in SST during the DJF, so we'll press on. I mean, one thing that could be considered in all this is as Tom mentioned yesterday as well, is some sort of artificial adjustment to correct the seasonality in the simulation before doing PDO simulations. Okay, so what basically I'm gonna show today as a first part of a series of experiments that we're conducting. So the first thing we wanted to do is we're hoping to look at what's the influence of the tropical Pacific Ocean on the PDO. So what we do is we take the model's own climatology in the tropical Pacific, so you can sort of see the, oh, going back forward there. And what we do is we wanna constrain the SST variability in the tropics, essentially to kill your ENSO in the tropics so that we can determine the relative incidence. Basically we wanna see this part of the puzzle. So if we kill our ENSO, how much are we killing that teleconnection in the atmospheric bridge to the illusion low, which is then potentially going on to the PDO? And the way that we do this is we nudge the tropical SSTs towards, like I say, the climatology through the heat flux term. So we, as you can see there you have your SST, you subtract your climatological SST and then you have a little restoring coefficient as well because the ocean's not seeing the atmosphere, the atmosphere's not seeing the ocean. So yeah, that's just an outline as the first set of experiments that we're sort of conducting here, so to go on. So effectively what we have here on the top is we have a zonal section through the tropics, two degrees south to two degrees north in the fully coupled simulation. So these are both for the same time period in the simulation as well. The second one is the one where we nudge towards our climatological SST, so effectively killing your ENSO. And this is apparent, so this is your temperature variability with depth, and so you can basically see killing your thermoclimate response, especially in the western tropical Pacific. And if you look at, if you do the difference plot, you can see there that you're effectively killing your ENSO on the upper layer. So if we then go back to a similar plot to what I showed before, so on the left we have our PDO index regressed onto SST. And then on the right we have the same period, but with our nudging of the SST and the tropics. So basically we've, and so what you can see here is effectively we've killed our ENSO in the model. You've also killed reduced a large percent, or 40% of your PDO signal in the extra tropics. So by killing your ENSO, you've significantly reduced it. Another feature of it is you can see that you sort of lose that inter hemispheric IPO pattern. So there's less of a pattern appearing now in the southern hemisphere. There's also a reduced variability in SST in the Indian Ocean, but it doesn't seem to be that much of an impact in the Atlantic from killing the ENSO, so that was all about that. And so this is, like I said earlier, there's a little bit of a caveat as to the climatology being used as well, which is something that I'll adjust when I go back to the lab. What we have here is we have the fully coupled simulation. This is just the mean fields, the mean SST fields along the equatorial Pacific, the nudged fields. So there is a very, very small change in the mean state. So you're getting a change in the variability associated with killing the ENSO, but your mean state doesn't show a large shift. But if you look at the difference plot, there is some sort of difference there at the bottom of the mixed layer. And that could be a result of using a climatology that's not directly over the period that we're concerning ourselves with. So this is the first in a series of systematic studies that we're hoping to complete over the next couple of months. So basically just at the moment, so the present conclusions that we have from this work is the IPSO reproduces the IPO variability reasonably well and this is despite the fact that there's a seasonal shift in the teleconnection patterns. So we've set up some sensitivity experiments to explore this interaction between the tropics and the extra tropics using a perfect model framework. And ENSO appears to contribute to about 40% of the SST variability that we're observing in the PDO and more in the western pole, but it seems that it has less of an influence over that eastern pole. And it also appears that constraining the tropical variability also produces only a slight change in the tropical mean state. So at the moment, we're sort of going up, one more point of course. And so the potential influence of mid-latitude variability on ENSO is also the focus of future experiments that we're planning on conducting. So at the moment, we're currently going on to do a series of forced experiments where we're forcing the extra tropics with prescribed teleconnection patterns associated with ENSO through the heat flux terms, fresh water flux and the momentum fluxes. And again, this is something that is we're just currently working on. So if anyone has any ideas or if anyone wants to discuss potentially what this could mean, then come see me and we'll have a chat. So yeah, I'll leave it at that. Keep it sharp.