 So, I'm going to talk about the study with colleagues here at ICTP, Ricardo Parnaty and Fred Kukarski and Tim Stockwell at TCMWF. It's a bit strange for a talk on decadal variability given by a person from an institution that doesn't do decadal studies by mandate, but of course we do a number of things which are related. And so my idea was to see whether we could say something about the decadal variability from basically what we do at TCMWF, we do reanalysis and we do seasonal predictions and longer sets of reforecasts. So just a few slides to start just to summarize where we started from and of course these things have already been discussed, some of them just in the past couple of talks so there's no need to spend too much time. So we have discussed the hiatus in global mean temperature and so, first of all, we can look at it in terms of the decline in the trend of annual mean, global mean temperature. And for example, if you use the Hadley Center, the Hard Crew T4 data set from the IPCC report, you can quote a trend of about 0.26 degrees per decade from 84 to 98 and almost flattening in the following 15 years. However, it's also been reported that this hiatus has a seasonality. You see much stronger if you actually look at the northern hemisphere winter. For example, if you look at the picture on the left hand side from Trenembert et al in nature climate change, you can actually see that actually if you look at the basically cold season in the northern hemisphere, you actually have a slightly negative trend in surface temperature. And even more so, if you concentrate on what happens in the northern extra tropics, then the decrease is even stronger. Now, we have heard a lot of discussions about the role of the PDO or IPO and in particular, we have heard from Dr. Kozaka about this very nice pacemaker experiments, which have actually shown a very good reproduction of the strands and one thing that perhaps she didn't mention was the fact that not only in this period they managed to reproduce a lot of the pattern and the overall amplitude, but also they managed to reproduce at least part of this seasonality with, again, the strongest hiatus being during the northern winter. However, if you actually look at the observations, you can actually see that the seasonality is even stronger than what is actually simulated in this particular experiment. And if you look, for example, at regional levels, you can actually see some discrepancies, which, of course, you may expect because of the chaotic nature of the response in the extra tropics. So one question is, is there a missing ingredient in simulating the seasonality and if you actually think about the global energy budget at the surface, and you do an annual mean, for example, this is a result from era interim, then, of course, what is very noticeable is that the regions where the ocean loses heat to the atmosphere are mainly concentrated in the regions of the western side of the ocean, the regions of the western boundary currents and this actually loses heat mainly, of course, during the northern winter. So a related question is, what is the role of the variability in this region in actually modulating the changes in near surface temperature? So as I said, we don't do the KDAL simulation at tcm.gov, but we produce the analysis. So I looked at some data from the era interim, reanalysis that now covers from 1979 onwards and then I looked at the set of reforecast that we use to calibrate our seasonal forecast system. The latest version is called System 4 and to calibrate this seasonal forecast, we have done reforecasts every month, starting from January 1981 and you can then add the operational forecast and every sort of continuous set. In particular, for if you consider the December, January and February season, we have runs from the 1. of November. In this case, the DJF season will correspond to forecast months 2 to 4 and for this we have 51 member ensembles in the reforecast. So it's quite a large set and basically we used it to estimate the internal variability arising from basically the internal spread within these ensembles. We have sufficiently long records so we can cover basically the latest transition to the hiatus. So we computed trends and we tried, since of course there's the issue of where actually you put the limits of your trend and the issue is that if you put it in 98, as many people do, then it's really the year of a very strong El Nino. So you may really get very large numbers or very small numbers simply because you start from an El Nino year. You get a more consistent estimate of trends if you actually run a 5-year running mean over the temperature first, which is roughly the main period of an answer cycle. So you get something which is more related to the Kedal modes. So we computed these trends from various quantities that you will see from 84 to 98 and then from 98 to 2011 in various areas. Basically an area just excluding the polar cap south to 70 north, the tropical belt to 20 and the northern extra tropics in various regions. And the first thing I wanted to check is this. If we make the hypothesis that the main contribution to the IATOS comes from this natural internal variability of the climate system including the PDO, including other possible, then a main contribution will come basically from a reduction in the heat transfer from the ocean to the atmosphere. So let's now take the, basically, the air entering fluxes and you actually plot, basically, the net heat flux into the atmosphere, which would be basically some of latent and sensible heat, the upwelling, long wave radiation from the surface and the absorption of solar radiation, which actually doesn't play a very big role in winter, then can we see IATOS? And the answer is yes, we can see it. These are plots starting from 1980. The red dots are individual yearly means and the blue dots are five year running means. So you can clearly see that the much stronger positive trend in the fluxes correspond about three watts per meter square per decade and after 1998 this flattens to about 0.6 watts per square meter per decade. And if you convert these numbers into some temperature trends using a simple, basically radiative equilibrium hypothesis, then you get numbers which are not very far from these estimates from the hot-coated data. So there's a consistency between the fluxes and the observed surface temperature. Now these are yearly means if you look at what happens during the northern winter, DJF, this again is 17.0 to 17.0 south and then you can actually see that consistently with the temperature record the trend in the fluxes is actually becoming slightly negative in the last 15 years. Okay, then you can ask where is this signal coming from and you can actually compare these two regions which have roughly the equal geographical size the tropics 20.0 to 20.0 south and the northern extra tropics 20.0 to 17.0. And again you can look at the total heat into the atmosphere from the tropics and again you see the Ayatos, you see it's the effect of dedicated variability in the tropical Pacific. But if you actually look in the northern extra tropics then you see a much stronger signal and a complete flattening of the heating into the atmosphere. In order to cut the story short you can actually find out that most of these signal come from the latent and sensible heat. And what we know is that latent and sensible heat are strongly modulated by the circulation by the intensity of the winds and by the capacity of the northern circulation to what we call there over warm ocean during winter. We wanted to explore this particular aspect and a few years ago we tried to link basically the observational definition that Mike Wallace and collaborators gave of the so-called cold ocean warm land pattern which is basically wave number two pattern that brings warm air over the continents during winter with the prediction of thermal equilibration theory which has been discussed by a number of authors. And basically what this theory says is that you can have basically two states of this wave number two pattern in one phase you actually have the highs over the ocean and therefore you reduce the heat exchanges and you have basically the atmosphere equilibrating to the surface temperature contrast or a force mode which is actually this one where you actually have a strong advection of cold air over the warm ocean you extract a lot of heat from the ocean and then this heat is actually advected over the continent. So we define this cold pattern actually using the wave number two component of the net surface heat flux in this way basically is the net flux out of the ocean in this region minus the flux into the land over the continents and if you actually look at the time series of this particular index over this last 35 years you actually see that in the last 10-15 years there has been a downward trend so we wanted to see whether this was actually related to the hiatus so we first look at how this pattern was simulated in the system four model and so we computed covariances of this index with the number of variables and in particular you can see for example in other intensification of near surface winds over the northern ocean this extracts more latent and sensible heat flux and then this actually turns into warm anomalies over the northern continents. So then what we said is okay probably worry too much about the details now we have this index we can compute it from each ensemble member every year we are 51 ensemble member and then we can somehow composite in this case is actually a weighted average basically those members which have very good representation of the variability of this particular index and those who actually have a very poor representation of this particular index in particular ensemble members do not simulate the variability of the car pattern and then we can construct ensemble means from these weighted averages and we compare what comes out in terms of for example surface temperature and this is what happens so you see here in red dots the time series of the observed car index that you have seen and in green is the time series of the index from the ensemble that reproduces the analysis is called NSA and you see that it fits quite well with the red dots and then we have this N0 which is basically an ensemble that doesn't reproduce almost at all the variability in this now since these are seasonal forecasts that are actually initialized with the same SSD in the first few months the SSD don't change very much so if you actually look at for example the most identical between the two ensembles so these two ensembles have basically the same SSD in the tropics so what they differ is simply the response in the extra tropics in one case you get the response similar to what has happened in the other case you have near zero response now let's look now at the two meter temperature over the northern hemisphere in this two ensemble we plot the two curves so you can see that the strong increase before 98 is well reproduced in both ensemble and then you see a decrease again in both ensemble but if you actually look at the black curve where there is no extra tropical variability you actually have I would say a flattening of the temperature if you actually compute a trend from the one that has the right variability it actually has a decreasing trend and in fact if you plot the difference you can see that just the difference in the extra tropical response accounts for a decrease basically a negative temperature trend over the northern extra tropics of about 0.1 degree per decade so we have talked about this missing 0.2 degrees overall on a global scale if you actually concentrate on the northern hemisphere in winter about 0.1 so half of that can be explained by the extra tropical signal and in fact this estimate it is quite well with the recent estimate by Sapfioti et al and he is here in the audience so I think he will talk about this probably tomorrow now the interesting thing is that well it is interesting to look at these integrated quantities but what about the regional effects and the regional effects are in fact quite strong over the land so this is the change between basically the beginning and the end of the last 15 year period this is the change in geopotential height from the reanalysis this is the difference in the trend between let's say the good ensemble and the bad ensemble again remember this is almost the same SST they only change in terms of extra tropical variability this is the difference in 2 meter temperature in global level you see this very strong warming actually over the Arctic but in general quite a strong cooling over the northern continents in fact if you just average for example over Europe you get actually a cooling of 0.34 degrees now if you look at the basically the pattern which is attributable to the change in this wave number 2 component then you see that this continental cooling is pretty well simulated and in fact for example if you concentrate here over Europe you see that the ensemble with the correct variability reproduces the cooling of 0.49 centigrade the ensemble with no variability in this wave number 2 actually gives a slight warming of 0.21 so at the regional level there is a very strong difference now but another interesting feedback maybe in terms of the feedback on the ocean circulation and we've actually seen that the ensemble with the correct tropical variability actually simulates a response in terms of near surface winds over the Pacific that tend to actually slow down the direct circulation and therefore the subtropical circulation so this may actually be relevant for the mechanism that Ricardo described in his earlier talk about the possibility of a feedback of the isotropical response on to the PDO itself ok? so coming to conclusions so although the recent reduction in global warming trends can be detected in both the tropics and the extra tropics the seasonality is which is strongest in the DJF season if you actually trust in the winter in fluxes you come to the conclusion that it's mostly originated in the northern extra tropics there is still a component from the tropics but the fact that it's so strong during winter to explain it you have to take into account what happens in the extra tropics so we've seen that changes in the phase of this wave number 2 kaul pattern and the associated heat flux anomaly gave a significant contribution to the slowdown in northern hemisphere warming trends during the winter of the last 15 years in many overland and basically by subsampling the different members in the ECM that we have system 40 forecast which have basically the same tropical sst anomaly and very very similar also extra tropical sst but by construction different terms of strong or weak variability then we can account for about 0.1 degree per decade in the IATO so it's about half of what you actually need to be explained and we have also seen this possible feedback on the wind circulation that we are going to explore in the future thank you