 But this also works, not that not okay. I think it's already obvious that we shifted the program around a little bit, so it's on the board. So I think the key change is that Victoria has a full minute talk to be out the noon session, but we'll start 15 minutes early in the evening, so we have 15 minutes left to go. We have to eat, smoke cigarettes, take a whatever. So we'll start now with Tito Sampson. Yeah, thanks very much for the organizers to give me the opportunity to talk to you today about our work at the Alfred Wigner Institute. So today I'm actually talking about one study that we did which is basically using NWP to assess the influence of the Arctic atmosphere on mid-latitude weather and climate. So with a relaxation method in a controlled way we are doing sensitivity experiments. So basically the questions we would like to answer is, so by how much could weather forecasts in the northern mid-latitudes be improved if we had perfect knowledge of the Arctic, but that actually links also to the question how much influence of the Arctic in what regions actually the Arctic will influence mostly the northern mid-latitudes. And under which large scale circulation conditions is the influence strongest? So basically we are using here an NWP technique to be able to also in a climatological sense determine the influence of the Arctic on the northern mid-latitudes. So for this we have been using the IFS model. We started basically experiments on the first and fifteenth of each month from 1979 to 2012 and we did it for all the seasons as well. So we can also determine the seasonal dependency without and with relaxation applied from 75 north to 90 north. I should say about the relaxation basically what we are doing is every time step of the IFS model which is three quarters of an hour. We are drawing the state of the Arctic atmosphere north of 75 degrees north with a smooth transition actually between 65 and 85 degrees north towards the state of the reanalyzed data. And the reanalysis data are actually the data from ECMWF which we are also using to initialize all these experiments. So that basically we have one set which freely evolves like a normal NWP forecast model and one experiment which is drawn towards the observed or reanalyzed state. So the good thing is in such an NWP setting we are of course able to afford with limited computing power to afford quite a large ensemble. So we have actually 204 start points for each season and we evaluate the difference in mean absolute error of the forecasts. So these curves look certainly quite familiar in fact actually Erland Cullen has shown something similar already kind of how the root mean square error is growing with increasing forecast time. And I think he showed just up to 10 days. I'm showing here up to 14 days which is the length of our experiments. And you can see here you're already going into the saturation meaning that we are already getting towards kind of the climatological error of the model. Or I should say the internal variability of the model and also I suppose of the observed atmosphere. So here I should now say we have the four different seasons as I indicated before and we have a solid line and a dashed line each. The dashed line is actually the relaxed forecasts where we have the information of the Arctic atmosphere from the reanalysis data. And the solid line is the one which is without relaxation. And we are looking here at averages over the whole northern mid-latitudes 40 to 60 degrees north. So as you can see it doesn't do an awful lot basically this relaxation. So you could argue the Arctic influence is not that strong on the northern mid-latitudes. So what one can also see of course what is an important feature that in summer you have less variability than in the other seasons. And yeah that's of course also an important information we come back to this later. So this is just basically the same just the difference between the two forecasts relaxed minus not relaxed. Kind of to see what is actually the forecast improvement averaged over the whole mid-latitudes if we relaxed the Arctic atmosphere towards the reanalysed state. And you see we are indeed just in the order of 5% if we consider the whole mid-latitudes. It's also true if we pick out the European region. But the interesting thing is over northern Asia actually we get quite a strong increase or quite a strong improvement of the forecasts through Arctic relaxation. So that is of course something interesting. Over northern America it's again similar around 5%. So and if we now look just to explain this a little bit if we look here at the just the mean circulation. This is just the mean Z500 field 500 hectopascal geopotential height. And it is basically as a mean as an end also the deviation from the zonal average to see a little bit the wave structure the climatological wave structure in winter and in spring. What we can see over the continents basically we have a climatological deviation from the mean westerly flow from the north with a northerly component. And over the maritime areas we are looking at the opposite kind of more a southerly component in the mean westerly flow. And this is of course very important for the Arctic influence. So now we are looking at summer and autumn basically the same pictures. First of all you see as I already said in summer we have of course actually not only less variability which can be explained by the weaker gradient in general. So even if you have few waves going around they will be with a with a weaker intensity. But also of course you see that that actually the deviation from the zonal average is is less. And then that explains the clearly lower standard deviation and also the lower forecast error in the Z500 field in in the summer season. And in autumn it kind of goes back again to a stronger meridional gradient of the Z500 and also stronger wave patterns. But of course the winter was clearly the strongest here. And so now if we look at the mean absolute error reduction in the Z500 again compared with relaxation and without relaxation. We see that that basically of course very obviously where we apply the relaxation we get a huge reduction in percent of the mean absolute error. This is of course by definition of the experiments so no surprise. What we are interested here is actually what happens in the in the mid latitude ring over 40 to 60 degrees north. And basically we see the strongest reduction of the forecast error here in this area in western Asia, eastern Europe. So here we are actually talking about something around 15 percent which is clearly more than the 5 percent in the average over the whole mid latitudes. And so in the days 4 to 7 the signal is still a little bit more confined but in days 8 to 14 the signal also spreads out to the more southern latitudes. And so in spring you generally still see a similar picture maybe with a little bit less intensity as I explained before which was expected. In summer the whole thing gets more patchy and in autumn we are getting back again to the stronger influence of the Arctic. So because of course it's also important to know what I mean we were talking all the time about Z500 and more than 5 kilometers height above the surface. So how about close to the surface which matters more for society? So actually here the main message is that we are still getting similar features. It's a little bit more patchy though but nevertheless the main message I would say still remains the same also close to the surface. So now I would like to come to another topic. So basically we have now seen what is the mean influence of the Arctic to the northern mid latitudes. But how about if we are now looking at forecasts which have been especially strongly improved compared to the non-relaxed forecasts. So if we apply the relaxation and we get a especially strong improvement of the forecast. So here we are actually looking at composites where we only pick the forecasts with the strongest reduction of the forecast error. And we compare those basically to the ones where we have basically a normal forecast reduction or even a very small forecast reduction. So we are actually using one standard deviation to define this cut off of the forecast improvement. So basically it is of course interesting to know now in what situations happens that the Arctic has strongest influence on the mid latitudes. And as you can see I mean is in a way so we focused actually on different regions but especially over this western Asian Eastern Europe area. We found basically the strongest patterns and that's why I focus here. So it is going back to again also the influence from the barren sea, carousel area which we have heard already about from Paolo and also which has been studied in previous literature. It has been already identified as a key area. So basically when we have an anomalous northerly flow from the Arctic towards western Asia, eastern Europe we get the strongest forecast reductions. This is especially true in winter, in spring and autumn it's kind of also visible a little bit but in summer hardly visible. Again due to the arguments I had given before. So it is of course also interesting what happens to the two meter temperature if we look at the strongly improved forecasts over this western Asia region. So basically what we see is we actually get cold anomalies up to three Kelvin which is quite substantial in these situations. And so these cold anomalies due to basically the circulation anomalies which we had seen before. So basically if you imagine that this anomalous flow goes around here it will of course influence then eastern Europe and central Europe as well. That's why you can see that these cold anomalies actually are not only over western Asia but also over central and eastern Europe. And then of course these are the experiments which have the strongest Arctic influence where basically the relaxation helps the strongest to actually improve the weather forecast. So you can of course draw the conclusion or hypothesize that in such situations the performance of the model may be poor. And you are getting an extremely strong improvement due to the relaxation. So and this is of course going back a little bit to this discussion how about basically can the models really capture these Arctic mid-latitude linkages as well. So basically now I'm coming to the conclusions. We have basically over northwestern Asia we have the strongest forecast improvements and therefore the strongest Arctic northern mid-latitude linkages. And we have also a minor pathway over northern North America. It could be basically there of course when we also included a little bit more the west of northern America. We also included a partition where is a southerly flow. One could also play around with these target areas a little bit more. But in any case there is a secondary pathway also to north from the Arctic to North America. So and as I discussed already maybe we are here finding an indication for poor representation of these northerly flow anomalies in the model. And so also what is interesting what I didn't go into detail during the presentation is we also looked at because of course the Arctic sea ice has been shrinking a lot during the last 34 years. And we were basically interested would we actually see an increased marionality as some studies show like for example Francis at all. But we can actually not detect that the Arctic influence has been increasing during this time period we investigated. Basically because we have the whole set from 1979 to 2012 and we were thinking maybe we get kind of a stronger forecast reduction if the marionality increases in these key areas. But we don't really see that. So thanks very much for your attention.