 Great. So our next speaker would be Fredrik Wieter from ECMW. Fredrik also gave a talk during the colloquium. Thanks for that, Fredrik. When are we are ready? Okay, I would share my screen. Can you see it? Yeah, I can see it. Okay, so thank you for the invitation. And so we give no presentation on focusing on MGO-T connections in ECMW. I am talking almost the other T connections over the Euro-Atlantic weather regions. So this talk will have three parts. First, I will mention the impact of the MGO on Euro-Atlantic weather regions. Then the presentation of MGO on ECMW connections in S2S models. And the third part will be about the modulation of MGO-T connections by Enzo. So first part, I mean the impact of MGO on the weather region, Euro-Atlantic weather regions. So that's quite a long story. I mean, there has been many publications on the topic. And from Ferranti, Tino Sardesmer and Minetal. And this paper by Kassou also showed very nicely this impact on the four weather regions, so negative and Euro, positive and Euro-Atlantic, which can be very blocking. So this paper, Kassou looked at the analysis data for each day where the MGO in phase one, then he looked at the frequency of those weather regions. And then he draws this sketch where the shaded area means there is a significant impact. And it shows here the evolution of the population of those weather regions relative to the climate from day one to day 15 following an MGO from phase one to phase eight. So the main conclusion here is that the impact of the MGO is mostly on the NEO. And following an MGO in phase three, we tend to get positive NEO about 10 to 15 days later. And following an MGO in phase six, seven, we tend to get a negative NEO, higher-poverty or negative NEO to 10 to 15 days later. So this is quite an important impact and an important source of predictability for an ocean hemisphere and therefore it's quite very important for S2S model to be able not only to represent the MGO correctly, but also to be able to represent this link accurately. So to evaluate this as the case, this slide shows the teleconnection, which are here measured as a composite of that 500 sweep and toss after an MGO in phase three, that's for the extended winter for November to March. And the top left panel represents that form of the composite in Aeronterium. The contour is very clear. North America is on the left, Europe is on the bottom right. And then we have this very typical positive NEO signal in Aeronterium, which if we use an NEO index, if we project it into an NEO index, which has a mean of zero and star deviation of one, has a value of 0.5. So it's about our first standard deviation. And the bottom panel represents the same composites, but in the various models from the WAP, WCAP, S2S database. So showing that most of them represent the pattern relatively well. All of them go to positive NEO. But the amplitude is strongly underestimated by all the models. But otherwise the models have been ordered as a function of their horizontal resolution, so showing that, so there is a feeling here that the higher resolution models tend to get stronger teleconnections than the weaker, than the lower resolution models. If we look at individual inseparable member, which has the same population of the Aeronterium, not a single inseparable member from all those models get close to 0.5 to 2, this value. So which means that there is clearly a significant underestimation of those state connections in the S2S models, which means that the model are missing part of the source of predictability. So the positive view here is that we can do better. There is a room for improvement in the S2S forecast scale over Europe and North America if we are able to improve those state connections. So then the next question, another plot I wanted to show is the evolution of this state connection as a function of lead time for longer lead times. Here we are looking at seasonal forecast, C5 at SMWF. And it shows that top panel is for three panels after an M joint phase three, and bottom panel three panels after an M joint phase seven. And it shows that as a lead time increase, those state connections get weaker and weaker over the electronic sector and same for M joint phase seven. So by month four, the state connection in the model start to be really, really weak. By the way, those state connections are computed from the MGO produced by the model, not from the observant. So it means that the connection between the MGO and the NEO start to be lost after a few months of integrations. So the main question is then where does this error come from? Well, the main suspect is of course the representation of the MGO itself. We know the model have improved. There are much better representation of the MGO than they used to have 20 years ago, but there are still a lot of issues. And if we look for example the error in the representation of the amplitude of the MGO, which is relative to error in Terim. So this is the evolution of this error from day one to day 32. This shows that if we are negative means that you have a weaker MGO compared to error in Terim. And this plot shows that all the model, particularly most of the models, tends to have a two-week MGO. In the case of CMVF, we are in the middle of the pack with an MGO which is about 20% two-week by day 30. In term of error in the propagation of the MGO, this represents the MGO phase zero relative once again to error in Terim. So positive value means that MGO is too fast, negative value means too slow. And we can see that some models tend to get an MGO starting too fast and then slowly down on becoming progressively too slow. What is the question with amplitude is that the error is already quite strong already by day one. So it's really an error that is very short, that is already apparent in the short range. So it's likely that those error must have an impact on the TLA connections. Intuitively, if you have a stronger MGO, we expect stronger TLA connections. And indeed, I mean if we look at the TLA connection in the CMVF model, and if we consider all the MGO events which are characterized by an amplitude larger than one star deviations, using the willer and the IMS index. So this is the TLA connection we get. And if we restrict it to the stronger MGOs, one which has an amplitude larger than 1.5 star deviations, we tend to get stronger TLA connection over the URA connected sector. And if we go to more two star deviations, we get even stronger. So if the model underestimates the MGO amplitude, it's clear that they should have an impact and they should partially explain why the MGO TLA connections are too weak in the model. But the question is that the whole story. And to check that, we did an experiment where we nudged the tropics to what there are in Terim. So in this experiment, we have a perfect MGO, representation of the MGO. And the left panel here shows those TLA connections are in Terim. The middle represents the control experiment. And the right one is the tropical relaxation experiment. So we see that the MGO TLA connections are a bit stronger, better defined. And the stronger positive MGO. But we are still very far from what we get in Terim. So this suggests that those error in the TLA connections are likely to come from, don't originate only from the tropics, but may come from errors in the Northern Isratropics. So a possible clue of that may come actually from a recent study, actually there has been several recent studies. For example, here from Li et al in 2019. We looked at the modulation of the MGO TLA connections by Enzo. So they say here they did the same study as Casu. So that's exactly the same plot as in Casu, using a different analysis on different periods, but the conclusion is the same. And the right panel here is if we consider any years on bottom, if you consider any line in your years. So the TLA connections, the project impact on the MGO is much stronger during any new years than during the line in your years. So this impact of the MGO is much, much quicker. And that's one of the mechanisms that should be quite important for model to represent accurately. So we looked at all the ECMVF model. ECMVF model was able to reproduce this modulation. So we looked at the differential Gaussian, we visited ECMVF couple of years last year. And here we look at the TLA connections, so 10 to 15 days after an MGO in phase one to phase eight. And we look at the NEO index from this composite 500 composites. So the black bars represent the error interim. So we get this positive NEO after phase three and negative NEO after phase seven. So this confirm what I just said that the model will produce fairly well modulation of the NEO by the MGO, but much, much bigger amplitude. And the right panel represents when we have any new learning ideas. So as in the previous paper, we get a much, much stronger amplitude of the NEO index after an MGO in phase during any new years and during learning years. But this is something that the model doesn't reproduce well. Actually, even for phase seven and eight, we tend to get the opposite signal. So it seems that the model fails to reproduce this modulation by NEO, which is quite a big problem because it means that during learning years, well, the model may overestimate the TLA connections or during any new years, the model may underestimate them. So an explanation for this impact of the NEO on NEO, so may come from the impact of NEO on the MGO itself. But another hypothesis is also the impact of NEO on the subtypical jet stream, which is extend much more to the east during any new years and retreat more to the west during any new years. So the jet stream plays a key role as a pathway for MGO, Rosby-Web, for MGO-generated Rosby-Web to propagate to the Atlantic. So errors in the jet stream representation may have some severe impacts also on the representation on TLA connections. So now if we look at the quality of the jet stream in the model, the climatology of jet stream, this is a topical jet. So we look at the climat of 300 zonal winds at 300 hectopascals, a different lead time, 10 to 7, 8 to 14, with 3, with 4. The orange bar represents the history extension in era 5 on black bar in the week 1, 2, 3, 4, and we see as the lead time increases, the model tends to push the jet stream much, much more to the west, increasingly to the west, but there is an error of about 15 degrees by week 4. And so this is, this may also affect the TLA connections. And if we look at the composite once again on that 500 hectopascals, so 3 penta after MGOs in phase 3. The left panel is when the suitable jet is more to the west, or it is when it's more to the east. We see that when it's more to the east, we get much stronger the connection of the Atlantic than when it's at the west, which is consistent with era 5. So this means that this bias we have in the jet stream may be so detrimental, maybe another piece of the puzzle to explain those two weak TLA connections. So this one actually is this plot here show actually the difference of the wind, 300 hectopascals. So the difference between the linear linear so we get this very strong negative anomalies in the central Pacific. One of the biases in January. So suggesting that the model there is a sort of a drifting towards a linear state. So we come from errors in the SSDs, I mean, if we we get the same errors if we force the model with observe SSDs. So it's a purely atmospheric error. That's the set model errors. And so as I said, so the model is sort of drifting towards linear and as we have seen during a linear years. The TLA connections tend to be richer than during any new years. Yeah, so I'm most finished. So here's the system with the same life model and then the question is, so what about those are models and if we look at other model from this to his database. We find that actually most of the model share the same patterns actually you can met a face in Canada or the same error. Meteor France has a very strong error to GMA and separate opposite error. But the majority of model also seems to share those this error with those are just twins of which is located too much to the West. We just conclude here. So to say that the window from literature from previous studies at the MGO has a strong impact on the NEO. So the stress models tend to underestimate this impact of the MGO and you're working with the regimes. So it's a very free forecast or felt to capture the strong addition of MGO to connections by and so. And those those are also maybe linked to the to the location of the pacific or so typical jet in the model, which the model has a tendency to to to move too much towards the West, and to other sort of learning aspect. So that's it and we stop here. Yeah, thanks a lot. Any questions for Frederick. I had one Frederick while we waste for others. So you're the teleconnections patterns you should they also resemble the P and a pattern and the pacific right. So how much does the internal variability P and a more compared to there's also studies showing that the end so projects on to like some part of the predictive mode of the P and a. How much does the P and a pattern modulate this teleconnection of El Nino versus La Nina. And the MGO and a teleconnection has Yes, I haven't looked too much at the P and a but yes I mean the P and a can can also impact I mean it's difficult to this entangle everything because it's a P and a can be also affected by and so so so I don't have an answer to these questions but I think it's not a simple simple to establish that Zane has a question on the chat Zane, if you would like to unmute and ask your question. Sure, thanks Anish. Thanks. That was interesting Frederick I wonder are you guys planning, you know you could do some sort of similar nudging experiment where you maybe only nudge around the, the subtropical jet or something or do you have sort of plans to do other experiments. Absolutely I mean we are doing we have done really quite a lot of experiments to try to understand where these bikes come from. The first idea was to come from the tropics itself. We know that the MGO modulates those is the project but if we relax the tropics actually we get the same error. It seems to come more from a hero in the stratropics so that's something we, so we don't just know the stratosphere there was no impact. But if we seem to if we relax the high latitude it seems to have a more impact so some impact but once again it's a worker ongoing work. Thanks. The problem of doing is that we impose a bit of solution but that's Thanks Zane. Hey me, you have a question would you like to unmute and ask. Sure, friend are you sure that the MGO amplitude reduces in day one, and I have also similar results but do, is there any plan in the same dev to improve this, this quick decay of MGO signal. We are starting, we are we are looking at it for for from day one actually we have a new person actually we started to look at that with Mark Woodwell I mean they are looking at the tendency during the first first first day to try to disent and go to try to understand where this error in day one come from I think if we try if we if we can understand there what happened in day one, I think it will help a lot to understand this error in the longer range. But yes it's it's a once again yeah there is work ongoing currently to focus more to the very very short range. Thanks. Thanks. Hi Lynn, you have a question. Next. Yes. Hi Frederick. Let's talk. There are quite a few recent studies show that the MGO teleconnection there is a stratosphere pathway. So I'm wondering if you see there's some models they have high top better resolved stratosphere as a better MGO teleconnection. That's a good question I mean I mean we could come back to to to this plot and so here it was ordered as a function of the horizontal resolution. So we could do the same with a high top. I think it's difficult because usually those models which have low resolution are very low top on the model here we perform the best as usually the one which has the highest top. So, where does it come from. So I think to answer your question I mean we're not sure we can address this question directly from the database itself. So I think we need to do sensitive experiments where you where you increase or decrease the strategic resolution. Thank you. Thanks. Thanks again Frederick. Really interesting talks.