 Okay good afternoon. So this work is a follow-up on the talk Paul gave yesterday on the Pacific session. Only this time I use a pattern scaling to try to look at the next 30 years trend. So as Paul mentioned our work is motivated by the poor capacity of coupled models to reproduce observed trends. So this is illustrated here. It's taken from a paper by Shin and Sardosh Mook. So we see here that the observed trend on the top panel and the coupled model and atmospheric model forced with observed SST and CI's concentration. So from a figure like this it's easy to see that AGCN seems to do a better job of reproducing the observed trends than coupled models. Now in that paper, they dig a little bit into this and they find that part of the reason why this is the case is because coupled models have difficulties to reproduce the observed sea surface temperature and they not only have difficulties to reproduce the surface temperature but it doesn't just come from internal variability. This really points towards biases in ocean models. So one way to get away from this we thought okay what if we could derive a pattern from observed surface temperature, try to have an idea what it could look like in the future time period and use this to force an atmospheric GCM. Can we say something about the data prediction? So of course because we don't really know how internal variability is gonna change in the future we only aim to estimate the global greenhouse gas warming component of sea surface temperature. So we'll call this SGW from now. So a quick reminder on the method Paul already explained a little bit yesterday. So the method we need a few assumptions for this method. So first of all we need to assume that the observed SSTs can be separated between a component that between a global warming component and a residual that includes the internal variability of the ocean but also short-term anthropogenic forcing such as aerosols. And then we assume a spatial temporal decomposition of this SGW. So basically we assume that it can be separated into a time invariant map that is scaled with a temporal pattern. So the map because we don't really we try to get away from the spatial pattern from coupled model we derive this from observation. And the way we do this but the temporal pattern has to be taken from the semi-five models because we want to go into the future. So the way we obtain this map is that we simply regress the observed SSTs onto the red line here. And the red line here is simply a smoothing version of the sea surface temperature taken from global mean annual mean, multimodal mean semi-five models. So what Paul has shown yesterday it stops in 2010 here. So this curve was a little bit more linear than the one you can see here but when you go into the future it's close to linear but it's a little bit more curvy. And so the maps he's shown yesterday tends to be a little bit warmer than what we would see from coupled model but what I will show today I will talk about this player here. Ours tend to be a little bit colder than what you can see in coupled models. So I will basically compare three ensemble of time-varying simulations. So the ensemble we have performed which I will call GW, global warming ensemble here. As 10 members we run it with the atmospheric GCM camp five. It's at two degrees and we use our estimate of global warming sea surface temperature and sea ice concentration to force it. And then I will compare this with the CSM large ensemble. So it has 30 members. It's basically the coupled version of atmospheric GCM. So this time the ocean is interactive and it's one at one degree. And I will compare with semi-five multimodal mean which I took 26 models. So all these ensembles I will show trends for 2010-2040. And they all use RCPA 8.5 radiative forcing. So the map here shows the sea surface temperature pattern we obtain when we do the pattern scaling method. So this is what we use to force a model. I show it as a trend for 2010-2040. So what's interesting about this map is that we see a lot of regional pattern. In particular we see no trend. So yeah this is our estimate of sea surface temperature response to global warming. So from what we can see there is no trend in the very little trend in the Pacific Ocean. We can see a north-south gradient in the Atlantic, a north-south gradient in the Indian Ocean. The some western boundary current seems to be warming a bit faster as well. And we see some kind of cooling here in the southern ocean which I don't quite explain. Then if we compare this with the coupled version of our model, so the CESM large ensemble, we see that the pattern is very different. The first thing we see is that as I mentioned earlier our pattern is generally colder. And this is mostly due to large differences in the Pacific Ocean. So in the in the in the coupled model the Pacific Ocean seems to warm quite a lot. A lot more than in our estimate. In general the northern hemisphere ocean seems to warm a lot more too. There are similarities. So we have a similar north-south gradient in the Atlantic Ocean. And then if I compare with a semi-five multimodal mean, we see similar differences. So the semi-five multimodal mean is not as warm as the CESM large ensemble, but still I mean compared to our estimate both coupled, one coupled model and multimodal mean coupled model show a very washed out warming trend in the CESM temperature. So nothing like a regional pattern that we obtain. Whether which one to trust, I don't have an opinion on this. I think they probably all have a little bit of right and a little bit of wrong. It would be very interesting to investigate a little bit and see which one of these maps we could trust more in which regions and why. So now this is a result from the GW ensemble. So this is a simulated result. It's the ensemble mean. So this is annual surface temperature trend for 2010-2040. And so here what we see is that so there is a big Arctic warming. We see some regional pattern for instance in the US. We see a faster warming in the southern part of the US. The Canadian islands also seems to warm quite a lot. There is a warming in the Indian Ocean. But except from that it's we see some regional pattern but it's not striking. If I compare this map with the same from the CESM large ensemble, again our estimate seems to be globally colder. Main differences lie in the southern ocean. There seems to be a lot more warming in the coupled model. Apart from that, except from the difference in absolute temperature, I wouldn't say there is a huge difference in the regional pattern between these two maps. There are some but it's not striking to me. Then if I look at the same in five multimodal mean, it's kind of a similar story. Again the same in five multimodal mean and the CESM large ensemble tend to look like each other a lot more than this two map look like each other. So this is just the atmospheric model and this is a coupled version of this. So it wouldn't have been surprising if these two were looking like each other. Then for precipitation, here we see more interesting regional features. So there are some noticable regional pattern in particular in the southeastern US. It seems to get drier. The tropical western Africa also seems to get drier and in India. Everywhere else we get an increase in precipitation, in particular in the southeastern Asia, eastern Australia and eastern Africa. Comparing this with the coupled model, here we see quite a lot of difference. The coupled model seems to show mostly a wetting in all the northern hemisphere. So here I would say that the difference in the SST pattern seems to matter a lot more for regional pattern in precipitation than it seems to do for the land temperature. Some similarities nevertheless as in the eastern Africa and southeastern Asia. For the semi-five ensemble, it's a similar story. As for the large ensemble, it shows mostly an increase in precipitation in the northern hemisphere or overland, which differs quite a lot with our estimate. At last, I would like to show some results for spring snow covered trends at high northern latitude. This is a snow-water equivalence. It shows snow depth. Here our estimate is that we see a snow melting in Scandinavia and a snow melting in the northeastern US. We don't see anything else in the other regions. If I compare with the coupled version of this model, the snow melting in Scandinavia seems to be here as well. So this seems to be from the global warming part of sea surface temperature. The snow melting in this part of the US is also reproduced, but we have a difference in Alaska, whereas our ensemble shows increasing snow as a decrease in the coupled model. For semi-five again, it looks very much like the CSM large ensemble and a little bit different from the global warming ensemble, but not that much. To summarize, we used a simple method to estimate the global warming component of future sea surface temperature and sea ice concentration. By using this to force an atmospheric GCM, we could get some insight into near term prediction. I have shown that the SST pattern we get shows some differences with the SST pattern one gets from coupled model. In particular, we see no warming in the Pacific Ocean and we see some north south gradient in the Indian Ocean and in the Atlantic Ocean. But mostly we see a lot of regional pattern which we do not see in coupled model. When we look at the impact on this on land climate, on the continental climate, we see small differences in the land temperature trends. They are generally colder, but to my opinion, the regional patterns are not so different. On the other hand, we see a lot more differences for the precipitation trend, in particular some drying area which do not appear in the coupled models. Again, for the spring snow trend at high latitude, there were some differences but not huge. And one last point I would like to make is that it really seems that the CSM large ensemble to the one coupled model as compared to the multi-model mean from CMIP5, they seem to look like each other quite a lot and they seem both to differ from our estimate. I forgot to mention that this is very preliminary, so I welcome any input on this work. I'm only starting to look at this so please give your comments. I have a couple of questions for you to work. Can we explain some of the patterns we see in the precipitation trend from the pattern we see in the CSF temperature trend? Also, what is the role of a direct relative forcing? And maybe most importantly, does this kind of experiment give any additional information when we want to know a little bit more about digital prediction? Can we say something more than seeing this experiment simply like a prescribed CSF temperature experiment? And I also would like to raise the question of linearity, so Paul has shown yesterday similar maps but for the past 30 years with the same experiment and I'm wondering, I didn't have time to look at it so much, but I'm wondering if the map he has shown look like the map I have for the future. Basically, does the continental climate respond linearly to the same to the same surface temperature? So this is something I'd like to dig a little bit into. And finally, I'll emphasize a little bit more about the limits of our approach. It only makes sense in regions that are strongly influenced by the ocean, in the other regions. Of course, it makes less sense. There are limits with our estimate of CSF temperature and CI's global warming component. So as I said earlier, it does not represent the totality of the anthropogenic forcing. And of course, it depends on the observation we choose to derive this pattern. And we have seen already that observations do not agree all the time in these historical periods. So our results will be very dependent on which observation we choose. And of course, it's also model dependent. Although as I've pointed out, it seems that the coupled model agree with each other more than they agree with our ensemble. That's it. Thank you.