 Rhaid i'w pwysig yw yng nghydfyrddol yw'r newydd, o'ch siŵr yma, sy'n gofyn i'w tref mwy pwysig. Rwyf i'n ceisio y cwrdd a'i dweud yn gwneud i'w cael prynyddau a writing o symud o ran o'u cyflwyno. Rwyf i'n ceisio efallai ar gyfer'r gweithreffod, ac â wneud i wneud i weithificio'r gwybu. OK. Let's do this. So, first I want to talk about some work that was led by a guy from the Met Office Halley Centre called Chris Roberts, involved myself and Matt Palmer briefly discussed it earlier in the week and Doug McNeill. What we were interested in was the question of what is the likelihood of the current hiatus event and what is the likelihood that it may continue so what we did actually a very kind of simple analysis we first estimated the global mean temperature trend due to the force component so anthropogenic and natural forcings and we did this by averaging all the C-MIT-5 simulations together to get a kind of smooth estimate of the force response it turns out that perhaps the C-MIT models may be slightly over responding to the forcing but we can take that into account by scaling our sort of background trend up and down. We then took the control runs and detrended them and then sampled periods of high and so negative sample periods of negative trends in global mean temperature from these control runs and kind of added them on top of the force response. By using control runs we get a bigger sample size so we generate kind of if you like a large number of synthetic ensembles by adding the control run variability with the correct phase to the force response and then we estimate the probability of different lengths of hiatus events. We also sub-selected some of the models based on their ability to simulate intranial variability that didn't make much difference so we estimate the probability and occurrence of hiatus events of different periods and of potential what we call surge or accelerated warming events following on from a hiatus and we can also look of course in the models at what's going on in these events. So here is the red is just the different observational records, different estimates of the observed global mean temperature. The grey lines are the individual C-MIT-5 ensembles and the blue lines are those models for which there were multiple ensemble members you could make so this is just really for display purposes. We made an average of all these to get the force response and the force response we estimate to be about 0.2 degrees C per decade so in order to overcome 0.2 degrees C per decade you must have a natural variability trend of minus 0.2 degrees C per decade. So what these figures show I'll take a bit of time to explain them so on the x axis is the trend length and on the y axis is the magnitude of the trend so this is in sampling all these different control runs so for example a trend length of 10 years and a trend magnitude of minus 0.2 degrees C per decade i.e. what we need to have to overcome the force warming trend and you read off here these are contours of frequency or probability and you see if I look on my screen the probability of a 10 year hiatus is about 10%. So we wouldn't be surprised given a force response of 0.2 degrees C per decade we wouldn't be surprised to see a hiatus event that lasted 10 years in length. For a 20 year hiatus event the probability is less than 1% so we might think that it's be extremely unlikely that we would see a 20 year hiatus event so maybe we've had about a 15 year event which is maybe about 5% probability. One interesting thing you can do is to say okay you know the probability of a 20 year hiatus event is really quite small however if we are already in a hiatus event so this is okay so if for example we have already experienced 10 years or indeed 15 years of a hiatus event what's the probability that it will continue and this turns out to be higher than just the raw probability of a 20 year hiatus event because if you're in the kind of low point of an oscillation or some kind of red noise process it will take some time to get out of it so it turns out that the probability of a continued hiatus of another further five years of a 0.2 degrees C is about 15% so much larger than the less than 1% we see here so if you're in an event it's harder to get out. What we also did was look at some of the kind of processes that were going on and it turns out that a hiatus event is associated with a transport of heat from the upper ocean once that's say 100 meters into the deep ocean and if you have a reversal or a surge then that heat gets transported from the deep ocean into the upper ocean and the net total earth system energy is kind of round about zero for these things so that in these hiatus events that we identified in the control runs there was no kind of energy input or output from the system it's just a reorganisation of heat content within the ocean and I should say that the majority of the hiatus events that we found in the control runs were of Pacific origin greater than 90% although we did find some which were of Atlantic origin and just a couple which involved heat exchanges in the southern ocean so it is possible in the models at least to see hiatus events which arise out of different ocean basins. This is the PDF conditioned on already having a 10 year hiatus period and it turns out that the probability of the hiatus continuing is about 14-15% as I showed on the previous graph but the probability of an accelerated warming or a surge is about 50% so and the patterns of SSTs that you see during hiatus this is like a continued five year hiatus event this is a five year surge or accelerated warming event look like this kind of PDO pattern although they're not quite a mirror image of each other whether that sampling or not I guess Chris has done some sort of statistical test on these. Okay, so in summary if we focus on natural variability as the cause of the hiatus and assuming an expected force response of 0.2 degrees C per decade the probability of a variability driven hiatus is about 10% but less than 1% for 20 years but if we're in a hiatus event then the probability of it continuing is much larger so slightly different numbers here and an accelerated warming is more likely than not which means 50%. Okay, so following on from that and following on from the meeting that we had in Aspen in the summer John Fife was interested in this problem too but you know we thought we could actually do a bit better by taking into account what the current state of the ocean atmosphere system is because you know we have the ability to do seasonal forecasting so what he's done here in this work which is as yet unpublished it's prepared in the summer this is if we focus here on the global mean temperature anomaly and this is from 1980 up to the present day and tacked on here in red is the Canadian seasonal forecasting system which actually runs out to a year initially initialized in the summer the scale on the x-axis changes a bit here so be wary of these plots but what you see not surprisingly is this big El Nino event that's happening at the moment and following the El Nino event a kind of rather rapid decline but not declining to below this kind of plateau period here we can take this kind of idea further and what what John did was to they have a large ensemble of canni SM2 historical type runs so just free running experiments but you can kind of it has a pretty good Enso cycle so you can select Enso events like this from the from this kind of database of years of historical simulation and then you can work out what the kind of trend in the following years is and there are a number of lines on this but the crucial aspect is this kind of light blue shading which hasn't come out very well and this blue blob with the error bars here which shows the global average temperature sort of 2016 to 2017 so this is kind of extending the seasonal forecast using an analog type approach and you see that even in a couple of years time the temperature doesn't get anywhere near the kind of twin 2000 to 2014 mean we were trying to proceed even to push this further he wasn't very confident in doing that but so you know maybe this is one way of assessing the likelihood of continued but it looks like you know the idea of diving back into a kind of hiatus type climate is is pretty small okay so that oh yeah and just to say this is the sort of slightly updated El Nino forecast from the now from the Met Office forecast system which again shows this you know we're just about there in terms of the current Enso and as I said large El Nino events have a kind of tendency to kill themselves off so we get this kind of rapid word you know back to neutral conditions quite rapidly next year although if you look at Nino 4 that the forecast suggests that the Nino 4 region the kind of central to West Pacific remains warm and so that's indicative of perhaps a longer you know transition to like more of a positive PDO type phase but again it's only a seasonal forecast so we're not really going far enough okay okay the the the the this is the second bit of the talk is really I can perhaps go through this quite quickly because it kind of follows on or it backs up what Scott said this morning in his in his talk he was talking about the trends over the you know from about 1979 you know where we have kind of good observations at beginning of the satellite record he was talking about the the strengths of the trade winds across the Pacific based on MSLP measurements here what I've computed is just the trend in haddist over the period 1979 to 2014 to a 35 year period so you know that this is a reasonably long period and what we see here is this kind of classical picture that we see many times already in this meeting cooling in the east pacific warming relative warming in the west and in the Indian Ocean and if you look in the models at this in so these are historical plus rcp 4.5 spliced on apologies for the the lines over the land this is because I'm using a sort of vintage piece of software to do this to do these graphs and if I compute an sst index which is this one this box minus this box as a kind of indication of the the gradient of temperature across the pacific here and plot that out as a as a time series it's clearly got lots of interannual and decadal variability but if you fit a trend it's it's a sort of slight negative trend as indicated in the the contour plot I'm not suggesting that it necessarily is a trend because of this interannual to decadal variability but if you take the null hypothesis that it is a trend and then we compute here the observed trends in the vertical bars from various different sst data sets these are their cmit 5 models a histogram of about 100 simulations different models different initial conditions and none of the models actually get their magnitude so this is consistent with scott's presentation that the models do not either that the real world has produced a very rare decadal stroke multi decadal event in the pacific or the models have not got high enough decadal multi decadal variability and of course this is all consistent with so this this sort of sst index is consistent here are the trade wind strength so you can match up things and here's the pdo index so if this meeting was being held in a couple of months time this graph would probably look a bit different because I could put on 2015 that's meant to say 2015 up here and well from the from the 10 months of data that I had up to this point well we see a pdo a pretty strong kind of pdo index as we might have expected because of the end so and and a fairly strong warm east pacific relative to west pacific so is it a trend is it variability um if it is a trend then it's not well it's it's it's partially consistent with the kind of ocean dynamical thermostat hypothesis which has already been mentioned and has published here by our new friend Amy Clement so this none of the models really do this so this kind of theory has largely been I would say discredited but it's not popular um so maybe we need to go back to revise our theories um and and as I said earlier on you know this these because because these trends are kind of broader than just the equatorial region none of these theories really necessarily completely apply where am I going next so this is kind of nearly near the end so are these really forced trends in the pacific if they are then we need to revise our theories about forced trends if it's decadal variability then either the models um uh don't have a large enough decadal variability so either of these two options is a kind of failure in theory and models uh it could be a fluke large amplitude amplitude natural variability but in other areas of climate science we usually call this detection if something is outside the range of natural variability we see in models so we can't have one one rule for things that we kind of know know about and understand one rule for things that we don't know about and understand and are we about to enter a surge and and finally I just maybe you can read these for yourselves but um I just make some notes about general about the general aspects of decadal climate variability and predictability um I think you know closing the energy budget has been a great triumph in in climate science but for these types of features we're looking at order point one what's been described per decade trends in in these you know the various terms in the heat budget and I guess well I'm not on I don't know much about measurements but I guess these are hard to measure hard to quantify so that means I think we're really forced to look at spatial patterns mechanisms and you know perhaps more impacts you know we focus a lot on oceanic variables from the point of view of understanding what's going on but maybe we need to think more about atmospheric processes storms etc etc and that might lead us also on to the impact part uh spatial uh you know forcing forced decadal variability tends to be thought about more at the global scale perhaps we need to think more about the spatial patterns of response of course there's been some work on that and my final conundrum uh which I've told a few people around the room is that uh at least it looks like the Atlantic seems in some way more predictable uh than the pacific but the pacific seems to be more influential at least on the global mean temperature so thank you