 I think our experiments are still running anyway. Yeah, it's a flame of knowledge. Knowledge. Like the three squares down for something, but I can't. Okay, so thank you very much for the introduction. And I'd also like to thank the organizers of this workshop for inviting me to give this presentation. So today I'm going to talk about some idealized experiments that I've been doing very recently with OpenIFS, looking at how the structure and characteristics and statistics of extra tropical cycles will change in the future. And I'd like to thank Bibi Hapanala and Yolanda Ryzenen, my colleagues in the University of Helsinki who've helped with this work. And I'd also like to thank Glenn Carver for trying to answer my never-ending stream of questions about how to use OpenIFS, Helen Daker and also Kevin Hodges for helping me with various bits of code and the technical aspects here. So I hope most of you know what extra tropical cyclones are, but I thought it would be a good idea to show a quick example of what I'm really talking about. So when I talk about extra tropical cyclones, I'm thinking about large-scale weather systems that are in the mid-latitudes. So this is an example of an infrared satellite image taken from Monday evening, actually. And over-plotted is the surface pressure. So we can see this is a large, mature, extra tropical cyclone. It's the centre here, trailing cold front and lots of cloud associated with it. We can also see there's an additional low-pressure system also here. So this is the United Kingdom here and we can see that there's another low pressure here. So both of these we can think of as an extra tropical cyclone. But we can see that both of these systems look very different. So it's important to remember that not all extra tropical cyclones are equal. They have different characteristics, different strengths, for example. They have different amounts of precipitation associated with them, even in our current climate. However, extra tropical cyclones can lead to quite strong impacts that affect society. Three of these pictures are taken from my adopted country, Finland. So the top two are from Finland. So often in winter extra tropical cyclones can bring quite heavy snowfall. This is in the centre of Helsinki. You can see that the transportation is having some challenges here. And also at the airport here, you can see in Helsinki we have very efficient snow clearing. But we still need to do it. And also this costs a lot of money in Finland to deal with snow every year. You have to budget for how much your snow clearing is going to cost. The next picture here, this is also in Finland. And this is another impact of extra tropical cyclones, which is maybe more common to people in its wind. And this can cause damage to forests, for example, which affects the economy in Finland. Forestry is a major industry and wood production is a major industry. The other three pictures I've taken from my native country, the United Kingdom. And we can see the other impacts. For example, again, you have wind damage, falling trees. We also have flooding. So this is taken from a winter a few years ago in the UK where there was a stream of extra tropical cyclones, one after another. That affected the UK and brought lots of precipitation, which led to sustained flooding. And this is on the south coast of the UK where the main train line here into the south west was basically swept away by damage related to extra tropical cyclones. So they can cause a large amount of disruption in our current climate. But the question I'm asking and thinking about is, well, how are these going to change in the future? Are they going to have more precipitation associated with them? Are they going to be stronger, so stronger winds? But are there going to be fewer cyclones or more cyclones? So this is the overall motivation for this work. And we need predictions of these systems on different timescales. For example, we need to think about individual weather events. So one individual storm. We need to be able to reliably predict this. There's also now the next step, which is the sub-seasonal prediction. So we need to try and understand, will the coming winter be more stormy than average or not? But we also need to think about these long-term changes. So this allows, for example, adaption and planning infrastructures. So coastal defences, for example. And also for development of power lines, for example, what do they have to withstand in the future? So is this part here about understanding really long-term challenges, sorry, long-term changes, which I'm going to concentrate on today? So you might think, OK, well, how about you just go and look at some C-MIP5 models. So the current generation of our climate prediction models. And what do they show? Well, I think this has been touched on already today. And there's quite a large degree of uncertainty in how the storm tracks will change in the future if we look just at the climate models, but also if you look at storm tracks. So this plot here shows the change in extra tropical storm track density, basically between the future period and a historical period for two different scenarios. So the RCP 4.5, which is a medium scenario, and then this much higher emissions scenario, RCP 8.5. And the shadow shows the change in basically the number of extra tropical cyclones. So when it's blue, the models are predicting that there'll be a decrease in the number of storms. Where it's red, there'll be more storms. But the hatching here, so the stifling, these dots, these show the areas where at least 90% of the models at least agree on the sign of the change. And you can see that there's not many areas that are the stifled areas. So these are the areas where we actually have confidence in these predictions. So for example, the Mediterranean, a decrease in the number of cyclones in the low scenario and a more pronounced one in the higher scenario. But we can see that, for example, especially in the North Atlantic here, so affecting the UK, Southern Sweden, and into Finland here, there's very little confidence in these future predictions. So therefore, although we can look at the CMIT 5 models and we can learn a lot from them, it's very hard to understand the basic dynamics that are leading to these signals because there's so much variation between the different models and how the different models respond. So I think the overall aims, which I've tried to introduce so far, are what I'm trying to do here is I'm trying to quantify how will the number and track intensity of the cyclones change in the future. And this is a topic that many people have addressed before. Many people look at this in climate models as well, for example. And there's lots of studies on this first question. The second question which I'm trying to address is, to me, much more interesting in a sense. And it's how will the structure of the extra tropical cyclones change? Will they have, for example, different frontal structures? Will they have stronger warm conveyor belts, for example? So this is what I hope to eventually manage to answer with open IFS. However, to answer the first one, I really think you need to have some understanding of the... Sorry, to answer the second question, you have to have an understanding of the first one, too. So, as I said, the CMIT 5 models are very complex. So I'm going to the complete other end of the spectrum and I'm taking a very idealised approach. I'm trying to simplify the problem as much as I can, but still retaining a realistic atmospheric setup. So I'm calling these simplified climate change experiments. This is maybe a stretch of the term climate change experiment, but essentially in this work we're planning to do four experiments. So we have a control experiment, but we're also looking at the effect of increasing the sea surface temperatures in a uniform manner, so just warming them everywhere. And there's two other experiments looking at the effect of CO2 concentrations and also both of these together. Unfortunately, I haven't had time to finish these results yet, so I'm only going to focus on the first one today. But if you're interested, I have some very preliminary results from this experiment. So we're going to do this using an aqua-planet configuration. So as well as having a very simplified type of experiment, we have a very simplified model. So we have the open IFS configured into an aqua-planet configuration. And some more information about the model. So as I said, I'm doing this with open IFS. So the aqua-planet I'm running has a resolution of T159 and the spectral resolution starts around about 125 kilometres, 62 vertical levels. There's no seasonal cycle in the model, so we have fixed solar radiation, I think it's March equinox value, so the Sun stays above the equator for the whole year. The sea surface temperatures are constant in time, and we have a zoneally uniform specification of the sea surface temperatures. And the way that we initialise the model is very simple. It's maybe a bit crude in a sense, but what we do is I just have taken a real atmospheric state, so I just picked some initial conditions I had already for open IFS. And then I essentially removed the change to land-sea mask, so set everywhere to sea underneath. I removed the topography, and then I did some very basic interpolation of the fields from above the topography to the surface. And then I basically initialised the model. Obviously it's not in a very balanced state when I start because of this very crude initialisation, so I need to run with quite a short time step to start with so the model can get into balance. But after about three months, the model has spun up into a very... into a realistic balance state, and we've checked the spin-up time by looking at the precipitation, for example. And I think even within two months, the model is very balanced. So the experiments I'm going to discuss today are changing the sea surface temperatures. So we have taken, in our control simulation, which is this blue line here, we've taken what's called the Q-OBs sea surface temperature distribution from the Neyland-Hoskins paper, and this is essentially a sinusoidal curve. So at the north pole, so at the south pole, which is here, we basically set the temperature to be freezing, so 273 Kelvin. Also at the north pole, and then we specify this function going up to 300 Kelvin on the equator. In the experiment that I'm perturbing the sea surface temperatures, all I'm doing is a uniform warming of 4 degrees Kelvin. So this line just jumps up by 4 degrees everywhere. I run both of these experiments for 11 years. I discard the whole of the first year just to be very safe to avoid any spin-up issues, although I'm fairly sure you could only discard the first three months and you'd be quite safe. So after this, I have 10 years left for the analysis, and when I run the experiments, the important thing I do is I output the model state every six hours, because I want to look at the extrapical cyclones and I need to track them, so I need this high-resolution output to do this. So here's an outline of what I'm going to present in terms of the results and the rest of the presentation. So first of all, we're just going to have a quick look at the zonal mean state of the control simulation, and my SST plus 4 Kelvin simulation, hopefully to ensure you or to convince you that the model is doing something sensible and it's producing a realistic state. We're then going to have a look at some of these, I'm calling the bulk cyclone properties. So these are really statistical properties, for example, looking at histograms of the intensity of the cyclones, for example. And I'm using an objective tracking algorithm that I'll discuss briefly to do this. And then finally, we'll look at the structure of cyclones by creating some composites of the strongest cyclones in my simulations. So here's the first results, and this is the zonal mean precipitation. So this is averaged over sort of year 2 to year 11. And the blue line here is the control simulation. So we can see that we have a peak in the tropics through around 10 millimeters per day. We have a slight double ITCZ problem. Then we have the dry subtropics, and then we have the peak in the extra tropics in both hemispheres. So we can see we've got around 4 millimeters per day in the extra tropics, which is associated with the extra tropical cyclones, the storm track. We can see that when we increase the sea surface temperatures, so remember they're increased uniformly. We don't change any gradients here. So this is the orange line. So we see we really increase the tropical precipitation quite significantly. And we can also see that in the extra tropics in both hemispheres, we increase the precipitation slightly, but we also shift the peak polewards slightly. And this is consistent with the jet response. So now going on to look at those jet streams. So this is now my control simulation. So what we have here in the colors are the zonal mean, zonal wind. So a cross section from going with height and going from the south pole to the north pole. So we have the two jet streams up here, relatively strong, almost 50 meters per second. The black contours here are the temperature. So we see we have the warm part in the tropics, the baroclinic zones in the extra tropics, and then we resolve the stratosphere somewhat. But we can see that we have relatively realistic looking zonal jets, which if you compared this to, for example, the klemphological mean from Eritre interim, you'd have reason that it would be incomparable. One thing to note here is that the structure is symmetric, but I've said roughly, and that it's not a complete mid or opposite. There are some subtle differences between the two hemispheres, very subtle ones in this experiment. In my CO2 experiment, which I'm not going to discuss today, there are more differences between the different hemispheres, which may mean that a 10-year simulation is not quite enough to get rid of your natural variability. So then if you look at the response of the jets to the warming of sea surface temperatures. So the black lines here are from the control. So we have the maximums around 200 hectopascals. The shading here is the difference. So it's the experiment minus the control. So where it's red, the wind speeds have increased. And where it's blue, the wind speeds have decreased. And I've only plotted the values which are statistically significant here. Almost all of the changes here are actually statistically significant. And we can see to start with that there's more red than blue. So we seem to have stronger jets, but they also seem to move somewhat poleward and also upwards. And I think this is easier to see if we look at the same information, but in a slightly different way. So what I've plotted here is I've only looked at the northern hemisphere now. So we have 20 degrees to 80 degrees and then the zonal mean wind speed at three different pressure levels. So we're almost the maximum of the jet is. So 175 hectopascals, 250, and then much lower down, 850. And you can see actually the response of the jet is not a simple poleward shift in some ways. If we look, for example, at upper levels, the blue line is the control, the orange line is the warmed SSTs. And really what we see here, so the wind speeds increase at this level, but what's actually happened is the jet is moved up. And that's because the tropical, the tropical pause is moving up because we've got more convection in the tropics. So we have a deeper convection and then the divergence into the extra tropical atmosphere is occurring at a higher level. If we look at 250 hectopascals, we can see a more complicated response. So actually towards the equator we see a decrease in the wind speed whereas here we have a slight increase in the wind speed, which you could think of that as a poleward shift, but because the response is asymmetric we've got more increase on the poleward side. The jet also appears to be broadening slightly. At low levels, so this is really what you can think of as the eddy-driven jet. So this is the part that the extra tropical cyclones are really influencing. Here we see a very standard, well a much more typical just poleward shift in the jet speed. And this relates very well to the precipitation plot that I showed you where you saw the precipitation peak in the extra tropics moving poleward. So as I said, I track my extra tropical cyclones using an objective measure. So this is Kevin Hodges' algorithm, originally named TRAC. So it basically aims to identify synoptic scale cyclones. I'm not interested in polar lows or mesoscale cyclones, I'm interested in the very large scale features. So this algorithm essentially looks for localized maximums in the 850 hectopascal relative vorticity field. It's truncated to a coarse resolution because vorticity is a noisy field. We then do some filtering on the tracks. So to make sure I have large scale synoptic systems that are transient, so they're moving, so they're not stationary eddies, I need to ensure that they move at least a thousand kilometers and it lasts for at least two days, so they have to be relatively long lasting systems. I also exclude some cyclones which do not have at least one point north of 30 degrees north. The tracking algorithm actually picks up quite a lot of tropical cyclones which are generally very weak and they generally stay around in the tropics and don't do very much, but there's quite a few of them and they have very weak relative vorticities. So I remove all of them. And then the output of tracks is essentially you get the maximum vorticity but you also get the position of each individual cycle in every six hours. And from this you can obviously calculate statistics such as the genesis latitude, lysis latitude, et cetera. So some results from the tracking algorithm. So what I have plotted here, so we have a histogram, a normalized histogram of the maximum vorticity. So this is times 10 to the negative 5 seconds to minus 1. And this is the orange colors are the control and then the blue is the experiment with the warmed SSTs. So you can see in this distribution that they're relatively similar in a sense and if we look at the mean values of the maximum vorticity they're actually very similar in both simulations. If we look at the standard deviation we can see that if we increase the SSEs we have a slight increase so that the distribution is broadening. But also if we think about the extreme cycles. So if we look to the tail here we see the blue parts are poking up above the orange parts. So this means that in the SST plus 4 distribution simulation we've got more extreme cyclones in a sense. And we see this, so if you look at the maximum ones this is the strongest cyclone and the strongest cyclone is quite a bit stronger. And also I've written here this 200 threshold. So this means when I create the composites I'm looking at my 200 strongest storms. So I want to see this is the vorticity value which the 200 storm has in the data set. And again this increases slightly. So when we warm the SSTs it appears that the mean and the median intensity of cyclones does not change but maybe there are some more extreme cyclones. And we can also think about the number of cyclones so that's the top line here and we can see that warming the SSTs leads to a small reduction in the number of cyclones. And this obviously makes sense if you decrease the number of cyclones each cyclone probably has to do more work so you have to have slightly stronger cyclones. So this is again the same type of plot but looking at the Genesis and Lysis latitudes so Genesis on the left, Lysis on the right. And again we see what we saw in terms of when we looked at the jet shift and the precipitation shift that the Genesis region is moving forward so again we have the blue parts here and also the Lysis region is moving forwards in both of these cases. And the typical shift so down here so the negative sign indicates a poleward shift in terms of degrees of latitude so they're moving between around 1.5 and 2 degrees towards the pole. Interestingly the Lysis regions are moving less forwards than the Genesis region so in a sense on the average the cyclones are not travelling as marionally. Okay so now we're getting on to look at the composites of cyclones so the structure of cyclones and I'm using a method which Jennifer Cato used as part of her PhD and it's published in his paper in Journal of Climate. So there's three steps to this process so the first one is we track our cyclones so this is the step here so we use track to find the track of each individual cyclone and then I select the tracks I want to use so I filter them to only pick 200 tracks so I want the 200 strongest storms and then for each track I find the position of the maximum intensity along the track and I call this time equals zero so t equals zero so this is essentially equal to each of these black dots here so there's three different cyclones here and we find the maximum intensity and what we do when we find the maximum intensity we then combine the output from track with the output from open IFS so what I do is I take the output from open IFS in a circle so this is a spherical cap so we have a spherical grid I overlay this on my cyclone's center and then I interpolate from the open IFS output grid to the spherical cap which is normalised so centred on the position of the cyclone I do this for lots of different offset times so here this would be time equals zero but I also look at time equals minus 24 hours minus 48 hours and then on the other side I look at time plus 24 and plus 48 hours and then we have we do this for each cyclone and once we've done that we need to add them all together to take an average there's one last complication we need to rotate all of our cyclones so that they're travelling in the same direction because obviously if we just average these on a north south grid this one is travelling this way and this one is travelling that way the same features in terms of cold fronts and warm fronts are not going to be in the same place so we need to rotate them which is shown here so they all travel directly east and then all we do is we take an average of the 200 storms to get our resulting composite cyclone and we get the composite for each different offset time so when we do this this is the results so this is the composite for the mean sea level pressure the top line is showing the control simulation and the bottom one the warmed SSTs so time equals zero the time of maximum intensity this is 24 hours before the time of maximum intensity and then 24 hours after so if we just focus on the top line now we can see that we have a very strong low pressure system here it's around 960 I think I guess it's hard to see the numbers 962 is the minimum pressure about 950 at the maximum intensity and then it starts to fill later on and we can see it looks like a relatively sensible cyclonic structure so at the bottom this is the second experiment but I think it's more intuitive to look at a different plot so now this is the difference so the blue parts here this shows that the pressure is lower in the warm sea surface temperature so all offset times the pressure is lower to the north and slightly higher to the south and also the pressure in the center of the cyclone here is actually slightly higher at the early part at all of these times but the pressure gradient across the cyclone is essentially increasing when we warm the sea surface temperatures we can look at the same types of figures now for the total column water vapor so here the blue colors are the moist parts so here you see the warm conveyor belt developing the time equals zero you see it wrapping up cyclonically into the center of the cyclone you see this drier air being pulled down back behind the cold front and then later on you see that the overall the the cyclone is drier later on this is usually because it's moved further north but you see again that there's drier air to the north and moisture air to the south you can see very clearly in the sst plus four that almost everywhere the moisture increases and this is not surprising when we increase the sea surface temperatures the water by the clausia-clapron equation there's going to be more moisture in the atmosphere the total column water vapor is going to increase but we can maybe see that there are some structural differences for example if we compare time equals zero this one maybe is more cyclonically wrapped up in a sense than at this point here so maybe there are some subtle structural differences here again this is the difference plot on the bottom now and again you see so most of the strong differences occur earlier on I should say this is maybe a misleading color bar the smallest value here is actually zero there's moisture increasing everywhere but you see there is a very large increase here it's almost 10 kilograms per meter squared so it's a large increase here but then also you see that later on especially at the time of maximum intensity there's not a uniform the spatial pattern is not uniform in the increase there's this part here which increases less for example which is really in the warm sector part of the cyclone here so then if we go on to look at the precipitation so you can see that if we just look at the top line here we see that most of the precipitation actually occurs before the time of maximum intensity this is an agreement with previous studies these precipitation values are around 2 millimeters per day so they're relatively sensible and you see it really maps out this typical warm conveyor belt area of a cyclone but you can also see there's some so you see that there's if we compare this one here and also then in the sea surface temperature you can see the position of the maximum is shifted so here it seems it's moved slightly further north in a sense than in the control simulation and this is more evident if we look in the difference plots so here when these blue colors show an increase in precipitation these brown colors show a decrease in precipitation so you can see that there's the area of the precipitation is shifting and it's moving more sort of to almost like ahead of the warm front you could almost think that the precipitation associated with a cold front is decreasing but the precipitation is associated with the warm front is potentially increasing so this maybe suggests that there's some change in the structure of the fronts and the cyclones this is then moving on so related to the vertical the precipitation is obviously the vertical motion and I've only looked at this at 700 hectopascals and I'd say this is omega so in units of pascals per second so the blue colors here which are negative are ascent and the red colors are descent and this matches relatively well with the precipitation patterns particularly if we compare the time equals zero so if we look at the difference plots for the vertical motion again so here this is red so it's less ascending motion and here there's more ascending motion and this matches up very well with what we saw in the precipitation pattern so these are very new results I have to say I'm not sure I fully understand them but I've been having a think and a bit of a hypothesize about what might happen with the structure of cyclones in the future climate and this is a paper from Barnes and Hartman 2012 and they were essentially looking in this is CMAP 3 models actually so two different scenarios so one is a climate change simulation and one is a control and they were looking at how wave breaking changes so you can have anti-cyclonic wave breaking or cyclonic wave breaking and this in a sense can affect the structure of the cyclones you have so in their analysis they found that cyclonic wave breaking decreases in their climate change experiment where anti-cyclonic wave breaking increases in the in the future in their climate change experiment so this is one thing that I hope to look at in the future so one question I could ask are the cyclones in my warm SST experiment are they more anti-cyclonic than the control simulation and I think the best way to address this is to look at some composites of isentropic potential vorticity I think so this is hopefully what we plan to do next to try and understand these if there are any structural changes in these cyclones so some preliminary conclusions in a sense so when we increase the sea surface temperatures in our Aquoplanet configuration the number of cyclones decreases slightly but the mean intensity does not change there are maybe there's more extreme cyclones so ones with very high vorticities and also the genesis and lysis regions move forwards as do the zonal jets there are some subtle changes to the cyclone structure which I don't fully understand yet the ascent and the precipitation definitely seem to change and maybe there's some indications of a different type of cyclonic or anti-cyclonic wrap up I have to say these are very new results so hopefully there's more to come soon but I would like to stop now and thank you for your attention and answer any questions