1
00:00:07,730 --> 00:00:11,750
Thank you very much, Tim. And it's a pleasure to be here and to speak to you.

2
00:00:12,350 --> 00:00:18,140
So just to quickly say a lot of what I'll say here, those are a couple of background papers here.

3
00:00:18,950 --> 00:00:24,409
Review prospective papers, one a couple of years ago on circulations, climate change in general,

4
00:00:24,410 --> 00:00:29,330
and then one just this past spring on event attribution and how we think about circulation there are

5
00:00:29,330 --> 00:00:35,450
so you can find these on my website and so if you don't catch up with them here and follow up there.

6
00:00:36,050 --> 00:00:45,590
Okay. When we talk about climate change and and anthropogenic climate change, this is, if you like, this the poster for that.

7
00:00:46,370 --> 00:00:53,209
We know that there's long term warming of the climate system and we know it's the issue and the anthropogenic causes.

8
00:00:53,210 --> 00:01:00,380
And so this is one of the map that's been updated a few times, but this is from the last IPCC report, the Intergovernmental Panel on Climate Change.

9
00:01:00,860 --> 00:01:05,599
What you see are a bunch of a postage stamps and they either represent the the red ones.

10
00:01:05,600 --> 00:01:10,700
Here are a surface air temperature over all over a continent, an entire continent.

11
00:01:10,700 --> 00:01:18,409
So South America, North America, Europe, Asia, Africa. And the blue ones are the the the ocean heat content theory.

12
00:01:18,410 --> 00:01:24,200
That's the the upper part of the ocean over the over the ocean basins, North Pacific, South Pacific and so on.

13
00:01:24,650 --> 00:01:29,000
And then there's some ice as well, the sea ice and the in the in the polar regions.

14
00:01:29,000 --> 00:01:37,790
And in each case, what you see is a band that's in purple, which is the as a climate models without anthropogenic climate change.

15
00:01:38,150 --> 00:01:44,360
And maybe I'll focus on temperature here and pink as the models with climate change and blue are the observations.

16
00:01:44,360 --> 00:01:51,680
And in each case, you see it's clear that you can only explain the observed record if you include the effects of climate change.

17
00:01:51,980 --> 00:01:56,000
And the observations are incompatible with with the model without climate change.

18
00:01:56,000 --> 00:02:05,630
This is the simplest attribution statement and it holds if you look at the scale of entire continents or ocean basins.

19
00:02:06,590 --> 00:02:14,240
So this is, you know, from a classical causation point of view, this is what you would call sufficient causation.

20
00:02:14,520 --> 00:02:19,190
They remember this is necessary and sufficient will get to necessary in a moment.

21
00:02:19,190 --> 00:02:23,239
So sufficient causation, if you just consider this is a cartoon,

22
00:02:23,240 --> 00:02:29,090
if you think of a of a geophysical quantity like surface air temperature, of course there's natural variability.

23
00:02:29,090 --> 00:02:35,480
So if you talk about temperature over a certain continent or let's say you have a PDF,

24
00:02:36,110 --> 00:02:44,149
and so this is just a cartoon and we would represent a distribution under a world without climate change,

25
00:02:44,150 --> 00:02:48,740
with a peanut and a world with climate change with P one.

26
00:02:49,130 --> 00:02:53,720
So Pinard is strictly a song is called The Counterfactual.

27
00:02:53,720 --> 00:02:56,670
It's the world that would be if it wasn't for climate change course.

28
00:02:56,690 --> 00:03:05,719
We often use the pre-industrial period as a proxy for that counterfactual so that the thing about these PDFs is that they're well separated.

29
00:03:05,720 --> 00:03:15,530
So if you if you take a single realisation, you know, with with pretty high confidence, which of these two distributions you're you're you're in.

30
00:03:16,430 --> 00:03:23,569
And this is of course you could say an extreme case where we've really shifted the PDF and this value out here,

31
00:03:23,570 --> 00:03:29,000
which was extreme in the counterfactual climate, is now the new normal.

32
00:03:29,000 --> 00:03:31,940
So this is a case where it's actually manifest.

33
00:03:31,940 --> 00:03:39,980
This is a very but this is what we had on the previous plots, right, where the PDFs get well separated.

34
00:03:40,280 --> 00:03:43,849
It's called predictive power for individual realisations and for climate.

35
00:03:43,850 --> 00:03:48,139
A Germany applies if you look at at land surface temperature over sufficiently large

36
00:03:48,140 --> 00:03:52,970
areas and sufficiently long times like a climate over one year or something like that.

37
00:03:53,630 --> 00:04:03,020
And so this is just an example. I like this figure. It's a, it's a graphic of, of the hottest the decade of the hottest summer.

38
00:04:03,020 --> 00:04:08,150
And this is using climate proxy records. So it's a 500 year record.

39
00:04:08,630 --> 00:04:17,270
And when you add ten years to that record, so the, the, the vertical scale is the temperature and the colour is the decade.

40
00:04:17,600 --> 00:04:24,350
And so what happened is that basically the whole content becomes much more red, which means that the records happened in the last ten years.

41
00:04:24,350 --> 00:04:27,830
And of course, if we updated this now, this is only up to 2010, it would become even more red.

42
00:04:28,430 --> 00:04:36,020
So basically the whole map got redrawn in ten years, which couldn't have happened by chance.

43
00:04:36,020 --> 00:04:43,550
So this is a great example of sufficient causation, but the problem is that it's not only about temperature.

44
00:04:43,790 --> 00:04:49,729
And so this contrast of precipitation and temperature, precipitation, of course, is very, very important because we need for water.

45
00:04:49,730 --> 00:04:53,870
And you can have, of course, too too much as bad and too little as bad.

46
00:04:53,870 --> 00:04:58,190
So from an impact point of view, precipitation is often one of the key variables.

47
00:04:58,880 --> 00:05:04,459
So these plots, you may not have seen these before, but these are common in the IPCC.

48
00:05:04,460 --> 00:05:10,220
So what's plotted here is a change in on the left. Surface temperature on the radius of precipitation.

49
00:05:10,790 --> 00:05:20,030
Precipitation is always in percent, so blue and green means wetter in red and in yellow means drier.

50
00:05:20,360 --> 00:05:28,340
Temperature is is just a scale of warmer or colder. This is at the end of the century under one of the one of the climate change scenarios.

51
00:05:28,880 --> 00:05:31,820
So and then on top of that are the symbols.

52
00:05:31,820 --> 00:05:40,360
And you can see that there's either stippling, which is the dots, and simply means that the models all agree on the sign of the chain.

53
00:05:40,370 --> 00:05:47,960
So it's a robust change. Hatching means that the change is small compared to natural variability.

54
00:05:48,380 --> 00:05:54,530
And if it isn't, if anything, if you don't have either hatching or shading, it means that the models are disagreeing.

55
00:05:54,890 --> 00:06:02,270
So if you look at the temperature, it's stippled everywhere, except for a little patch in the North Atlantic Ocean.

56
00:06:02,270 --> 00:06:08,030
Because of circulation issues, there changes, but basically everywhere else and certainly over land it's stippled.

57
00:06:08,930 --> 00:06:12,350
But if you look at precipitate, so this is the case of a sufficient causation.

58
00:06:12,350 --> 00:06:14,450
But in the case, if you look at a precipitation,

59
00:06:14,450 --> 00:06:25,440
you see that the stippling is confined pretty much to a high latitudes where not many people live or over the ocean where not many people live.

60
00:06:25,460 --> 00:06:31,700
So if you look at the land areas in the in the temperate latitudes or the tropics,

61
00:06:31,700 --> 00:06:35,990
you find a lot of catching and a lot of areas without any any stippling at all.

62
00:06:36,350 --> 00:06:40,729
So this is basically saying that the model that we don't have robust predictions either

63
00:06:40,730 --> 00:06:47,210
because of just the climate noise and natural variability or because the models don't agree.

64
00:06:47,480 --> 00:06:53,240
So it's quite a contrast between what you get for precipitation and temperature and going to be really the main theme of my talk.

65
00:06:54,410 --> 00:07:02,540
So we'll talk first about natural variability. We know, of course, the variability is a coupled have a sort of ocean phenomena.

66
00:07:03,470 --> 00:07:09,980
I'm interested in the atmospheric manifestation of that. But of course it should be remembered that the ocean is part of that whole system.

67
00:07:10,700 --> 00:07:18,560
So one of the people talked about most modes of variability, and one of them that the climatologists talk about is the North Atlantic oscillation.

68
00:07:18,920 --> 00:07:25,100
And it basically affects whether the jet stream is is is heading over Britain or heading for the south.

69
00:07:25,100 --> 00:07:30,050
And so that will affect whether the northern part of Europe, let's say, is is dry or wet.

70
00:07:30,410 --> 00:07:36,889
And so certainly ever since I've been in this country, everything's blamed on the jet stream.

71
00:07:36,890 --> 00:07:40,010
So if you want to explain why things are the way it is, it's because of the jet stream.

72
00:07:40,010 --> 00:07:43,489
And so that's basically this is one of them, one of the ways of characterising it.

73
00:07:43,490 --> 00:07:49,370
There's other other ways, too. But this just shows one of the classical variations.

74
00:07:50,930 --> 00:07:57,889
It's a convenient one because it's it's some the index actually is based on sort of surface pressure measurements over Iceland in the Azores.

75
00:07:57,890 --> 00:08:03,260
So we have a pretty good time series of that. So that shows the record of the observations.

76
00:08:03,260 --> 00:08:08,930
That's about 110 years. And you see that this North Atlantic Oscillation Index goes up and down and up and down.

77
00:08:09,800 --> 00:08:12,860
So this is the problem that we're faced with, with chaotic variability.

78
00:08:13,160 --> 00:08:21,050
And it's what very long timescales. You can see that there's there's year to year changes, but there's also changes that are multi decadal.

79
00:08:21,710 --> 00:08:29,750
And of course, this makes it very hard to pull out a climate change signal because there's this whole trend that you can pull out of this time series.

80
00:08:30,200 --> 00:08:36,860
In the early two, thousands people were quite excited about this this apparent trend here, and there were papers published.

81
00:08:37,130 --> 00:08:44,360
Even the IPCC attributed this to climate change, but I think no one would would make that claim anymore.

82
00:08:44,360 --> 00:08:47,990
And actually, no one actually believes that an EL probably should have a trend.

83
00:08:48,380 --> 00:08:55,700
So this is really a confounding issue when we talk about trying to identify climate change in the circulation.

84
00:08:56,960 --> 00:09:04,510
So of course, precipitation, as you can imagine, is controlled by temperature because the warmer the air,

85
00:09:04,530 --> 00:09:08,120
the more water vapour you can have because our closest cooperation.

86
00:09:08,210 --> 00:09:11,960
So that's kind of a thermodynamic control, but it's also controlled by the circulation.

87
00:09:12,560 --> 00:09:18,860
And if you read the IPCC reports, basically all the statements about precipitation that have any level of conflict,

88
00:09:18,860 --> 00:09:23,689
any any reasonable level of confidence behind them are based on thermodynamic arguments.

89
00:09:23,690 --> 00:09:29,659
Basically, just as you warm up the air and holds more moisture and then you can make some arguments around that.

90
00:09:29,660 --> 00:09:35,150
And the if the the way it's sometimes summarised to me and my guess is like the weather and the dry good dryer.

91
00:09:35,660 --> 00:09:41,030
But if you look at many and I think we're realising now that this just doesn't hold over land,

92
00:09:41,030 --> 00:09:45,650
which is kind of most important place for rain, but in many.

93
00:09:46,490 --> 00:09:52,069
So it's say it's controlled both by circulation and by temperature. So this is what this is showing is simple model.

94
00:09:52,070 --> 00:10:00,549
Of course, we only have a single realisation in the atmosphere, but with with a model you can generate an ensemble.

95
00:10:00,550 --> 00:10:05,510
And this is one of the first what was called, I guess at the time, a large ensemble.

96
00:10:06,020 --> 00:10:10,240
It's about 40. Combo members, which isn't that large, I suppose, but at the time it was large.

97
00:10:10,810 --> 00:10:14,350
This is with me with the anchor model from the US and it's PDF.

98
00:10:14,350 --> 00:10:19,770
So the winter time. So is December, January February widget on trends over 55 years.

99
00:10:20,140 --> 00:10:30,250
So going in the future in a pretty large area, the Eurasian North Atlantic sector, each part has two PDFs.

100
00:10:30,280 --> 00:10:35,859
One, the great is the control that's without climate change and the red is with climate change.

101
00:10:35,860 --> 00:10:38,950
So for example, if we look over here and this is surface air temperature,

102
00:10:39,250 --> 00:10:44,049
you see that the grey has a mean that's within noise of being zero because there shouldn't be any trend.

103
00:10:44,050 --> 00:10:49,450
There shouldn't be any mean trend in the in the in the in the control simulation.

104
00:10:49,450 --> 00:10:56,320
But you can see that you can certainly get members that will have trends that are either positive or negative over 55 years.

105
00:10:57,400 --> 00:11:01,780
But if you look at the at the PDF with climate change, you can see there's a well separated.

106
00:11:02,650 --> 00:11:07,059
So from a single realisation, you know, you'll know whether you're climate change or not.

107
00:11:07,060 --> 00:11:10,690
Again, this is consistent with the kind of figures I was showing at the beginning.

108
00:11:11,110 --> 00:11:17,950
But if you look over here, this is sea level pressure. So that's an index or a measure of the circulation and you see that the PDF.

109
00:11:17,950 --> 00:11:20,230
So there is a change here, there is a shift.

110
00:11:20,920 --> 00:11:26,860
It's significant because you've got enough ensemble members to see it, but it's basically about one standard deviation.

111
00:11:26,860 --> 00:11:34,900
And so if you only have a single realisation, you're not going to be able to tell which these two populations are from the strongly overlapping.

112
00:11:35,170 --> 00:11:40,959
And if you look at a precipitation, it looks a lot more like sea level pressure than it does like surface area temperature.

113
00:11:40,960 --> 00:11:52,570
So in this case, a precipitation is really being controlled by the circulation and we've got strongly overlapping PDFs and the signals noise is small,

114
00:11:54,160 --> 00:12:02,290
but the fact that is so IPCC of what we call this hash because it's small compared to climate change, but of course it's it's small in some sense.

115
00:12:02,290 --> 00:12:08,980
But if you look at the risk of and these are not even extremes in a tail sense, but just on the shoulders of the distribution,

116
00:12:09,400 --> 00:12:18,250
the chance in this case of being dry has gone down by about a factor of two, and the chance of being wet has increased by both by a factor of two.

117
00:12:18,670 --> 00:12:25,840
So the change in risk is is not small yet. We're dealing with a prediction with only one, one realisation.

118
00:12:25,960 --> 00:12:30,310
You can never of course validate this with a single realisation.

119
00:12:30,310 --> 00:12:40,330
So this is the challenge that we're facing. And what do you do with, you know, of a winter like 2013, 2014?

120
00:12:40,330 --> 00:12:48,430
This was not long after I arrived and and you may recall there was all this flooding in January.

121
00:12:49,300 --> 00:13:00,709
This was the This is the Time series from the Met Office of the Precipitation in southern England and Wales for all of January over 130 years.

122
00:13:00,710 --> 00:13:08,080
So good long record and you see that that winter was or was a record amount and there was all this flooding.

123
00:13:08,410 --> 00:13:16,000
Now, this is a weather map it's I guess you don't at writing this is the standard theta on V2 which I won't try to explain,

124
00:13:16,000 --> 00:13:22,840
but it's basically a map of the weather systems and what some people would now call the polar vortex, the tropospheric polar vortex.

125
00:13:23,410 --> 00:13:30,280
And and the blue is cold air. And so actually on this is on January 5th, 2014.

126
00:13:30,280 --> 00:13:37,299
And actually, you can see this big cholera outbreak all over the central U.S. And if you look at the newspapers at that time,

127
00:13:37,300 --> 00:13:41,260
there were all these cold, cold snaps and everything. It was a lot of a lot of talk at the time.

128
00:13:41,620 --> 00:13:46,600
And you see this big storm travelling towards Britain. This was the first of many storms that came in.

129
00:13:46,900 --> 00:13:52,479
None of the storms were particularly extreme. But what happened was that the jet stream got quite stuck.

130
00:13:52,480 --> 00:13:59,200
So the proximate explanation for why there was so much rain was just that the jet

131
00:13:59,200 --> 00:14:03,730
was stuck in the same place and the storms just kept rolling in one after another.

132
00:14:04,300 --> 00:14:10,900
But if you ask is is a stock jet just re more or less likely under climate change?

133
00:14:10,900 --> 00:14:14,080
There's no no accepted view on that.

134
00:14:14,080 --> 00:14:17,620
And certainly there's no accepted view on on how much that would change by.

135
00:14:18,520 --> 00:14:25,300
So this is the kind of challenge that we're facing with a lot of extreme events that that are controlled by by the jet stream.

136
00:14:26,710 --> 00:14:34,690
So if you have a small signal to noise ratio, then we're talking about not not sufficient but necessary of causation.

137
00:14:34,690 --> 00:14:39,280
So this is a this is a case where we have strongly overlapping PDFs.

138
00:14:39,280 --> 00:14:46,930
So if you have an extreme, you already have to be extreme in the counterfactual climate because you have to be out in the tail of the distribution.

139
00:14:47,380 --> 00:14:54,550
So actually for all of these extremes, it's correct to say it's mainly natural variability.

140
00:14:54,550 --> 00:14:57,850
So you have papers, I will say this and they'll say it's mainly natural variability.

141
00:14:58,090 --> 00:15:03,280
And that's true in the sense that you have to be out in the in the tail to actually get the thing in the first place.

142
00:15:04,600 --> 00:15:11,659
So a stuck jetstream would be an example. Of that. So that means we're dealing with with multiple causal factors.

143
00:15:11,660 --> 00:15:20,390
We can't just say climate change. We can't say climate change caused it. In a sufficient sense, it just was one of many factors.

144
00:15:21,290 --> 00:15:27,620
So it's a bit like an accident investigation, and you really should look at it with all these multiple causal factors.

145
00:15:28,460 --> 00:15:33,560
It's got no no predictive power for single events because if you just have a single event,

146
00:15:33,560 --> 00:15:36,800
you don't know if you're in the counterfactual or the factual distribution.

147
00:15:37,460 --> 00:15:40,070
As I said, the anthropogenic effects are in some sense small,

148
00:15:40,490 --> 00:15:48,530
but the the impacts depend in a very non-linear way on the on the hazard, as it's called.

149
00:15:48,530 --> 00:15:49,159
So, for example,

150
00:15:49,160 --> 00:15:58,160
flooding will depend will the flooding damage is a very steep in terms of cost is a very steep function of the flood level and that kind of thing.

151
00:15:58,160 --> 00:16:00,200
So small effects really do matter for impacts,

152
00:16:00,620 --> 00:16:06,380
but it means that the attribution problem is very different from what we talked about with climate change in terms of temperature.

153
00:16:07,340 --> 00:16:11,150
It's fundamentally a probabilistic you're just talking about a change in likelihood.

154
00:16:11,840 --> 00:16:18,170
So if you want to quantify this, you're going to, if will with a model, you're going to need very large ensembles of simulations.

155
00:16:19,460 --> 00:16:27,080
So this is what I call the risk based approach. And and Oxford really pioneered this, I would say.

156
00:16:27,740 --> 00:16:31,450
But there's a number of issues with that that I want to highlight.

157
00:16:31,460 --> 00:16:36,680
One is that the event needs to be generalised because in fact every event is unique, right?

158
00:16:36,680 --> 00:16:41,060
It's like you'll never have exactly the same event.

159
00:16:41,070 --> 00:16:44,690
So if you want to talk about statistics, you need to sample population.

160
00:16:44,690 --> 00:16:51,170
And to get a sample of population, you have to blur the event. So, for example, we're talking about a heat wave, you might say,

161
00:16:51,410 --> 00:16:55,580
exceeding a certain temperature threshold over a certain area, over a certain length of time.

162
00:16:55,910 --> 00:16:59,330
But of course, as many other events that might cause it to do that for other reasons.

163
00:16:59,330 --> 00:17:06,950
So you blur the event to get enough of the sample on, and that will certainly lose local features and may lose connection to the actual event.

164
00:17:06,950 --> 00:17:14,749
So you're not actually asking the question, you're not addressing the question that people are probably asking about if there's, say, a heat wave,

165
00:17:14,750 --> 00:17:20,360
which was really because of, you know, warm temperatures in the city for the or for for the hottest day,

166
00:17:20,360 --> 00:17:23,390
you're going to take aa5 day average or something like that.

167
00:17:24,650 --> 00:17:30,469
The definition, the event is generally arbitrary. I mean, there are categorical events like a hurricane.

168
00:17:30,470 --> 00:17:36,500
But in general, if you're talking about temperature or precipitation or something, you're talking about a continuous variables.

169
00:17:36,890 --> 00:17:42,350
And so you just have to pick a threshold, you have to pick an averaging time, you have to pick an averaging domain.

170
00:17:42,380 --> 00:17:48,800
Those are arbitrary. And the quantification, the actual number that you get can depend a sensitivity on that.

171
00:17:48,800 --> 00:17:52,610
As you can imagine, if you try to estimate the change risk here,

172
00:17:52,940 --> 00:17:56,290
if you're in this part of the curve or way out here, you're going to get a very different ratio.

173
00:17:56,300 --> 00:18:00,140
So the numbers that you get are very sensitive to this definition.

174
00:18:01,310 --> 00:18:07,660
There's a question that Tim Tam certainly asks all the time if you want to get these good statistics,

175
00:18:07,670 --> 00:18:13,010
even if you have to run the model for a very long time. Well, if you have to run the model for a very long time, the model has to be cheap.

176
00:18:13,520 --> 00:18:17,960
If the model is cheap, is it doing a good job? Cheap in the sense of computationally.

177
00:18:18,470 --> 00:18:25,070
So the problem is you may not have the spatial resolution that you need if you want to do thousands of years of runs,

178
00:18:25,070 --> 00:18:29,210
which is what you need, you need to do to get out the tail. So that's a very serious question.

179
00:18:29,900 --> 00:18:34,580
Also, this construction of the world without climate change can be a major source of uncertainty.

180
00:18:34,580 --> 00:18:40,340
The so-called counterfactual, and in particular, since I've already raised this issue of circulation,

181
00:18:41,120 --> 00:18:47,839
if the circulation changes were important for your result, why would we actually believe those if we don't have any?

182
00:18:47,840 --> 00:18:53,900
And this is what I'll now talk about, is if we don't have any basis for that, why would we trust that result?

183
00:18:54,320 --> 00:18:58,610
This is an example from a from Oxford Lee scholar who isn't here anymore.

184
00:18:59,840 --> 00:19:04,459
I really like this paper because I think it highlights the other issues very clearly.

185
00:19:04,460 --> 00:19:11,480
So this is getting back to January 2014. And as I say, there was this very persistent jet stream or stuck jet.

186
00:19:11,930 --> 00:19:16,309
So what's done here is this is with the weather at home here with very large simulations.

187
00:19:16,310 --> 00:19:20,450
And what they did was they was they estimated they made different estimates.

188
00:19:20,450 --> 00:19:29,450
So to get the counterfactual, you have to estimate how much of the sea surface temperature changes were anthropogenic.

189
00:19:29,450 --> 00:19:33,769
And we don't know, of course, how much was anthropogenic and how much might be variability.

190
00:19:33,770 --> 00:19:41,090
Plus, I guess there's observational uncertainties. So you use a model, the climate model, to determine what the half of the time part of that is.

191
00:19:41,510 --> 00:19:47,150
And the answer that you get depends on the climate model. So what they've plotted here is this is zero is the ozone hole.

192
00:19:47,870 --> 00:19:53,989
I'm not not with your stands for but it's this so the was only a persistent state and versus

193
00:19:53,990 --> 00:19:59,959
the number of days in the month on the vertical axis and the return period out here.

194
00:19:59,960 --> 00:20:04,910
So further to the right means it's more rare. And then these are offsets just so you can see them in the red.

195
00:20:04,910 --> 00:20:11,870
In each case, is the is the. Present day condition and the light blue are the different models.

196
00:20:11,870 --> 00:20:16,400
So you see, just to take this case here of 22 days in the state.

197
00:20:17,120 --> 00:20:21,440
Some of these models say that there's been no change in the likelihood of the state.

198
00:20:21,470 --> 00:20:27,450
Some of them say there's been quite, quite a big change. In other words, moving from the right to the left, from blue, blue to red.

199
00:20:27,470 --> 00:20:33,320
So in some cases, you find that there's there's been a big change in the persistence of the jet.

200
00:20:33,980 --> 00:20:41,970
In other cases, not. And then if you if you propagate that to actual flood risk in terms of numbers of properties at risk, the answer depends.

201
00:20:41,990 --> 00:20:44,990
The sign of the answer depends on that state.

202
00:20:45,350 --> 00:20:52,430
And what they did in this paper was also show that the which sea surface temperature changes that that you change really affected the jet.

203
00:20:52,970 --> 00:21:00,070
Wally Well, you can see it here. So basically the circulation changes make the difference between increased and decreased flood risk.

204
00:21:00,080 --> 00:21:07,910
So it means we can't really make a statement about this this risk unless we can get some confidence in which of the circulation changes is correct.

205
00:21:09,170 --> 00:21:16,069
So you won't read all this, but what's there's been a lot of interest, of course, in extreme events.

206
00:21:16,070 --> 00:21:24,860
And the American Meteorological Society publishes a report every year on the on the extreme events in the previous year.

207
00:21:24,860 --> 00:21:29,629
It's a report of opportunity. People send in papers on almost different extremes.

208
00:21:29,630 --> 00:21:38,959
This is the one for 2013. And what's what's done is there's the there's a statement, but then there's in this column,

209
00:21:38,960 --> 00:21:45,380
it's the case where you can attribute the the event or at least an increase in risk due to climate change.

210
00:21:45,380 --> 00:21:52,070
And it's a decrease in risk. And this there's no answer basically null result can't tell.

211
00:21:52,490 --> 00:21:59,719
And for hot events, everybody finds there's an increase in likelihood of hot events.

212
00:21:59,720 --> 00:22:04,130
That's not too surprising for cold events. Several times there's a decrease in cold events.

213
00:22:04,460 --> 00:22:14,720
That's not too surprising. Once you get to heavy precipitation and drought, you can see that things are mixed, maybe up, maybe on no result.

214
00:22:14,960 --> 00:22:17,990
Once you come to storms, you basically can't conclude anything.

215
00:22:18,380 --> 00:22:25,370
So what this is saying is that when you try to do these studies, and especially if you really try to take the part of the uncertainties,

216
00:22:25,370 --> 00:22:30,440
which I just I think most of the studies probably didn't do because they're often just using a single model, so on.

217
00:22:30,890 --> 00:22:36,020
But what you're going to find is that if the attribution really depends on the circulation,

218
00:22:36,020 --> 00:22:39,320
you're probably not going to be able to make make a statement.

219
00:22:39,740 --> 00:22:45,719
But that, of course, doesn't mean that there was no effect. On the longer time scales.

220
00:22:45,720 --> 00:22:50,550
This uncertainty also affects, well, long term climate change.

221
00:22:50,580 --> 00:23:00,120
So, for example, this is of matrix, I guess of of this is contributions to uncertainty of the wintertime changes for ten year averages.

222
00:23:00,120 --> 00:23:06,240
So, you know, so it's a climate variable. It's a ten year average of the wintertime precipitation per breadbox.

223
00:23:06,600 --> 00:23:11,399
And there's basically three sources of uncertainty and what's going to happen in the world.

224
00:23:11,400 --> 00:23:17,100
One is internal variability. We just have chaos and the butterfly effect and we and we always have to live with that.

225
00:23:17,460 --> 00:23:23,730
There's also the model uncertainty, some models, the models of errors, and that that means that we can't that there's going to be issues there.

226
00:23:23,730 --> 00:23:29,640
And then we have the scenario uncertainty, which is we don't know what the greenhouse gases are going to be in the future.

227
00:23:29,940 --> 00:23:33,059
This uncertainty, of course, is one that we can do something about.

228
00:23:33,060 --> 00:23:41,520
So the policy angle is to focus on on the CO2 levels and in the other climate forces and focus on the scenario.

229
00:23:41,520 --> 00:23:47,970
So if you do this for temperature, what you find is that the and I trust that this is a fraction of the uncertainty

230
00:23:48,590 --> 00:23:53,069
that these authors have partitioned into these three different sources.

231
00:23:53,070 --> 00:23:58,350
For all the different climate models that are used and blue is it's a small fraction and red is a large fraction.

232
00:23:58,350 --> 00:24:03,960
It goes up 200%. So if you do this for temperature, what you find is that it's a diagonal.

233
00:24:04,440 --> 00:24:11,100
And so if you look at mine decades ahead, it's basically dominated by the scenario uncertainty and the fact the models disagree.

234
00:24:11,310 --> 00:24:18,900
The fact we have internal variability doesn't matter. The leverage on the uncertainty is, is from how much the CO2 is there.

235
00:24:19,230 --> 00:24:22,890
But if you look at a precipitation, you see that it stalls.

236
00:24:23,310 --> 00:24:27,930
And if you were trying to set your your carbon targets based on precipitation,

237
00:24:28,350 --> 00:24:33,780
as some people talk about a two degree target or something like that or what, one and a half degree target, that's what temperature.

238
00:24:33,780 --> 00:24:38,250
If you tried to do it for precipitation, you couldn't because the models are just too uncertain.

239
00:24:40,780 --> 00:24:47,800
So when you look at the IPCC, this lack of confidence in the future circulation changes propagates to the impacts.

240
00:24:47,820 --> 00:24:55,270
So this is just because the air 51i couldn't fit on the page anymore, but this was from a slightly older one on extremes.

241
00:24:55,270 --> 00:25:02,500
But if you look at droughts or floods, which are two of the important circulation related impacts, these are observed changes.

242
00:25:02,800 --> 00:25:06,670
This is the Archbishop observed changes and the projected future changes.

243
00:25:07,000 --> 00:25:11,709
And you see that there were some medium confidence statements are actually downgraded to low confidence.

244
00:25:11,710 --> 00:25:15,760
So we got in the error five. So we have low confidence in the past changes.

245
00:25:15,760 --> 00:25:19,930
We have a low confidence in the reasons for that and we have low confidence in the future.

246
00:25:20,350 --> 00:25:23,500
But maybe this is the wrong way to ask the question.

247
00:25:24,850 --> 00:25:32,530
So the normal side of a prior in our business and physics is is no, no effect at least, right?

248
00:25:32,530 --> 00:25:35,530
Well, maybe yeah. I guess that's the normal scientific prior.

249
00:25:35,800 --> 00:25:43,600
So you you try to reject the null hypothesis, but of course, given the magnitude of the natural variability, given the model uncertainty,

250
00:25:44,020 --> 00:25:50,469
given the limited observational record because it's often fairly short, even for the best things,

251
00:25:50,470 --> 00:25:56,290
that's 100 years or so, which is not not nearly enough for the really multidecadal variability.

252
00:25:56,920 --> 00:26:02,620
It may be that you can't reject the null hypothesis, even if even if there was a real signal there.

253
00:26:03,430 --> 00:26:06,760
But of course, we must remember that failure to reject,

254
00:26:06,820 --> 00:26:10,600
reject the null hypothesis doesn't mean that the null hypothesis is true, and that's a mistake.

255
00:26:10,600 --> 00:26:13,810
But I think a lot of people in climate science who seem to make.

256
00:26:14,560 --> 00:26:16,420
But maybe that's the wrong null hypothesis.

257
00:26:16,420 --> 00:26:25,880
We know that there's climate change, so maybe that the null or the prior hypothesis should be the robust aspects of climate change that we know about.

258
00:26:25,930 --> 00:26:29,410
We are very confident about those aspects, the warming and the poisoning.

259
00:26:29,680 --> 00:26:34,360
It seems a bit strange to basically throw it all that knowledge each time that we do a new analysis.

260
00:26:34,870 --> 00:26:38,559
And from a practical point of view, I've just shown examples of time series of things.

261
00:26:38,560 --> 00:26:44,680
If you wait till the observed changes are unambiguous, like they are for temperature, we wait for that for storms, it'll be much too late.

262
00:26:45,310 --> 00:26:48,730
And certainly a reinsurance firm would take a precautionary approach.

263
00:26:48,730 --> 00:26:52,630
So should we shouldn't we be thinking about that for climate science?

264
00:26:53,440 --> 00:27:01,120
So there's what I've called the the storyline approach in this in this paper, and it's what you might call dynamically conditioned.

265
00:27:04,570 --> 00:27:07,640
Attribution, which is a growing theme in the field.

266
00:27:08,170 --> 00:27:17,379
So basically this is a formulation of the National Academy of Sciences in the U.S. produced an extreme report last year, this past year, I should say,

267
00:27:17,380 --> 00:27:23,970
where this is the formulation of a different one in this paper, which is a little more clumsy, actually, this is more elegant version of it.

268
00:27:23,980 --> 00:27:31,580
So you have a a joint probability. So most on, you know, a circulation of a related extreme.

269
00:27:32,440 --> 00:27:34,480
So there'll be, say, a heat wave.

270
00:27:34,500 --> 00:27:44,560
It's normally connected with an insect on a circulation or in the case of the flooding and the heavy rain that was connected with this structure.

271
00:27:45,190 --> 00:27:52,090
So you can express a joint probability of the extreme that you care about flooding or rain or heat or something.

272
00:27:52,300 --> 00:27:59,830
And then C is the C for circulation, but C is the synoptic or the weather situation that's conducive to that extreme.

273
00:28:00,670 --> 00:28:02,090
So this is very trivial.

274
00:28:02,110 --> 00:28:09,909
The the probability of a of A, the joint probability B and C is equal to the conditional probability of A given C times, the probability of sea.

275
00:28:09,910 --> 00:28:14,200
That's just that's just sort of straightforward. And we do this for the counter.

276
00:28:14,200 --> 00:28:19,200
If you do this for the factual of the current climate, P wanted to do this for the counterfactual of you,

277
00:28:19,280 --> 00:28:25,659
not you just take the ratio, you get this very nice Factorisation So what this is saying,

278
00:28:25,660 --> 00:28:33,130
if you want to look at the change in risk of this combination of the of the circulation, that's a conducive to the extreme and the extreme.

279
00:28:33,310 --> 00:28:39,400
And people often look at these joint joint probabilities. It's composed of a product of two terms.

280
00:28:39,400 --> 00:28:47,260
The first is the change in the probability of the event given the synoptic situation.

281
00:28:47,800 --> 00:28:53,170
And then the other is the change is the risk ratio or the change in probability of the circulation.

282
00:28:53,170 --> 00:29:03,850
So this first one is a is a conditional probability ratio and it's really got the purely thermodynamic effects of climate change.

283
00:29:03,850 --> 00:29:08,709
You're saying given the synoptic a situation, how much did the event change?

284
00:29:08,710 --> 00:29:13,240
Maybe because of having more warmer temperatures or more moisture?

285
00:29:13,600 --> 00:29:19,690
And what happens then is that you really tighten up. This is a cartoon, but it's been shown in examples.

286
00:29:19,690 --> 00:29:26,110
Now we get if we're way out in the tail here, you you can't separate these guys in grey in the unconditional.

287
00:29:26,350 --> 00:29:29,350
But now when you a condition all of a sudden you get high signal to noise and

288
00:29:29,350 --> 00:29:35,080
actually what you restore is sufficient causation but in a conditional way.

289
00:29:35,860 --> 00:29:39,759
And then there's this other term, which is the ratio of the circulation.

290
00:29:39,760 --> 00:29:47,110
So the real point about this is that now you can probably do a you could probably compute this first or ratio with with a fair

291
00:29:47,110 --> 00:29:52,989
amount of accuracy because when we have a lot of confidence in the thermodynamic effects of climate change and this other term,

292
00:29:52,990 --> 00:30:01,330
it may be small or it may be highly uncertain, and you should not necessarily ignore it, but you should certainly be treated differently.

293
00:30:01,330 --> 00:30:04,920
And you shouldn't mix it in this because you shouldn't be combining things with different uncertainties.

294
00:30:04,940 --> 00:30:06,850
That's what we argued for in this paper last year.

295
00:30:08,410 --> 00:30:15,819
Now, why would you defend the the argument that we might want to have a null hypothesis of no circulation change?

296
00:30:15,820 --> 00:30:22,840
Well, if you go to the IPCC and you basically that any statement that you pick about a circulation is very, very weak.

297
00:30:22,840 --> 00:30:30,910
This is for it acts to tropical storm tracks, which is what we would care about in terms of, you know, winter time flooding in the Britain.

298
00:30:31,090 --> 00:30:34,960
This low confidence in near term projections of low confidence.

299
00:30:36,670 --> 00:30:41,680
Over the longer term, the global number of cyclones is unlikely to decrease by more than a few percent,

300
00:30:42,490 --> 00:30:50,740
likely to be small compared to a interannual variability. So basically IPCC is saying we have no basis to assume anything other than no change.

301
00:30:51,220 --> 00:30:57,160
And if you assume it, if you find if you have a model that does give you a change, you would have to ask yourself, do I believe the model?

302
00:30:57,160 --> 00:31:00,460
Because basically there's no no confidence in any of those changes.

303
00:31:01,210 --> 00:31:06,610
So I'll give a few examples of how this kind of reasoning works.

304
00:31:08,630 --> 00:31:14,000
This is from a paper a few years back by a French group, so there was a cold winter in 2012.

305
00:31:14,030 --> 00:31:18,440
I guess the climate sceptics said, What about climate change?

306
00:31:18,980 --> 00:31:23,960
So what they did here was they said, Well, how cold would it have been under the circulation regime?

307
00:31:23,970 --> 00:31:31,010
So they used a relationship historical between circulation and temperature and they computed what they call the analogue state,

308
00:31:31,010 --> 00:31:38,090
which is the is the the temperatures that they would have had under the same circulation regime in the past.

309
00:31:38,240 --> 00:31:40,910
And what you see so this is the the observe state,

310
00:31:40,910 --> 00:31:46,370
this is what they estimated would have happened 50 years earlier if they'd had the same configuration.

311
00:31:46,940 --> 00:31:50,780
And the difference shows a warming. So that's an attributable warming.

312
00:31:51,140 --> 00:31:55,040
If you just look at the Time series over 50 years, you certainly can't pick out a trend there.

313
00:31:55,040 --> 00:32:00,470
But if you look at the difference between this analogue state and the actual state, then then you can pick up the trend.

314
00:32:00,830 --> 00:32:04,550
It's showing this signal to noise gets very, very high.

315
00:32:05,360 --> 00:32:15,860
Once you condition on the circulation, this is another example, a recent study of Hurricane Sandy which led to flooding in New York City.

316
00:32:16,250 --> 00:32:19,940
So, of course, a hurricane is something that Clinton models calculated.

317
00:32:19,940 --> 00:32:28,370
But the weather models can you you may know that Hurricane Sandy was forecast quite accurately about ten days ahead by the European Centre.

318
00:32:28,790 --> 00:32:31,490
So this is what the American model called morph, which is a weather model.

319
00:32:31,790 --> 00:32:37,399
And you can forecast the hurricane and then you can ask how would a hurricane have been if the sea surface

320
00:32:37,400 --> 00:32:43,110
temperatures had been cooler as they would have been in the past or warmer as they would be in the future?

321
00:32:43,130 --> 00:32:51,440
And this is the Time series. This is an hour of a run over over fourth of four days.

322
00:32:51,740 --> 00:32:55,610
So you see that the surface pressure goes down. This is the minimum surface pressure.

323
00:32:55,970 --> 00:33:02,750
This was the actual well, the observed state when in the in the Dawson model ensemble under president conditions.

324
00:33:02,750 --> 00:33:09,110
This is rerunning under a conditions of 100 years ago and this is 100 years in the future.

325
00:33:09,110 --> 00:33:17,300
And you can argue that the hurricane has was a bit stronger than it would have been a hundred years ago and it will be much stronger in the future.

326
00:33:17,350 --> 00:33:23,990
This is another conditional attribution where you do it using forecasts so you stay close to the observed state.

327
00:33:24,350 --> 00:33:30,379
Another example the California drought has been a very source thing in the U.S., has been a lot of papers on this, a lot of debate.

328
00:33:30,380 --> 00:33:32,720
Of course, it's a big issue for poor California.

329
00:33:34,670 --> 00:33:43,040
I like this paper, which are basically you can you can argue that drought depends on a combination of temperatures and and precipitation.

330
00:33:43,040 --> 00:33:48,319
You have, of course, low precipitation, but it tends to be exacerbated by hot high temperatures.

331
00:33:48,320 --> 00:33:51,140
And in California, that's because of the melting of the snowpack.

332
00:33:51,860 --> 00:33:58,189
So if you look at the at a precipitation time series over 100 years, you can see that it's up and down and up and down.

333
00:33:58,190 --> 00:34:03,590
And these are low values here and there's not no apparent trend. If you look at the temperature, it's clear that there is a trend here.

334
00:34:03,860 --> 00:34:08,089
This Palmer Drought Index is is some synthesis of these two different factors.

335
00:34:08,090 --> 00:34:14,180
People argue over it, but it's it's a combination of the two and you can see that there's been going down means more drought.

336
00:34:14,630 --> 00:34:20,240
So the argument here is that if you look at the long term record over a hundred years,

337
00:34:20,240 --> 00:34:29,510
you see that if you ask whether the temperature was high or so warmer or cooler and the precipitation that your high or low basically,

338
00:34:29,510 --> 00:34:34,010
of course, it's evenly spread around the mean. But if you look in the last 20 years,

339
00:34:34,310 --> 00:34:41,690
you can see that the that the probability of both being both dry and warm at the same time has increased a lot because of the warming.

340
00:34:42,350 --> 00:34:47,059
So if you ask what's going to happen to a precipitation in California, it's highly uncertain.

341
00:34:47,060 --> 00:34:51,320
It depends on the storm track that that comes in. There's a lot of uncertainty about the storm track changes.

342
00:34:51,740 --> 00:34:54,860
You really you really wouldn't want to trust models for that.

343
00:34:55,280 --> 00:35:02,509
So what you can say is that we don't know what's going to happen to to to precipitation storms, but we do know it'll get warmer.

344
00:35:02,510 --> 00:35:06,150
So the risk of drought has has increased. So again, it's a conditional statement.

345
00:35:06,830 --> 00:35:09,409
This is another example that's a little bit different.

346
00:35:09,410 --> 00:35:21,590
But L.A., because it really to me emphasises the signal to noise when you condition there was a course, a terrible heat wave in France in 2003.

347
00:35:22,040 --> 00:35:27,700
These are images, maps that this scale is 500 metres.

348
00:35:27,710 --> 00:35:32,780
These are three kilometres by three kilometres in central France.

349
00:35:33,260 --> 00:35:36,409
The top panel shows a vegetation index from satellite.

350
00:35:36,410 --> 00:35:41,480
This the bottom is surface temperature. So this is a grid box of three kilometres by three kilometres.

351
00:35:41,840 --> 00:35:47,149
So what you see is this is one day in August 2000, one day in August 2003.

352
00:35:47,150 --> 00:35:52,070
And you would say, how can you compare one day in 2000 with one day in 2003?

353
00:35:52,370 --> 00:35:58,519
Of course, there's synoptic variability, but the synoptic variability is zero over a three kilometre scale.

354
00:35:58,520 --> 00:36:03,200
So you can think of it as is basically that we have a uniform state over the grid blocks and

355
00:36:03,200 --> 00:36:07,550
what happened the red means that there's a vegetation and the remains of that there is an.

356
00:36:07,620 --> 00:36:14,669
So what happened in 2003 is all the passengers and crew all claims to have died out.

357
00:36:14,670 --> 00:36:20,580
So the red became green, but the forest is still there, so the red is still red.

358
00:36:21,090 --> 00:36:27,180
And if you look at the warming between here and here in the forest, it's 11 degrees C,

359
00:36:27,480 --> 00:36:33,440
but in the pastures it's 20 degrees C, and if you isolate to avoid the hedgerows, it's 24 degrees.

360
00:36:34,200 --> 00:36:37,889
So this is the impact of the vegetation on the surface temperature.

361
00:36:37,890 --> 00:36:42,930
And I'm just it's the signal to noise as you go across the sky is to me amazing.

362
00:36:42,930 --> 00:36:46,260
So it's the same kind of philosophy that we're talking about here.

363
00:36:46,270 --> 00:36:50,670
A few condition on a synoptic regime. You can get very high signal to noise.

364
00:36:50,970 --> 00:36:54,270
In this case, it's not the climate change that we're talking about.

365
00:36:54,270 --> 00:36:57,930
We're talking about the the land use effects on the on the extreme.

366
00:36:58,920 --> 00:37:06,239
Okay. So getting back to the to to to the stock jet stream and the flooding Commonwealth is

367
00:37:06,240 --> 00:37:11,160
actually do predict a strengthening of the winter time jet and the storm track over Britain.

368
00:37:11,610 --> 00:37:15,900
This is plots from up ahead. So if you take all the different climate models,

369
00:37:16,200 --> 00:37:23,100
what they predict is going to happen at the end of the 21st century under under a particular scenario, this is a number of storms, basically.

370
00:37:23,370 --> 00:37:31,429
And what you see, blue means less. And so we what this is predicting is that the Iceland gets some relief in a bit.

371
00:37:31,430 --> 00:37:37,170
But if you're storms and and also fewer storms in the Mediterranean but more more storms over Britain.

372
00:37:38,340 --> 00:37:42,360
But the problems here, first of all, we have very large model biases.

373
00:37:42,360 --> 00:37:47,069
So this is the the bias in the models. And you can see just the different scale here.

374
00:37:47,070 --> 00:37:50,130
So the bias is four times as large as the signal.

375
00:37:50,820 --> 00:37:54,990
So you generally would be very nervous about as a signal.

376
00:37:55,200 --> 00:37:58,650
That's one quarter of the size of your bias in a nonlinear system.

377
00:37:59,280 --> 00:38:02,880
There's no mechanisms for for the strengthening jetstream.

378
00:38:03,050 --> 00:38:10,440
No, no, no one really has any theory for that. Certainly there's no theory that everybody would agree and it's not seen in the observations.

379
00:38:10,440 --> 00:38:16,019
So if we don't have an observed signal, if we don't have a theory, if models have huge biases, of course we have low confidence.

380
00:38:16,020 --> 00:38:21,090
And that's why the IPCC gives us gives us this prediction, very low confidence.

381
00:38:22,530 --> 00:38:29,939
Now, this this has an extension. So the strengthened storm track is reflected in extension of the jet over Central Europe.

382
00:38:29,940 --> 00:38:36,419
And if you think about a signal to noise that's often quantified in what's called the time of emergence,

383
00:38:36,420 --> 00:38:42,360
so it's when does the when does the climate change signal become manifest in the observable record?

384
00:38:42,780 --> 00:38:48,660
When will with a single realisation. So this will be the sufficient causation and you can see that.

385
00:38:49,440 --> 00:38:53,249
So this is by year and you can see the most of this is, is, is white.

386
00:38:53,250 --> 00:39:00,990
So it's basically well, well past the end of the century. So temperature changes are manifest very quickly everywhere.

387
00:39:01,260 --> 00:39:06,989
But circulation changes are actually seem like they're going to be very small, almost everywhere.

388
00:39:06,990 --> 00:39:09,180
And there's a couple of and they're going to be very localised.

389
00:39:09,180 --> 00:39:17,640
So what you see over Europe is that there were there's there's a predicted changes in the winter time over Central Europe and over North Africa.

390
00:39:18,930 --> 00:39:25,290
And so I think that's the first important point is that the the circulation changes that are going to merge are probably very localised.

391
00:39:26,460 --> 00:39:38,130
If you look at the Central Europe case, although the models agree that there will be a strengthening of the jet, when a well will happen varies a lot.

392
00:39:38,130 --> 00:39:39,030
So some of the models,

393
00:39:39,450 --> 00:39:47,940
the models with the strongest response say that we might begin to see this signal emerging and the observations lie about in about 50 years.

394
00:39:48,270 --> 00:39:55,830
Others say it'll be after the end of the century. So there's a huge uncertainty in in the strength of that response.

395
00:39:56,460 --> 00:40:04,830
One place where you do see a strong circulation, that response really the biggest is in this area is is is to do with the Mediterranean.

396
00:40:05,940 --> 00:40:09,120
And so drying the models do I won't go back to it.

397
00:40:09,120 --> 00:40:13,980
But if you look at the precipitation maps that I showed in the first second slide probably or something,

398
00:40:14,880 --> 00:40:19,590
the one place where there's a stippling over populated areas is the Mediterranean.

399
00:40:20,010 --> 00:40:26,190
And so the models are consistently predicting a drawing of the Mediterranean, which of course will be very serious for Europe.

400
00:40:27,810 --> 00:40:30,959
So if you look at the time of emergence in the precipitation,

401
00:40:30,960 --> 00:40:40,470
now you see that that is that it's that the time of emergence may be much earlier, like even within the next ten years.

402
00:40:40,470 --> 00:40:45,150
And even the weakest response is the middle of the century.

403
00:40:46,470 --> 00:40:49,350
So this is a place where circulation is really the big story.

404
00:40:49,860 --> 00:40:56,010
And if you look at the at the drawing, it really is is associated with a circulation change.

405
00:40:56,400 --> 00:41:00,090
So this shows the precipitation response and this is the wind.

406
00:41:00,090 --> 00:41:04,620
This is the other 50 is the solar wind in the lower troposphere so close to the surface.

407
00:41:05,160 --> 00:41:10,480
And if you look so this is the average of all. The models. These are the models with the strongest drawing.

408
00:41:10,490 --> 00:41:15,670
These are the models with the weakest drawing. And you see it's highly correlated with with the circulation.

409
00:41:15,670 --> 00:41:19,330
And we use the the window of a North Africa as an index for that.

410
00:41:20,950 --> 00:41:26,349
And so if you look at the relationship between this circulation index, the wind over North Africa, solar,

411
00:41:26,350 --> 00:41:30,690
wind and the precipitation, if you look in the in the observations there, they're highly correlated.

412
00:41:30,700 --> 00:41:34,960
So the year to year variability is a very strong relationship.

413
00:41:35,560 --> 00:41:43,420
If you look at the models, the the black is the models under under historical conditions, the red is the models under future climate change.

414
00:41:43,420 --> 00:41:47,290
And basically what's happening is the models are just moving down this down this line.

415
00:41:47,740 --> 00:41:53,110
So what that saying is that they're forming the same relationship between circulation and precipitation.

416
00:41:53,410 --> 00:41:57,670
It's just becoming more extreme and it's controlled by the circulation.

417
00:41:57,670 --> 00:41:59,140
There is a slight drying as well.

418
00:41:59,380 --> 00:42:05,650
And it depends a little bit on these are individual models, but it's mainly coming from the shift in the circulation.

419
00:42:06,130 --> 00:42:15,790
And so if you correlate the precipitation, the response across the different models with which which is always a drying with the wind change,

420
00:42:15,790 --> 00:42:24,649
you see there's a very strong relationship. And so if it turns out that the models with the weakest response are correct,

421
00:42:24,650 --> 00:42:30,100
the by the end of the century, it turns out we'll have a shift that's like one standard deviation.

422
00:42:30,100 --> 00:42:35,380
But if the models with the strongest response are correct, they're going to be very well separated.

423
00:42:35,860 --> 00:42:43,640
So it turns out that 85% of the mean precipitation change and 80% of the model spread is related to changes in the circulation.

424
00:42:43,660 --> 00:42:50,860
So it's the same we're not going to be able to place any constraints on this cold season drawing unless we get to grips with the circulation response.

425
00:42:52,360 --> 00:42:57,219
So why do the models differ so much in the in this response?

426
00:42:57,220 --> 00:43:01,450
Those this is just a clue. It's something that we're working on at the moment.

427
00:43:02,020 --> 00:43:07,839
This is an analysis of the different models. The sign up is the archive used for the IPCC.

428
00:43:07,840 --> 00:43:11,829
How much done here is to partition? This is the mean sea level pressure.

429
00:43:11,830 --> 00:43:19,090
So it's a circulation index and basically say how much of the model uncertainty comes from uncertainty and the tropical warming,

430
00:43:19,090 --> 00:43:21,160
which has to do with with the climate sensitivity,

431
00:43:21,640 --> 00:43:28,360
how much has to do with the Arctic amplification, how much has to do with the stratosphere, which is also changing under climate change?

432
00:43:28,870 --> 00:43:33,339
And you see that, first of all, the conch, it's the same number of contours, more or less on each plot.

433
00:43:33,340 --> 00:43:38,830
So it's a relative plot, but it's saying that these three factors are equally important in terms of the model spread.

434
00:43:39,280 --> 00:43:43,030
So not necessarily for the total change, but for the uncertainty.

435
00:43:43,570 --> 00:43:48,400
And the tropical warming tends to push the jet poleward as people have seen in lots of studies,

436
00:43:48,670 --> 00:43:53,799
the Arctic warming tends to push it back and the stratosphere, if it warms, tends to push it back as well.

437
00:43:53,800 --> 00:43:58,840
And these these kind of patterns people have found in single forcing studies,

438
00:43:59,140 --> 00:44:03,760
in the case of the stratosphere, it's it's also what happens in the risk of some warming.

439
00:44:03,760 --> 00:44:07,450
So I think this all makes a lot of sense. So you can argue that there's a tug of war.

440
00:44:07,480 --> 00:44:17,799
These are telling actions really between that the the mid-latitude circulation is strongly affected by but by telling connections from other regions.

441
00:44:17,800 --> 00:44:22,780
And these are uncertainties in the climate response which are exerting a tug of war.

442
00:44:22,780 --> 00:44:27,310
And whenever you have two things acting in opposite directions, you can get uncertainty.

443
00:44:27,700 --> 00:44:36,549
So just just to give an example, I think this is why we have such differences on what the models are, projects under climate change.

444
00:44:36,550 --> 00:44:39,280
So this is just for different models of the end of the century.

445
00:44:39,280 --> 00:44:45,040
And you see that this is this this is this 852 scale, this lower troposphere, it's all wind.

446
00:44:45,040 --> 00:44:48,580
And you see some models show a strengthening over Central Europe.

447
00:44:49,150 --> 00:44:53,830
This model that doesn't show anything of that, the shows a strengthening over the Mediterranean and so on.

448
00:44:53,830 --> 00:44:59,530
So there's models that are doing very different things, which is probably because of this tug of war.

449
00:45:00,070 --> 00:45:04,209
So to summarise the growing emphasis on first of all,

450
00:45:04,210 --> 00:45:09,280
there's a growing emphasis on regional climate and there's, of course, on also near term projections.

451
00:45:09,280 --> 00:45:18,790
People want to know what might happen in 20 or 30 years is really exposing the limitations in our understanding of atmospheric circulation.

452
00:45:18,790 --> 00:45:24,180
And if you go to the IPCC, you start to find very weak statements with very large uncertainties and very

453
00:45:24,250 --> 00:45:28,360
low confidence once you get to to to the regional scale and the short term.

454
00:45:29,470 --> 00:45:39,160
So that put the signal to noise of a projected changes in circulation is small in general, but that doesn't mean that the risk is small.

455
00:45:39,160 --> 00:45:43,870
So we actually can't ignore this and we have to develop a scientific language,

456
00:45:43,870 --> 00:45:50,409
something incorporating risk that will express these uncertainties, but without losing sight of what we know.

457
00:45:50,410 --> 00:45:55,239
So this is what I for example, this factorisation of the uncertainties is a way of doing this.

458
00:45:55,240 --> 00:45:59,260
If you just put everything in the bin, then you just say, well, we we don't know.

459
00:45:59,260 --> 00:46:05,169
We have low, low confidence, but that it may just be that we don't have confidence about circulation, but we have confidence about other things.

460
00:46:05,170 --> 00:46:07,090
So we somehow want to be able to talk about.

461
00:46:07,570 --> 00:46:16,059
I think a lot of climate scientists are afraid to talk about uncertainties because they figure it'll give the sceptics too much from the material.

462
00:46:16,060 --> 00:46:20,230
But we have to find ways to talk about uncertainty without losing sight of what we know.

463
00:46:21,820 --> 00:46:27,040
Wintertime circulation chains over Central Europe are highly uncertain in this,

464
00:46:27,220 --> 00:46:31,630
probably because of a tug of war between these different drivers and certainly has

465
00:46:32,170 --> 00:46:38,860
implications for important climate change impacts like windstorms and Mediterranean drying.

466
00:46:39,250 --> 00:46:49,760
And in a way, to me, the fundamental challenge of climate science is that we need to talk about we need to we have model projections of the future.

467
00:46:49,780 --> 00:46:55,959
A projection is just a conditional prediction based on an assumed scenario of greenhouse gases.

468
00:46:55,960 --> 00:46:59,140
So we need to use those models to get information.

469
00:46:59,830 --> 00:47:03,420
But the only observations we have are in the past.

470
00:47:03,430 --> 00:47:10,240
The classic scientific paradigm is your theory. You make experiments, you test the theory, you go back and you go back and forth.

471
00:47:10,720 --> 00:47:21,640
We can't do that in a simple way because the space that we're making the observations in isn't the space that we're making, but the predictions.

472
00:47:21,640 --> 00:47:27,220
And so I think this really challenges the classic deduction and side of the paradigm.

473
00:47:27,550 --> 00:47:31,209
And certainly I think one of the key fields for for the future will be trying to how do

474
00:47:31,210 --> 00:47:35,620
we get information out of the short timescales in order to constrain long timescales?

475
00:47:35,850 --> 00:47:37,270
Thanks very much.