1
00:00:02,430 --> 00:00:06,870
OK. Welcome everyone.

2
00:00:06,870 --> 00:00:16,350
My name is Tim Palmer, I'm at the physics department here, and I'm going to say a few words of welcome to our speaker, Typekit Schneider.

3
00:00:16,350 --> 00:00:25,740
I should explain that we've been having, I think, a fairly successful workshop on the application of artificial intelligence,

4
00:00:25,740 --> 00:00:36,210
machine learning all that stuff into to improving climate modelling and Typekit, as well as real well experts at that.

5
00:00:36,210 --> 00:00:43,170
So the workshop was kindly, very kindly co-sponsored by the Oxford Martin School.

6
00:00:43,170 --> 00:00:51,030
So we kindly agreed to give a talk here this afternoon after the workshop finished.

7
00:00:51,030 --> 00:00:56,580
So it's happened in his Ph.D. at Princeton University.

8
00:00:56,580 --> 00:01:03,510
He had a spell as professor at Swarek, but has returned to the US,

9
00:01:03,510 --> 00:01:08,880
where he's currently a professor of environmental science and engineering at

10
00:01:08,880 --> 00:01:16,500
Caltech and also a senior research fellow for NASA's Jet Propulsion Laboratory.

11
00:01:16,500 --> 00:01:24,480
Tucker, I was, you will find out, is very much one of our world experts in the interaction between clouds and more generally,

12
00:01:24,480 --> 00:01:28,710
the hydrological cycle with the atmospheric circulation patterns.

13
00:01:28,710 --> 00:01:36,060
And if there was a single problem that really is critically important for understanding

14
00:01:36,060 --> 00:01:40,830
climate change and what direction we're heading to and how bad things will be,

15
00:01:40,830 --> 00:01:46,310
it's understanding this interaction between clouds and circulation.

16
00:01:46,310 --> 00:01:54,050
Typekit has got numerous awards, which I went I went embarrassing with entirely embarrass, embarrassing with one, actually.

17
00:01:54,050 --> 00:02:03,230
He was named as one of the 20 top brains of people under the age of 40 by Discovery Magazine.

18
00:02:03,230 --> 00:02:08,900
So that isn't embarrassing. I don't know what is, but so.

19
00:02:08,900 --> 00:02:15,590
And he's also won the James Holtmann Award for the American Geophysical Union and various other awards, which have, as I say, I won't go into.

20
00:02:15,590 --> 00:02:27,430
So can we just give a quick round applause and welcome to temptation?

21
00:02:27,430 --> 00:02:32,140
Thank you, Tim, and thanks for coming. So indeed, I'll tell you the story of clouds and climate,

22
00:02:32,140 --> 00:02:36,460
and I'll talk a fair bit about some of my favourite clouds, which should be your favourite too,

23
00:02:36,460 --> 00:02:43,900
because of the most frequent on earth, stunning cumulus clouds, the ones that you see on this, the satellite image from a space station.

24
00:02:43,900 --> 00:02:48,310
And I'll come back to studying cumulus clouds in just a little bit.

25
00:02:48,310 --> 00:02:52,120
But the story really starts in the cold place and Ellis Island.

26
00:02:52,120 --> 00:02:57,820
This is an island in the Canadian Arctic archipelago. It's an Arctic island.

27
00:02:57,820 --> 00:03:03,070
There are. There are flowers like the river beauty flower growing there.

28
00:03:03,070 --> 00:03:07,550
But the largest plants you'll find on this island right now is about two foot tall.

29
00:03:07,550 --> 00:03:13,600
Two Willow. I think Tyler grows there and it's fairly barren, but it was.

30
00:03:13,600 --> 00:03:17,770
It wasn't always like that about 50 million years ago.

31
00:03:17,770 --> 00:03:23,380
This is roughly what Ellesmere Island looked like. Once upon a time, it was lush and warm.

32
00:03:23,380 --> 00:03:31,060
There were a hippopotamus grazing on large firms that were top years peak like herbivores in the corner of the image.

33
00:03:31,060 --> 00:03:37,330
They're grazing along their deciduous trees, conifers and the like.

34
00:03:37,330 --> 00:03:48,730
And we know that because scientists have found fossils there, fossils of leaves larger than 10 centimetres tall or the jaw bones of tough years.

35
00:03:48,730 --> 00:03:52,750
They are from crocodile bones. There are crocodiles and Ellesmere Island.

36
00:03:52,750 --> 00:04:01,870
Fifty million years ago and don't need to know much about crocodiles or climate, but you probably can appreciate that they don't like it cold.

37
00:04:01,870 --> 00:04:13,270
In fact, they can't deal with frost. So we know that Ellesmere Island, the Arctic was frost free and that you've seen from 50 some million years ago.

38
00:04:13,270 --> 00:04:19,480
Now, 50 million years ago is a long time. And maybe sometimes hard to put these geological timescales in perspective.

39
00:04:19,480 --> 00:04:25,610
So does this post dinosaurs, but well before humans? And you can ask, Oh, maybe I was.

40
00:04:25,610 --> 00:04:32,470
It was a tropical island, and it wasn't. So here is a reconstruction of what continent look like at around the time.

41
00:04:32,470 --> 00:04:34,280
Ellesmere Island was a red star.

42
00:04:34,280 --> 00:04:42,520
It's right next to Greenland, as it is now is as much an Arctic island in the Eocene 50 million years ago as it is now.

43
00:04:42,520 --> 00:04:46,600
So there were an trees and ferns and all these things because the island migrated.

44
00:04:46,600 --> 00:04:52,570
There were trees in Ferguson and crocodiles because it was very different.

45
00:04:52,570 --> 00:04:56,730
We have some idea of how climate was different from.

46
00:04:56,730 --> 00:05:05,700
Looking at the shells of tiny marine organisms that fossilised marine sediments and from decomposition of the carbonate and the shells,

47
00:05:05,700 --> 00:05:09,750
we can tell to some degree how warm it was.

48
00:05:09,750 --> 00:05:18,890
So here is an attempt to reconstruct the last sixty five million years of the temperature history of Earth and.

49
00:05:18,890 --> 00:05:24,860
First thing, maybe you noticed this in week also that the climate has varied a lot.

50
00:05:24,860 --> 00:05:29,960
If you look at the scale on the vertical axis, assisted temperature change relative to today.

51
00:05:29,960 --> 00:05:37,730
So in the U.S., in early Eocene, fifty five million years ago, it was about 10 degrees warmer than it is now.

52
00:05:37,730 --> 00:05:43,140
We don't know the numbers yet, very precisely, but about 10 degrees is a pretty good estimate.

53
00:05:43,140 --> 00:05:49,640
And by and large, climate has been calling for the last 65 million years I've been gradually sliding downward.

54
00:05:49,640 --> 00:05:54,140
The temperature scale was a gradual, slight levelling off.

55
00:05:54,140 --> 00:06:01,190
There are some, some bumps and spikes in between. For example, there's this one spike you see at the forty three million years ago.

56
00:06:01,190 --> 00:06:03,110
It's that you're seeing climatic optimum.

57
00:06:03,110 --> 00:06:13,100
The temperature spiked for geologically brief period, which is still millions of years, and then went down again or even earlier.

58
00:06:13,100 --> 00:06:21,440
There's the Palio seen using thermal maximum of five or six degree temperature spike on top of an already very warm climate.

59
00:06:21,440 --> 00:06:28,310
The reasons for these spikes are for thermals, as they're called are not very well understood.

60
00:06:28,310 --> 00:06:35,290
It is clear they have something to do with greenhouse gases, but how exactly they came about is unclear.

61
00:06:35,290 --> 00:06:43,270
Just to orient you where we are Homo sapiens, we we appear on the far right of this plant about 300000 years ago.

62
00:06:43,270 --> 00:06:45,370
Temperatures have been decreasing at some point.

63
00:06:45,370 --> 00:06:54,220
Ice sheets started to form for a standard article because there is a land Antarctic ice sheets started to solidify around 10 million years ago.

64
00:06:54,220 --> 00:06:56,470
Ice sheets in the northern hemisphere are more recent.

65
00:06:56,470 --> 00:07:04,180
Around three million years ago that we started to form ice sheets in the northern hemisphere simply because there's no continent at the North Pole,

66
00:07:04,180 --> 00:07:09,100
and we don't know a lot about what drives these temperature changes,

67
00:07:09,100 --> 00:07:13,900
but we do know that the principal driver over these timescales are changes in the composition

68
00:07:13,900 --> 00:07:18,850
of the atmosphere and principally in the greenhouse gas concentration in the atmosphere.

69
00:07:18,850 --> 00:07:26,110
So here is a reconstruction of carbon dioxide and the atmosphere over these timescales.

70
00:07:26,110 --> 00:07:32,770
And well, first thing to note is this temperature was weakly and is somewhat uncertain.

71
00:07:32,770 --> 00:07:37,870
But the CO2 reconstructions are really uncertain. We don't know them very well.

72
00:07:37,870 --> 00:07:45,760
And that even if you look at any given time, you see a lot of scatter and the reconstructed CO2 values they come from such things as fossilised soils,

73
00:07:45,760 --> 00:07:52,790
where you can try to infer how much CO2 was an atmosphere based on the composition of the soil.

74
00:07:52,790 --> 00:08:01,160
It doesn't work all that well, but it works well enough that we can say that again, if you go back to the crocodile in the Arctic period,

75
00:08:01,160 --> 00:08:09,890
fifty five million years ago or so, CO2 concentrations were maybe 6500 parts per million at most.

76
00:08:09,890 --> 00:08:16,400
Most data seems to suggest they were lower than that, but there's a lot of uncertainty in it.

77
00:08:16,400 --> 00:08:24,770
But you can see that CO2 concentrations went down as the climate cooled from perhaps 6500 or so parts per million to around 400 parts per million.

78
00:08:24,770 --> 00:08:30,500
The Pliocene so three to five million years ago, 410 parts per million is just about where we are now.

79
00:08:30,500 --> 00:08:37,340
Come back to that number. So in the early years, maybe there was four times as much CO2 in the atmosphere.

80
00:08:37,340 --> 00:08:41,390
Maybe not even we don't quite know. But it was about 10 degrees warmer.

81
00:08:41,390 --> 00:08:43,610
And again, this number, we don't know precisely.

82
00:08:43,610 --> 00:08:50,030
But we do know that the Arctic was frustrated because there were crocodiles and to make the Arctic frost free,

83
00:08:50,030 --> 00:08:58,280
the entire globe would have to have been warm. So there are two numbers we can use here to do some sort of magic physics, if you like.

84
00:08:58,280 --> 00:09:05,810
So there's about four times as much CO2 in the early Eocene than in a Pliocene, and I'm picking the Pliocene here advisedly.

85
00:09:05,810 --> 00:09:11,500
I'm not comparing with the present day for a reason I'll come to come to in a moment.

86
00:09:11,500 --> 00:09:16,090
And temperatures were about 10 degrees warmer in the early U.S. and then apply as seen.

87
00:09:16,090 --> 00:09:21,940
I'm picking the Pliocene because I want to estimate roughly how sensitive the climate

88
00:09:21,940 --> 00:09:27,520
system is with respect to perturbations and greenhouse gas concentrations CO2.

89
00:09:27,520 --> 00:09:35,650
And if you pick a later time, you have ice sheets changing change the albedo or is that make these estimates a lot more complicated and the Pliocene?

90
00:09:35,650 --> 00:09:40,420
They play some role, but they're relatively minor role compared to the greenhouse gases.

91
00:09:40,420 --> 00:09:48,970
So we can take these two numbers about 10 degrees warming for quadrupling CO2 and combine them into something we call the climate sensitivity.

92
00:09:48,970 --> 00:09:57,820
The climate sensitivity is the way it's defined as the global mean temperature change you get in response to doubling CO2 concentrations.

93
00:09:57,820 --> 00:10:01,450
So quadrupling as twice and doubling and for twice doubling.

94
00:10:01,450 --> 00:10:07,090
We had about 10 degrees global warming or so. And all these numbers to be taken with a grain of salt is really the order of magnitude.

95
00:10:07,090 --> 00:10:15,280
Physics is not accurate even to the first digit and definitely not to any digits after the point here.

96
00:10:15,280 --> 00:10:22,390
But when you get out of this, if you combine the two, is that the climate sensitivity to get from the Pliocene to the U.S. from the U.S.

97
00:10:22,390 --> 00:10:27,580
into the Pliocene must have been something like five degrees or two by 10 degrees,

98
00:10:27,580 --> 00:10:35,050
but two doubling by two and must be around five degrees. And that's what the data seem to suggest.

99
00:10:35,050 --> 00:10:40,690
It might well be higher because CO2 may well not have been four times as high in the U.S.,

100
00:10:40,690 --> 00:10:44,590
but less than that, but at least around five degrees it has to have been.

101
00:10:44,590 --> 00:10:48,220
And here's the problem. If you look at the climate models today,

102
00:10:48,220 --> 00:10:56,710
and here are the twenty nine climate models for which the data were available from the most recent IPCC report, here is their climate sensitivity.

103
00:10:56,710 --> 00:11:05,590
It ranges from two to almost five degrees, not quite five degrees, but there is no model with the climate sensitivity higher than five degrees.

104
00:11:05,590 --> 00:11:11,680
And that assessment that's about to change with the next assessment that's coming out.

105
00:11:11,680 --> 00:11:14,560
But what this means is that these models just looking at these data alone,

106
00:11:14,560 --> 00:11:18,040
you think have a hard time reproducing and using climate if you just give it the

107
00:11:18,040 --> 00:11:21,280
right continents and the composition of the atmosphere as best as we know it,

108
00:11:21,280 --> 00:11:27,940
and even change the ice sheets as best as we know them. You would think it will be difficult for these models to reproduce.

109
00:11:27,940 --> 00:11:31,480
The first three are taken. In fact, that's the case a number of people have tried to do.

110
00:11:31,480 --> 00:11:38,230
That were at Purdue University, for example, but whenever they have tried to simulate and eocene climate,

111
00:11:38,230 --> 00:11:45,580
they could succeed, but only if they increase CO2 concentrations to something like 4000 parts per million.

112
00:11:45,580 --> 00:11:53,080
It was less than that there was frost in the Arctic. It was not consistent with the data, but we know it wasn't for a thousand parts per million.

113
00:11:53,080 --> 00:11:57,620
It was somewhere maybe 600 parts per million, maybe less, but not for a thousand.

114
00:11:57,620 --> 00:12:09,430
We can't exclude that. So the models are not sensitive enough to perturbations as a suggestion here that you get from looking over Earth's history.

115
00:12:09,430 --> 00:12:15,010
And the other thing you see from this graph here is that this kind of sensitivity varies.

116
00:12:15,010 --> 00:12:21,460
A lot varies by almost a factor of two from model to model, meaning there are large uncertainties.

117
00:12:21,460 --> 00:12:25,750
And the primary driver of these uncertainties are those clouds.

118
00:12:25,750 --> 00:12:33,190
And I think they may also hold the answer to what that missing X Factor and global warming and do is Eocene loss to the clouds.

119
00:12:33,190 --> 00:12:41,560
I'm talking about these guys on the left or strata cumulus clouds that cover vast areas of tropical oceans.

120
00:12:41,560 --> 00:12:47,410
It's the most frequent cloud type on Earth. They cover about 20 percent of the tropical oceans.

121
00:12:47,410 --> 00:12:57,510
This is off the coast of California, Baja California. I live in the upper left quadrant of this image and.

122
00:12:57,510 --> 00:13:04,840
If you fly from, say, Los Angeles. Goes here to the Hawaiian Islands, which are on the right.

123
00:13:04,840 --> 00:13:09,160
You look down and this is, by the way, how I got interested in this problem.

124
00:13:09,160 --> 00:13:14,140
Looking out of aeroplane windows and thinking we ought to be able to explain a wet blanket over the ocean.

125
00:13:14,140 --> 00:13:19,090
What happens is as you fly towards the Hawaiian Islands at some point, actually fairly suddenly,

126
00:13:19,090 --> 00:13:22,480
these whites try to cumulus clouds give way to more scattered cumulus clouds.

127
00:13:22,480 --> 00:13:28,060
These white clouds you see around the Hawaiian islands and where you have these cumulus clouds,

128
00:13:28,060 --> 00:13:31,930
there's more dark ocean surface exposed to less sunlight is being reflected.

129
00:13:31,930 --> 00:13:36,130
Meaning it's warmer the way strutting cumulus clouds reflect a lot of sunlight.

130
00:13:36,130 --> 00:13:38,200
They call the ocean underneath.

131
00:13:38,200 --> 00:13:44,950
And the fundamental problem we have inside modelling is that we don't know if we get more strident cumulus or cumulus or neither of the above,

132
00:13:44,950 --> 00:13:49,330
as the Climate Forum's climate models give wildly divergent answers.

133
00:13:49,330 --> 00:13:54,370
So if you go back to this graph shows the climate sensitivity for twenty nine models,

134
00:13:54,370 --> 00:14:00,070
the models on the left that have a relatively low climate sensitivity tend to produce more low clouds as the climate warms,

135
00:14:00,070 --> 00:14:04,960
meaning they reflect more sunlight. And that reduces the warming you get,

136
00:14:04,960 --> 00:14:12,870
whereas the models on the right tend to produce fewer low clouds that reflect less sunlight and that amplifies the warming.

137
00:14:12,870 --> 00:14:20,260
So that's the principle difference or the principle driver of uncertainties in climate predictions, even for the next few decades,

138
00:14:20,260 --> 00:14:26,850
and it's certainly one big uncertainty when you go back millions of years to where clouds hard, the various ways of thinking about it.

139
00:14:26,850 --> 00:14:33,750
I think one way I find intuitive and helpful is this If you take all the water in the atmosphere,

140
00:14:33,750 --> 00:14:39,900
everything vapour condensed water does bring it to the surface as a liquid layer of water.

141
00:14:39,900 --> 00:14:46,200
You get a liquid layer that's twenty five millimetres stakes for America to spot an inch thick, right?

142
00:14:46,200 --> 00:14:52,750
It's easy to remember. So all the water in the atmosphere is the layer that thick on the ground.

143
00:14:52,750 --> 00:15:00,230
That's not much the water is a trace constituent. But now the the fact that surprises many people,

144
00:15:00,230 --> 00:15:09,140
even in our field is if you ask how much of that water is actually in clouds as opposed to being vapour.

145
00:15:09,140 --> 00:15:19,460
It is exceedingly little like clouds. It's about two and 50 times more water in vapour form than it is in condensed form droplets or ice crystals.

146
00:15:19,460 --> 00:15:24,110
So you can take all the cloud water and bring that to the surface as a liquid layer.

147
00:15:24,110 --> 00:15:30,380
You get a liquid layer at 100 microns. Then it's the thickness of a human hair of a coat of paint here.

148
00:15:30,380 --> 00:15:37,940
So now imagine you're a climate model and you have to predict the tiny, tiny residual of a trace constituent of the atmosphere.

149
00:15:37,940 --> 00:15:43,340
Predict how much of that condenses as air moves around. That's a really challenging task.

150
00:15:43,340 --> 00:15:49,400
Of course, clouds form as air masses rise. As they rise, their cool water reaches, saturation condenses,

151
00:15:49,400 --> 00:15:54,380
and you need to predict that small residual of the total amount of water in the atmosphere that's condensing.

152
00:15:54,380 --> 00:15:58,850
That's hard. Climate models are not made for it. They don't do it well.

153
00:15:58,850 --> 00:16:04,830
Another way of looking at it is that climate models try to simulate the motion in the atmosphere using Newton's laws of motion,

154
00:16:04,830 --> 00:16:09,530
the loss of some dynamics with grids that are typically 100 kilometres wide.

155
00:16:09,530 --> 00:16:14,060
Maybe you're going towards twenty five 10 kilometres and some simulations and these

156
00:16:14,060 --> 00:16:19,490
clouds say strata cumulus cumulus damped dynamical scales of 10 to 100 metres.

157
00:16:19,490 --> 00:16:23,190
They literally fall through the cracks of the climate. Models fall through the mesh.

158
00:16:23,190 --> 00:16:30,170
It's no way to resolve it. We need to represent these clouds empirically in some fashion, and that's hard.

159
00:16:30,170 --> 00:16:37,910
And in fact, that's not working very well. It's just one climate model that I think is representative of all climate models.

160
00:16:37,910 --> 00:16:45,630
What is shown here is the cloud cover relative to observations so widespread that means the model has fewer clouds and observe it's blue,

161
00:16:45,630 --> 00:16:49,070
which means it's more about 10 than what's observed.

162
00:16:49,070 --> 00:16:55,250
And you see these vast red stretches in the subtropics or regions with a lot of low clouds of cumulus clouds.

163
00:16:55,250 --> 00:16:59,630
This climate model, as well as every climate model we have vastly under,

164
00:16:59,630 --> 00:17:05,720
predicts the amount of clouds by almost a factor of two is this colour Skelly goes to minus 40 percent.

165
00:17:05,720 --> 00:17:12,170
It's saturated, so it's even even more than 40 percent under prediction.

166
00:17:12,170 --> 00:17:18,890
So that's the problem with climate models that these uncertainties going forward and that leads to questions

167
00:17:18,890 --> 00:17:24,800
how reliable they are if you go back for millions of years and try to explain climates of the past.

168
00:17:24,800 --> 00:17:31,550
So we cannot resolve clouds, we cannot simulate clouds explicitly in a climate model and will not be able to do it any time soon, not for decades.

169
00:17:31,550 --> 00:17:36,950
Bigger clouds we can resolve. But these small clouds, you would need a computer billions of times faster than the fastest.

170
00:17:36,950 --> 00:17:41,840
We have to actually resolve them. This is not going to happen, but clouds are still physical objects.

171
00:17:41,840 --> 00:17:49,580
We know the equations governing them. We can solve the equations quite accurately and it can solve them just not on the globe.

172
00:17:49,580 --> 00:17:59,540
It needs to be in a smaller box. So here's a simulation of the cloud we did to cumulus clouds and the Caribbean over some ocean surface,

173
00:17:59,540 --> 00:18:06,050
and it's a computer simulation that looks pretty lifelike. We can compare those simulations with field data.

174
00:18:06,050 --> 00:18:12,740
They reproduce well. We have great confidence that we can solve the equations in small boxes quite well.

175
00:18:12,740 --> 00:18:15,170
The boxes just aren't as big as a globe.

176
00:18:15,170 --> 00:18:22,700
In fact, we have so much confidence in the simulation of this cloud that was cast in stone, says Karen LeMond,

177
00:18:22,700 --> 00:18:26,090
a sculpture who approached me with a question I have about this and saying she

178
00:18:26,090 --> 00:18:29,780
wanted to make a sculpture that weighs as much as a cloud and marble and said,

179
00:18:29,780 --> 00:18:33,640
well. You need a very big block of marble and you're very busy.

180
00:18:33,640 --> 00:18:39,010
You'll be very busy for a long time, but I didn't know at the time is that sculptures of robots and they can deal with cat files.

181
00:18:39,010 --> 00:18:46,960
So there was a robot sculpting for for a good six weeks and Karen than finishing the work by hand.

182
00:18:46,960 --> 00:18:52,090
It became the sculpture called Cumulus that was exhibited at the piano in Venice in 2017.

183
00:18:52,090 --> 00:18:55,790
So this this is this weighs about as much as a real cumulus clouds.

184
00:18:55,790 --> 00:19:02,380
It's a few times before in there. We can also simulate stroke cumulus clouds quite well.

185
00:19:02,380 --> 00:19:04,450
This is a simulation of strutting cumulus clouds. With this,

186
00:19:04,450 --> 00:19:09,550
you're looking down on if you wish to try to calm the clouds or you can think of looking up into

187
00:19:09,550 --> 00:19:15,110
the blue sky and what you see and there's wide scale is how much water there is in the clouds.

188
00:19:15,110 --> 00:19:20,190
See, is this just measured as a thickness of a liquid layer? It's about these 100 microns or so the dimensions.

189
00:19:20,190 --> 00:19:26,110
So an individual cloud that's about how little water there is to try to cumulus clouds.

190
00:19:26,110 --> 00:19:28,360
I'm particularly interested in, well, either the most frequent,

191
00:19:28,360 --> 00:19:39,070
but b they have very special dynamics that I find intriguing and that led us to questions what happens to them in a global warming scenario?

192
00:19:39,070 --> 00:19:42,820
So A. They're really important.

193
00:19:42,820 --> 00:19:52,460
And for Earth's climate, they cool Earth by about eight degrees globally as the subtropical strata cumulus clouds simply by reflecting side.

194
00:19:52,460 --> 00:19:56,750
What makes them pretty interesting is that.

195
00:19:56,750 --> 00:20:03,830
Every cloud in some fashion is driven by turbulence to run emotions that lead to air masses rising and air condensing.

196
00:20:03,830 --> 00:20:07,610
And in most clouds, that turbulence is driven from below by heating.

197
00:20:07,610 --> 00:20:11,990
Send a tropical cumulus clouds you heat the surface air masses rise. Eventually, water vapour condenses.

198
00:20:11,990 --> 00:20:15,920
You form a cloud of Australia. Cumulus clouds of turbulence is driven from above.

199
00:20:15,920 --> 00:20:21,410
It's driven by a cooling instrument by radiative cooling. So these clouds does.

200
00:20:21,410 --> 00:20:27,560
The 100 micron thick layer of of liquid water is an incredibly good absorber

201
00:20:27,560 --> 00:20:31,490
of infra-red radiation that's basically opaque to infra-red radiation means,

202
00:20:31,490 --> 00:20:38,390
which means they emit thermal infra-red radiation upwards, and that's how they cool themselves.

203
00:20:38,390 --> 00:20:46,520
And the question I had a number of years ago is, well, if not, you put more greenhouse gases in the atmosphere.

204
00:20:46,520 --> 00:20:54,020
This cooling of the clouds should become weaker. Well, the cooling does is it leads to cooling of air mass at a sink to the surface.

205
00:20:54,020 --> 00:20:59,870
They pick up moisture from the ocean surface, bring the moisture back up, and that nourishes the clouds in a convective cycle.

206
00:20:59,870 --> 00:21:04,580
If this cooling gets weaker is convective. Cycles will get weaker and the clouds can get thinner.

207
00:21:04,580 --> 00:21:10,310
And the question was, well, now you put more greenhouse gases in the atmosphere that cooling off to get weaker and much the

208
00:21:10,310 --> 00:21:16,190
same way that the humid summer night is warmer than the clear summer night and humid summer night.

209
00:21:16,190 --> 00:21:23,720
The Earth's surface can't cool radiative lee as well as it can in a clear summer night and hence the human nights of warmer,

210
00:21:23,720 --> 00:21:29,430
which is why I prefer Los Angeles summers over in New York summers.

211
00:21:29,430 --> 00:21:34,680
Same for the clouds, so you put more greenhouse gases, carbon dioxide, water vapour, anything else in the atmosphere above it?

212
00:21:34,680 --> 00:21:39,090
They won't be able to cool themselves as well. And the question is what would happen?

213
00:21:39,090 --> 00:21:43,470
So we asked the question by building a computer model to answer it.

214
00:21:43,470 --> 00:21:47,200
And a global climate model can simulate the clouds seen.

215
00:21:47,200 --> 00:21:53,790
So what we have to do is build a computer model that simulates the clouds in a box like the box you see.

216
00:21:53,790 --> 00:21:56,310
But then we have to represent the rest of the climate system somehow.

217
00:21:56,310 --> 00:22:01,080
So in a usual climate model, what you do is you resolve all the large scales of the motion quite accurately.

218
00:22:01,080 --> 00:22:03,390
We have great confidence we are doing this well.

219
00:22:03,390 --> 00:22:10,110
And then you represent all the small scales semi empirically and we don't have confidence we are doing that well.

220
00:22:10,110 --> 00:22:11,640
So we did two years. Just do.

221
00:22:11,640 --> 00:22:21,030
The exact opposite is similar to exactly small scales accurately, but that represented all large scale semi empirically, meaning the large scales.

222
00:22:21,030 --> 00:22:25,080
There's a lot of uncertainty in what we do there, but the small scales, if you have a good simulation,

223
00:22:25,080 --> 00:22:30,120
so in a cartoon, what it looks like, what we did is something like that.

224
00:22:30,120 --> 00:22:40,740
You had a box sphere. You simulate part of the subtropics, a small patch of ocean sea off the coast of Angola, Namibia, Peru, California,

225
00:22:40,740 --> 00:22:49,050
and that's coupled to the rest of the climate system to some column representing tropical dynamics to energy moisture transports in the atmosphere.

226
00:22:49,050 --> 00:22:59,820
It's coupled to the surface temperature and energy budget for the ocean to radiative fluxes of solar radiation, along with variation and the like.

227
00:22:59,820 --> 00:23:09,060
And then we the simulations. We did a lot of simulations, and what comes out in the end is a simple looking plot that is maybe too simple looking.

228
00:23:09,060 --> 00:23:15,020
This is a few million hours of computing to get a few dots of.

229
00:23:15,020 --> 00:23:21,500
I don't want to convert it into dollars or pounds for this plus plot costs, but it's expensive.

230
00:23:21,500 --> 00:23:26,390
And what do you see as the cloud fraction in the simulation and in the subtropical patch?

231
00:23:26,390 --> 00:23:35,120
So we start out at 400 parts per million CO2, about present day climate and this red upward pointing error that you can call our present day climate.

232
00:23:35,120 --> 00:23:41,790
So this is a situation where we have sort of cumulus clouds covering the entire subtropical part,

233
00:23:41,790 --> 00:23:46,760
simulating 100 percent cloud cover, and then we started increasing CO2 concentrations.

234
00:23:46,760 --> 00:23:51,570
And that's. As far as cloud cover is concerned, boring for a long time,

235
00:23:51,570 --> 00:24:01,260
you go from 400 to 600 to 800 and and 2000 parts per million and cloud cover is at 100 percent.

236
00:24:01,260 --> 00:24:05,700
But then there's a point where cloud cover falls literally off a cliff.

237
00:24:05,700 --> 00:24:15,020
Here it goes from 100 percent to around 30 percent. As happens between twelve hundred and thirty million parts per million in the simulation.

238
00:24:15,020 --> 00:24:22,430
The precise CO2 concentration where that happens is not crucial because it depends a bit on what you assume for the rest of the climate system,

239
00:24:22,430 --> 00:24:30,140
but that it happens is crucial. So we start from 400 parts ppm having strong cumulus and then we go into something like 30 percent cloud cover.

240
00:24:30,140 --> 00:24:37,280
But we just have scattered cumulus like around here. Why? And then he can increase CO2 concentrations even more.

241
00:24:37,280 --> 00:24:41,990
You lose some more cloud fraction. Otherwise, nothing too exciting happens.

242
00:24:41,990 --> 00:24:48,230
But here's what's interesting now let's go backwards. So we figured out how to scrub the atmospheric CO2 and our computer model.

243
00:24:48,230 --> 00:24:57,210
That's just literally a novel term. So you remove CO2 from the atmosphere and then you're walking back along the blue areas here.

244
00:24:57,210 --> 00:25:03,860
And the interesting part is that we are not reforming shredded cumulus clouds where they first broke up.

245
00:25:03,860 --> 00:25:12,550
But here you have to go to two below 300 parts per million CO2 before you get back to 100 percent cloud cover and start to cumulus.

246
00:25:12,550 --> 00:25:18,800
So we say there's hysteresis, meaning that this the state of the system depends on its history.

247
00:25:18,800 --> 00:25:24,170
And I'll explain it in just a moment how it comes about, but that's just cloud cover.

248
00:25:24,170 --> 00:25:28,190
So I have these clouds so enormously important for climate because it reflects so much sunlight.

249
00:25:28,190 --> 00:25:33,410
So you, you perhaps are asking and should be asking what happens to temperature.

250
00:25:33,410 --> 00:25:38,210
And he said, what happens to temperature in these simulations? So this is the temperature in the tropics,

251
00:25:38,210 --> 00:25:45,330
which is the temperature change in the tropics as a reasonably good approximation of global mean temperature changes here.

252
00:25:45,330 --> 00:25:51,470
So again, we started out at 400, we're going up and 2100 parts per million or so.

253
00:25:51,470 --> 00:25:56,210
The cloud cover drops and temperature goes up in this case by about eight degrees centigrade.

254
00:25:56,210 --> 00:26:02,000
So this is roughly the cooling subtropical marines try to cumulus clouds currently provide.

255
00:26:02,000 --> 00:26:09,140
So you lose that cooling, you get to about eight degrees warming in addition to the global warming that was gradually happening before.

256
00:26:09,140 --> 00:26:15,680
So it's to go from, say, from four to eight and reports familiar and you see global warming of this model of three and a half degrees or so.

257
00:26:15,680 --> 00:26:19,430
So the fairly canonical climate sensitivity that you see in climate models.

258
00:26:19,430 --> 00:26:23,600
But then when these clouds break up and that happens quite abruptly as a function of CO2,

259
00:26:23,600 --> 00:26:30,380
here you see an additional jump in temperatures and then as you go down, remove CO2,

260
00:26:30,380 --> 00:26:35,030
you stay on the warm branch of the climate system and in this model system and temperatures

261
00:26:35,030 --> 00:26:41,200
stay seven degrees or so above the temperatures and this cooler range until the clouds reform.

262
00:26:41,200 --> 00:26:47,690
And in this case, 200 parts per million. So the lower concentrations then we currently have.

263
00:26:47,690 --> 00:26:52,840
That's just a simulation, and you could ask all sorts of questions.

264
00:26:52,840 --> 00:27:02,290
How relevant an idealised model is, but one one reason we have confidence in idealised models is when we understand physically what's happening,

265
00:27:02,290 --> 00:27:07,330
and the physics is pretty straightforward. So the way you lose the clouds,

266
00:27:07,330 --> 00:27:17,170
the principal factor responsible here is that greenhouse gas concentrations increase that long with cooling of the cloud tops gets weaker.

267
00:27:17,170 --> 00:27:23,530
That drives turbulence down, but the driving of the turbulence down gets weaker.

268
00:27:23,530 --> 00:27:27,520
Up to a point where the clouds, we say decouple from this surface moisture supply,

269
00:27:27,520 --> 00:27:30,760
basically even when the turbulent air masses are driven down by the hook,

270
00:27:30,760 --> 00:27:38,780
by the cooling are not reaching the surface anymore, then cloud cover bends as cloud cover thins.

271
00:27:38,780 --> 00:27:42,080
The surface warms because less sunlight is reflected,

272
00:27:42,080 --> 00:27:46,670
less sunlight being reflected means there's more water vapour, any atmosphere as the surface warms.

273
00:27:46,670 --> 00:27:50,880
That in itself is a greenhouse effect gas amplifying the effect of CO2.

274
00:27:50,880 --> 00:27:55,370
You have a feedback cycle that leads to the break up of the clouds here.

275
00:27:55,370 --> 00:27:59,180
There are other processes that are important, and one is counterintuitive.

276
00:27:59,180 --> 00:28:03,020
Evaporation is important, and you might think that there's more evaporation.

277
00:28:03,020 --> 00:28:09,230
You would get more clouds. But that's not correct for these clouds with more evaporation, you actually get fewer of them.

278
00:28:09,230 --> 00:28:15,170
And that has to do with whatever variation does in the clouds. Water vapour condenses in the clouds.

279
00:28:15,170 --> 00:28:20,990
If you have more of it, that can convince you get more latent heat release that enhances the turbulence and the clouds,

280
00:28:20,990 --> 00:28:25,480
and that leads to more mixing of dry air into the clouds, and it works against their existence.

281
00:28:25,480 --> 00:28:33,230
That's another process that plays a role here, and it's somewhat counterintuitive. But now you can see how has two reasons why stability comes about.

282
00:28:33,230 --> 00:28:38,840
Once you lose the clouds, it's warm. When it's when it's warm, there's more water vapour, for example in the atmosphere.

283
00:28:38,840 --> 00:28:45,500
It's an additional greenhouse gas. Now, when you just ramp down CO2, well, there's still more water vapour in the atmosphere.

284
00:28:45,500 --> 00:28:52,640
That's still a greenhouse gas that still prevents the long way of cooling and still prevents the clouds from reforming.

285
00:28:52,640 --> 00:28:58,490
Similar story Applying to evaporation evaporation jumps by 70 percent when these clouds break up.

286
00:28:58,490 --> 00:29:04,970
It also works against the existence of the clouds, and that's how you get the sense to resist look out.

287
00:29:04,970 --> 00:29:10,640
So I said everything about large scale and the rest of the climate system here is idealised and to be taken with a grain of salt,

288
00:29:10,640 --> 00:29:15,350
which means that things like the precise CO2 concentration where that happens,

289
00:29:15,350 --> 00:29:19,730
that's not necessarily quantitatively accurate account in that quantitatively.

290
00:29:19,730 --> 00:29:22,490
Here's an example of how we can.

291
00:29:22,490 --> 00:29:27,860
You can make different assumptions, different reasonable assumptions what the large-scale atmospheric circulation does,

292
00:29:27,860 --> 00:29:33,560
whether it weakens weakly as the climate warms or weakens more strongly as the climate warms.

293
00:29:33,560 --> 00:29:37,280
These are the results again for the tropical temperature as a function of CO2.

294
00:29:37,280 --> 00:29:40,760
And you see that the point where these clouds break up shifts quite a bit.

295
00:29:40,760 --> 00:29:49,220
It's a function of CO2. The points where they reform also shifts the width of this as Teresa's loop shifts and the like, which is all to say.

296
00:29:49,220 --> 00:29:57,530
We can say with precision whether these clouds would break up a twelve hundred ppm CO2 or 600 or whatever it exactly is.

297
00:29:57,530 --> 00:30:06,200
We don't know that, but that it could happen seems pretty clear. Simulations show it, and the physics of it is clear.

298
00:30:06,200 --> 00:30:11,420
So back to back to the Eocene and the past was the mean.

299
00:30:11,420 --> 00:30:18,770
So one possible explanation of how you could have gotten climates as warm as the Eocene was,

300
00:30:18,770 --> 00:30:26,360
as we know it was because there are crocodiles in the Arctic. Is that a cloud cover might have been a lot thinner?

301
00:30:26,360 --> 00:30:32,000
Maybe there were no strutting cumulus clouds at the time, subtropical strutting cumulus clouds at the time at all.

302
00:30:32,000 --> 00:30:39,480
And that would certainly suffice to create a very warm climate globally and make a lush Ellesmere Island possible.

303
00:30:39,480 --> 00:30:41,570
It's one possible explanation.

304
00:30:41,570 --> 00:30:48,320
There are other possible explanations of generally climate as much more sensitive than models suggest without dramatic changes in cloud cover.

305
00:30:48,320 --> 00:30:56,270
That's also a possibility. But at least that's one possible explanation for a climate state that we know occurred.

306
00:30:56,270 --> 00:31:01,220
That has confounded explanation for a while. And what does it mean for the future?

307
00:31:01,220 --> 00:31:09,530
So here is CO2 and a bit more. More recently, it's just the last thousand years, rather than millions of years over those timescales.

308
00:31:09,530 --> 00:31:10,820
We know CO2 precisely.

309
00:31:10,820 --> 00:31:19,460
We know it by just taking air bubbles frozen into ice from Antarctica and looking at the CO2 concentration of the air that's preserved there.

310
00:31:19,460 --> 00:31:27,560
And we know that CO2 was fairly constant at two hundred eighty or so parts per million for thousands of years into the past.

311
00:31:27,560 --> 00:31:31,940
And of course, it has been speaking up for us at the beginning of the industrial revolution.

312
00:31:31,940 --> 00:31:36,320
And the red curve are actually instrumental measurements also in Antarctica.

313
00:31:36,320 --> 00:31:44,000
So they consist of a person leaving a station and taking a flask of air and then going to allow but measuring how much CO2 is in there.

314
00:31:44,000 --> 00:31:51,590
And you see, now we are at around 410 parts per million CO2.

315
00:31:51,590 --> 00:31:57,600
How significant the change is that I think it's useful to put this again into a perspective of the past so years,

316
00:31:57,600 --> 00:32:08,420
is Zoom in now the last 50 million years again and and then projecting out in the future what might be happening so far in 10 parts per million?

317
00:32:08,420 --> 00:32:12,860
That's that's a CO2 concentration we haven't seen in at least two million years on Earth.

318
00:32:12,860 --> 00:32:18,570
So this is getting us back to the Pliocene.

319
00:32:18,570 --> 00:32:27,720
Somewhere between two or three or four million years ago and a Pliocene Earth was around three degrees warmer and three degrees.

320
00:32:27,720 --> 00:32:32,130
I don't know if that sounds significant to you or not the other two, but there is agreement and to ease and come back to it.

321
00:32:32,130 --> 00:32:43,290
Three degrees may not sound like much, but three degrees of warming sustained over thousands of years is enough to melt a lot of ice.

322
00:32:43,290 --> 00:32:54,250
So during the Pliocene bit about three degrees higher temperatures, sea level was a good and good 20 metre or so higher than it is now.

323
00:32:54,250 --> 00:32:55,590
Now that is significant.

324
00:32:55,590 --> 00:33:05,190
Life then, was perfectly possible for all sorts of animals, and there were no Homo sapiens around at a time, but they could have lived just fine.

325
00:33:05,190 --> 00:33:07,290
Just the challenges,

326
00:33:07,290 --> 00:33:16,150
the challenges that we have built cities and say this is what the artist called London be under with this case will be 40 metres sea level rise.

327
00:33:16,150 --> 00:33:24,410
That's melting two thirds of all ice sheets. London would be flooded as much as much of L.A. and New York and the like.

328
00:33:24,410 --> 00:33:35,060
So the problem here is that if you sustain a warming of that sort for a long time, for thousands of years, you're just not adapted to that situation.

329
00:33:35,060 --> 00:33:42,420
We have built cities and places where they perhaps wouldn't be in that case. Now, I don't.

330
00:33:42,420 --> 00:33:47,580
I hope you won't get to anything resembling the scenario.

331
00:33:47,580 --> 00:33:54,900
But here's where we are. So we had about one degree global warming over the last 100 150 years.

332
00:33:54,900 --> 00:34:01,710
That's not uniformly distributed as this in temperature change between 2000, around now in the 1850s.

333
00:34:01,710 --> 00:34:06,220
And the grey areas are just areas where we don't have enough data to say much. This is in the mean.

334
00:34:06,220 --> 00:34:09,420
It's about the degree there's more warming over high latitudes,

335
00:34:09,420 --> 00:34:16,410
more over continents for reasons we understand having to do with how much water is available to evaporate.

336
00:34:16,410 --> 00:34:19,080
Number to remember is a degree of warming we have had,

337
00:34:19,080 --> 00:34:26,280
and I've probably heard about the Paris Agreement that had two hundred and ninety six signatories a while back,

338
00:34:26,280 --> 00:34:28,740
and now it's down to one hundred ninety five.

339
00:34:28,740 --> 00:34:35,730
And the Paris Agreement stipulated that we should limit warming to two degrees above pre-industrial levels.

340
00:34:35,730 --> 00:34:43,830
So we had about a degree, so there's one more degree to go. Now you can take climate models we have and ask how much more CO2 can we put in

341
00:34:43,830 --> 00:34:48,090
the atmosphere before you get that extra degree before we reach this Paris target,

342
00:34:48,090 --> 00:34:52,890
the threshold is fairly arbitrary, but it's useful to pick a threshold to illustrate the point I want to make here.

343
00:34:52,890 --> 00:34:55,350
So here's the answer you get from climate models.

344
00:34:55,350 --> 00:35:00,810
How much CO2 can you put into the atmosphere before we reach two degrees warming above pre-industrial levels?

345
00:35:00,810 --> 00:35:08,310
And it's the same twenty nine models I've shown you before. The answer varies between close to 600 parts per million and 480 parts per million.

346
00:35:08,310 --> 00:35:11,700
The models here are ordered in order of increasing climate sensitivity,

347
00:35:11,700 --> 00:35:16,560
so the models that are right are more sensitive just to say that the models on the radar right,

348
00:35:16,560 --> 00:35:23,130
the more sensitive models are right, then we'll reach those two degree warming within 20 years or so.

349
00:35:23,130 --> 00:35:30,050
And that's pretty much irrespective of policy interventions and the like. That's energy infrastructure, for example,

350
00:35:30,050 --> 00:35:34,630
is so long live that this would be an unavoidable with the models on the radar right for the models on the left.

351
00:35:34,630 --> 00:35:43,710
All right. Well, we would have the twenty six days or so even under high emission scenarios before these two degrees of warming are reached.

352
00:35:43,710 --> 00:35:54,710
So just a full human generation difference between the members of the predictions here, and I find that wide range of predictions unsatisfying.

353
00:35:54,710 --> 00:36:01,820
And it's perhaps even worse than what this graph says, because climate models now coming out are even a lot of them right,

354
00:36:01,820 --> 00:36:08,120
even to the right, the most sensitive models we have here. So maybe that's not even spanning the range of what could be happening.

355
00:36:08,120 --> 00:36:12,380
The point is the uncertainties are large and there are there are enormously large.

356
00:36:12,380 --> 00:36:19,040
If you want to plan rationally for human generation difference when you hit a certain temperature target,

357
00:36:19,040 --> 00:36:25,220
temperature affects everything in the climate system, precipitation and the like, so all the other predictions will be equally uncertain.

358
00:36:25,220 --> 00:36:32,150
You want to be able to plan infrastructure in one storm water management infrastructure and want to do so rationally.

359
00:36:32,150 --> 00:36:37,640
A number of cities are building seawalls in New York, for example, and spending close to a billion dollars on that.

360
00:36:37,640 --> 00:36:45,950
It would be nice to know how to build the seawall that New York City is protected from, say, a 100 year flood 50 years from now.

361
00:36:45,950 --> 00:36:52,820
And we don't have the information to say accurately what would what that would take Sam for wildfire risk.

362
00:36:52,820 --> 00:36:58,480
Assessing risks for vulnerable communities, droughts and the like.

363
00:36:58,480 --> 00:37:09,340
So the cost of climate change adaptation, which will be unavoidable, is estimated to be above $200 billion annually by 2050.

364
00:37:09,340 --> 00:37:15,340
People have tried to calculate what the socio economic value would be of having the uncertainties

365
00:37:15,340 --> 00:37:20,740
in climate sensitivity within 10 years and the number of they come up with this from a paper.

366
00:37:20,740 --> 00:37:24,550
Chris Hope was 10 trillion dollars.

367
00:37:24,550 --> 00:37:31,030
I don't have a good intuition for $10 trillion and maybe it's only one trillion dollars, but it's a big number is a point.

368
00:37:31,030 --> 00:37:35,660
And and the other point is.

369
00:37:35,660 --> 00:37:41,450
Note note how the second point is phrased, having the uncertainties within 10 years,

370
00:37:41,450 --> 00:37:46,280
it's important to get better predictions fast because climate is changing and at some

371
00:37:46,280 --> 00:37:51,350
point the predictions won't be that useful anymore because we see what what is happening.

372
00:37:51,350 --> 00:37:57,910
So the name of the game here is to come up with better predictions and come up with them fast.

373
00:37:57,910 --> 00:38:03,880
You don't want to wait 10 years, and there are things we can do assign side.

374
00:38:03,880 --> 00:38:11,410
And here's a shot you simulations of clouds and small boxes so we can do them in one small box that can do them in two small boxes.

375
00:38:11,410 --> 00:38:16,840
Well, we can do them in tens of thousands of small boxes. You can embed in the climate model.

376
00:38:16,840 --> 00:38:21,830
We can do it everywhere. That's where you need factors of billions and extra extra in computing that we don't have,

377
00:38:21,830 --> 00:38:25,510
but we can do it in thousands and maybe 10 thousands of places.

378
00:38:25,510 --> 00:38:30,760
And these simulations are quite good, at least as far as emotion of the clouds is concerned.

379
00:38:30,760 --> 00:38:37,420
So that's something we should be able to use. And the other thing we have, of course, is a wealth of data.

380
00:38:37,420 --> 00:38:39,910
This is what's called the train of satellites.

381
00:38:39,910 --> 00:38:48,100
It's it's satellites flying and constellation crossing the same spot on Earth's surface within seconds of each other.

382
00:38:48,100 --> 00:38:52,210
It's called the A-Train because they cross it in the afternoon after afternoon.

383
00:38:52,210 --> 00:38:55,210
The measure all sorts of things in a coordinated fashion.

384
00:38:55,210 --> 00:39:01,690
For example, in front is OCO two is measuring carbon dioxide for the actress calypso and clouds that they measure.

385
00:39:01,690 --> 00:39:14,480
A cloud properties, the lighter radar, a number of other Aquos measuring water vapour, for example, if you have a wealth of data and.

386
00:39:14,480 --> 00:39:19,130
These data, they can inform climate models, we have used them to evaluate climate models,

387
00:39:19,130 --> 00:39:24,780
you have used them in a way I've shown you before to say that there's a large bias in cloud cover, but you can do more.

388
00:39:24,780 --> 00:39:31,820
You can directly learn from the data. They have seen so many advances in the data sciences in the last few decades.

389
00:39:31,820 --> 00:39:38,690
And what we wanted to do is bring these advances in a data scientist to bear on the climate problem by building a model that

390
00:39:38,690 --> 00:39:46,280
learns from these high resolution simulations we can do in many spots and learns from from the data we have could be space based.

391
00:39:46,280 --> 00:39:51,680
But the same is true, say, for autonomous vehicles in the ocean and the like.

392
00:39:51,680 --> 00:39:59,360
So we formed versus a number of people getting together over time, and companies come to call ourselves the Climate Modelling Alliance.

393
00:39:59,360 --> 00:40:00,470
And some are people at Caltech,

394
00:40:00,470 --> 00:40:09,770
my institution and the Jet Propulsion Laboratory at MIT Naval Postgraduate School and an increasingly increasing network of other collaborators.

395
00:40:09,770 --> 00:40:16,980
The idea of this group is let's just try to. Bring climate model.

396
00:40:16,980 --> 00:40:22,080
Put it on a data driven footing, and weather forecasting has happened decades ago.

397
00:40:22,080 --> 00:40:28,740
Weather forecasts have become very good and perhaps better than their reputation, in large part thanks to extensive use of data.

398
00:40:28,740 --> 00:40:32,820
What we want to do for climate modelling is, in some ways the same that has happened for weather forecasting,

399
00:40:32,820 --> 00:40:41,160
make much more extensive use of data to make the models better. So we are doing is built a model that is in some ways a traditional model.

400
00:40:41,160 --> 00:40:47,700
If you visit it as an atmospheric model and ocean model and land model, size models and the like, but wrapped around everything,

401
00:40:47,700 --> 00:40:57,330
so is a layer of machine learning data simulation tools that allow the model at the core to learn from observational data

402
00:40:57,330 --> 00:41:04,740
and from high resolution simulations of clouds of sea ice of ocean turbulence that you can spin out where you need them.

403
00:41:04,740 --> 00:41:11,680
When you need them. You can spin them out. Part of what's attractive, it lends itself very well to cloud computing.

404
00:41:11,680 --> 00:41:21,240
You can do this in a distributed computing framework. And of course, it's irresistible to simulate clouds on the cloud, which is what we're doing,

405
00:41:21,240 --> 00:41:25,020
what we want to do if we want to achieve predictive advances that are practical.

406
00:41:25,020 --> 00:41:28,920
In the end, we want to get better predictions of of rainfall, for example,

407
00:41:28,920 --> 00:41:32,460
rainfall extremes and things that really matter for infrastructure planning,

408
00:41:32,460 --> 00:41:39,180
for example, high impact weather, accurate risk assessment, heat waves, droughts, storms and the like.

409
00:41:39,180 --> 00:41:44,550
High latitudes are important and like to be able to say what happens to Arctic sea ice.

410
00:41:44,550 --> 00:41:47,100
You'd like to be able to assess risks for vulnerable communities.

411
00:41:47,100 --> 00:41:55,230
And what I'm hoping for is that a few years from now that most of you will have apps for climate just as much

412
00:41:55,230 --> 00:42:01,140
as you have them for a weather where you can get accurate climate information for whatever interests you say.

413
00:42:01,140 --> 00:42:04,770
If you want to purchase a house and a coastal area,

414
00:42:04,770 --> 00:42:12,820
the likelihood that it will be impacted by the storm surge or in California if you purchase a house near near brush land,

415
00:42:12,820 --> 00:42:17,530
what the likelihood is that there will be wildfires and the like.

416
00:42:17,530 --> 00:42:23,860
I'd like to have people to have apps that provide that information, much like you get weather information now, but now.

417
00:42:23,860 --> 00:42:29,140
Things a few decades ahead that are just as actionable for your decision to now say,

418
00:42:29,140 --> 00:42:34,180
if you plan infrastructure or purchase property or whatever it is as weather information is for you,

419
00:42:34,180 --> 00:42:36,610
currently, you decide whether to bring an umbrella or not.

420
00:42:36,610 --> 00:42:43,750
But for these billions of dollars, a decision that we have to make, it seems even more important to have accurate information.

421
00:42:43,750 --> 00:42:49,630
And hopefully you get there within within four or five years is our goal.

422
00:42:49,630 --> 00:42:54,790
So the rev up, let me say a little bit of the people who were involved here Colleen, Colleen,

423
00:42:54,790 --> 00:42:58,640
press or two research scientist or now with the Pacific Northwest National Laboratory,

424
00:42:58,640 --> 00:43:03,790
as they did these simulations of the straight cumulus clouds I showed you. And the Climate Modelling Alliance.

425
00:43:03,790 --> 00:43:09,850
That's most of us down there. Not quite all of us at the meeting a few months ago.

426
00:43:09,850 --> 00:43:14,020
This is our office space. It's the garden of the Joint Office space.

427
00:43:14,020 --> 00:43:19,270
We have it pretty luxuriously nice and part of what makes us exciting for us.

428
00:43:19,270 --> 00:43:24,340
Aside from the direct usefulness of it, is that the group of people you see,

429
00:43:24,340 --> 00:43:30,040
other oceanographers or atmospheric scientists are software engineers who applied mathematicians and computer scientists,

430
00:43:30,040 --> 00:43:34,780
Caltech, who are all in one building occupying joint office space.

431
00:43:34,780 --> 00:43:40,900
And the hope is, and we have seen that hope being materially assisted by fostering interactions

432
00:43:40,900 --> 00:43:48,430
between the disciplines in a much tighter way than what has happened before. But we can make progress.

433
00:43:48,430 --> 00:43:53,170
I should acknowledge people who funders Eric Schmidt, provide the bulk of our funding,

434
00:43:53,170 --> 00:44:01,240
the recommendation on the French futures programme and to a number of other philanthropic organisations and the public public foundations involved.

435
00:44:01,240 --> 00:44:07,570
Earthrise Alliance, the Family Foundation and National U.S. National Science Foundation is funding a

436
00:44:07,570 --> 00:44:12,850
software institute for us and Charles Trimble and Maxine Linder of trustees of Caltech,

437
00:44:12,850 --> 00:44:18,820
who provided seed funding for us and got us started. And especially Charles Trimble early on was a great inspiration.

438
00:44:18,820 --> 00:44:20,110
You may not have heard his name,

439
00:44:20,110 --> 00:44:27,040
but he changed your life in case someone who figured out how to miniaturise jobs and that he is the reason that you don't get lost anymore.

440
00:44:27,040 --> 00:44:32,350
And so now we want to make sure that in the future we won't get lost as far as climate change is concerned.

441
00:44:32,350 --> 00:44:47,650
So I'll stop right here. I'm happy to answer questions. OK, thanks to here.

442
00:44:47,650 --> 00:44:54,250
So the plan is to ask hopefully some questions or comments.

443
00:44:54,250 --> 00:44:57,550
I guess the plan is to use the microphone, is that right?

444
00:44:57,550 --> 00:45:04,830
So maybe what I'll do since the microphone is here is sort of start in this area of the room.

445
00:45:04,830 --> 00:45:14,590
And yes, I worked my way. Though it is quite interesting, right at the beginning of your talk,

446
00:45:14,590 --> 00:45:20,740
when you said of all the water vapour that's in the atmosphere, only a very small fraction of its clouds.

447
00:45:20,740 --> 00:45:32,010
And it made me think, what is the contribution of vapour trails from high, high flying aircraft to global cooling or global warming?

448
00:45:32,010 --> 00:45:38,160
So clubs have two important effects. One is reflecting sunlight, and the other is that they have a greenhouse effect.

449
00:45:38,160 --> 00:45:43,830
And these low clouds they've talked about at the beginning, the greenhouse effect is quite small because they're so low.

450
00:45:43,830 --> 00:45:49,560
So they mostly reflect sunlight on high clouds, contrail cirrus clouds.

451
00:45:49,560 --> 00:45:55,800
Their main effect, as is the greenhouse effect, and they contribute to warming.

452
00:45:55,800 --> 00:46:10,500
I mean, it is a measurable effect of what comes from aircraft, but it's a small effect.

453
00:46:10,500 --> 00:46:17,490
That's for a very interesting talk with the curve that you have on the top right, so the looking back at that patio,

454
00:46:17,490 --> 00:46:25,330
Seville Maxima and one of the things about the oceans then is that you had quite a deep thermocline and as that as a cool,

455
00:46:25,330 --> 00:46:29,760
the thermocline came up and you could shoulder thermocline and bring cold water to the surface.

456
00:46:29,760 --> 00:46:38,280
So in those experiments that you've done, I wondered how sensitive others shot shoddy chemist regimes to the ocean surface temperature.

457
00:46:38,280 --> 00:46:40,830
So we have motion dynamics and those experiments.

458
00:46:40,830 --> 00:46:49,440
I did suggest really it's an ocean layer as an energy balance, and the surface temperature is is not independently determined.

459
00:46:49,440 --> 00:46:54,890
It's determined through an energy balance. So you can't really ask the question how how sensitive are they to surface temperature alone?

460
00:46:54,890 --> 00:47:02,580
But what you can ask is say what surface temperature increasing surface temperature does as, for example, lead to more water vapour in the atmosphere?

461
00:47:02,580 --> 00:47:08,670
And you can ask how large is the effect of the added water vapour on the long wave cooling of the clouds?

462
00:47:08,670 --> 00:47:15,610
In addition to the CO2, and that's about half. Okay, so over here.

463
00:47:15,610 --> 00:47:18,320
Okay, thank you to you for great talk.

464
00:47:18,320 --> 00:47:24,710
So one of the big differences between weather forecasting and climate prediction is if you have a lousy numerical weather prediction model,

465
00:47:24,710 --> 00:47:31,010
you find out pretty quickly climate forecasting. We may not find out for several decades.

466
00:47:31,010 --> 00:47:33,830
So could you say a little bit about the challenges of, you know,

467
00:47:33,830 --> 00:47:39,640
what data is available and to what extent can you really constrain the climate forecasts with data?

468
00:47:39,640 --> 00:47:44,630
Yeah, I mean, that's an obvious challenge. A couple of answers.

469
00:47:44,630 --> 00:47:51,620
So a in many aspects, climate models are doing so poorly right now that simply having a better present,

470
00:47:51,620 --> 00:47:56,480
better representation of the present climate will be progress. A seasonal cycle.

471
00:47:56,480 --> 00:48:00,080
Cloud cover. Seasonal cycle in the land biosphere and the like.

472
00:48:00,080 --> 00:48:09,020
So there there is just a lot of areas. For example, photosynthesis in a biosphere models differ by an order of magnitude and that seasonal cycle.

473
00:48:09,020 --> 00:48:16,490
So just simply getting a better present climate with a model that didn't talk about is so process oriented.

474
00:48:16,490 --> 00:48:21,590
So it's not going to be model where fit millions of parameters will be sparse and parameter space.

475
00:48:21,590 --> 00:48:29,370
So getting getting a better simulation of a better climate with a sparsely parameter model, that challenge and if you.

476
00:48:29,370 --> 00:48:32,760
Some confidence that you have gotten a better model now you want to verify that.

477
00:48:32,760 --> 00:48:36,300
And that's where things get harder and you don't want to wait decades.

478
00:48:36,300 --> 00:48:44,040
I think what you have to do is find Short-Term prediction targets, say, predict the response and try to do that accurately.

479
00:48:44,040 --> 00:48:50,370
It would actually be interesting to use a climate model in a weather forecasting setting and see if you can see progress there as well.

480
00:48:50,370 --> 00:48:57,300
Even though that may not be the success metric that's most important for climate, but you have to find some,

481
00:48:57,300 --> 00:49:06,630
some targets that can build confidence over time, and the targets have to be short term because we don't want to wait 20 years.

482
00:49:06,630 --> 00:49:19,400
But it's something you probably right, that's a problem, and we have the climate modelling ready. Thanks.

483
00:49:19,400 --> 00:49:28,310
That was really interesting. If with increasing temperatures, you continue to get more water vapour in the atmosphere.

484
00:49:28,310 --> 00:49:35,660
What happens to types of clouds that aren't destructive cumulus? Do we see a compensation with more cumulonimbus clouds or something else?

485
00:49:35,660 --> 00:49:41,810
Where's that water vapour going? Mostly in vapour and not so much rain clouds, right?

486
00:49:41,810 --> 00:49:45,500
Because clouds are just this tiny, tiny residual of water that that condenses there.

487
00:49:45,500 --> 00:49:56,520
So. You you cannot make any direct connexion between the amount of vapour and the atmosphere and and how many clouds you have.

488
00:49:56,520 --> 00:50:03,300
If anything, what you would need to do is think about how far from saturation you are.

489
00:50:03,300 --> 00:50:09,090
How much do you need to move an air parcel to reach saturation, and the interesting part is as you warm the climate,

490
00:50:09,090 --> 00:50:13,830
you have more water vapour in the atmosphere in such a way that the relative humidity doesn't change very much.

491
00:50:13,830 --> 00:50:21,150
But what that implies is that the what you might want to call the distance to saturation actually increases some.

492
00:50:21,150 --> 00:50:27,300
But that metric alone, it would be harder to form clouds. But I'm not saying this is actually how it works.

493
00:50:27,300 --> 00:50:32,400
Just to say that the intuition more water vapour translates into more clouds can easily lead to Australia.

494
00:50:32,400 --> 00:50:40,860
It's more complicated than that. For.

495
00:50:40,860 --> 00:50:47,790
Thank you. You've pointed out that we trust these high resolution models, and therefore they're a good tool to build a climate model.

496
00:50:47,790 --> 00:50:55,290
I mean, you've heard earlier this week already, I'm much more cautious about models of deep conviction, also in high resolution settings.

497
00:50:55,290 --> 00:51:03,930
And we have very limited observational constraints. So I wondered how long do we have to wait until we have some constraints on these models

498
00:51:03,930 --> 00:51:07,810
that we trust them more and how this industry of time scales that we're worried about now?

499
00:51:07,810 --> 00:51:11,850
I mean, a couple of aspects, right? So there's you're talking about the micro physical issues that we discussed.

500
00:51:11,850 --> 00:51:18,400
So yes, that's what Philip is referring to, is that even human, you can resolve the dynamics of motion in the clouds.

501
00:51:18,400 --> 00:51:25,080
There's still the physics of how droplets and ice crystals form or micrometre scale and those who can still resolve.

502
00:51:25,080 --> 00:51:30,480
So that that remains is a challenge. So I think there are two pieces that help, or maybe three.

503
00:51:30,480 --> 00:51:38,210
I mean, the first one is. Even if he couldn't do anything about the micro physics, I think they can, but even if he couldn't.

504
00:51:38,210 --> 00:51:44,060
Wouldn't it be nice if all uncertainties are reduced to that and the dynamics as we have confidence and right?

505
00:51:44,060 --> 00:51:48,140
So my first answer would be, well, let's just deal with the pieces we can deal with.

506
00:51:48,140 --> 00:51:49,760
That might be something left.

507
00:51:49,760 --> 00:51:56,910
And I think, however, what we can do in the micro physical side or how we can reduce high resolution simulations in that context.

508
00:51:56,910 --> 00:52:00,920
So what do you need to make sure that your global model,

509
00:52:00,920 --> 00:52:04,550
your high resolutions are consistent with one another so that whatever is wrong with the

510
00:52:04,550 --> 00:52:09,890
micro physics is consistently wrong so that you can still infer dynamical properties?

511
00:52:09,890 --> 00:52:15,760
But then we do have data and we do know a lot about the how.

512
00:52:15,760 --> 00:52:20,350
The deep convective clouds and main issue is the ice face a mixed phase, right?

513
00:52:20,350 --> 00:52:24,440
We have a lot of data about.

514
00:52:24,440 --> 00:52:31,140
The properties of clouds higher up and to what degree they're liquid or ice, and we do know that models are doing this poorly.

515
00:52:31,140 --> 00:52:38,010
And I think those data alone paired with whatever you can get measuring effect of radius of droplet droplets and the like.

516
00:52:38,010 --> 00:52:41,700
I think they do provide constraints we haven't even begun to exploit.

517
00:52:41,700 --> 00:52:50,510
Now, whether they do, I don't think they will do some scientists zero, but I think they can reduce uncertainties in that area as well.

518
00:52:50,510 --> 00:53:02,800
Josh, did you? As you said, you can't run the pretty high resolution simulation or the grip boxes,

519
00:53:02,800 --> 00:53:08,920
but have you got a sort of sense as if you did couple it to a full model,

520
00:53:08,920 --> 00:53:17,740
whether that might kill your hysteresis loop, whether it might make it wider if you got any sort of sense of what you'd expect to see?

521
00:53:17,740 --> 00:53:24,280
Ask me again, and I mean, six months, OK, I am trying to do it.

522
00:53:24,280 --> 00:53:29,060
They're challenging simulations. It's computationally expensive to do, but yeah, it's the obvious thing to do, right?

523
00:53:29,060 --> 00:53:34,890
And it's a good point. OK.

524
00:53:34,890 --> 00:53:44,740
Anymore for any.

525
00:53:44,740 --> 00:53:53,500
So you mentioned some of the funding sources and also the all the costs that could be avoided if we actually have a better prediction of climate.

526
00:53:53,500 --> 00:54:02,380
And I was just wondering why not? Well, a signal of the big tech giants that we have in the world like Google, Facebook and so on.

527
00:54:02,380 --> 00:54:09,220
Why are they not on board with these kind of projects? Because I mean, yes, making money and avoiding costs.

528
00:54:09,220 --> 00:54:19,800
Still something different. But in the end, it's in the information that they could really well well use for any kind of future investments, right?

529
00:54:19,800 --> 00:54:25,230
Short answer, as people are on board. I mean, if.

530
00:54:25,230 --> 00:54:28,800
It's a huge business opportunity, it's a reality, right?

531
00:54:28,800 --> 00:54:33,930
And there is a lot of interest, it's just that this problem is not going to be solved by having 100 software engineers.

532
00:54:33,930 --> 00:54:40,230
That's a challenge. OK.

533
00:54:40,230 --> 00:54:49,250
Just maybe take two, but we leaving the front first.

534
00:54:49,250 --> 00:55:01,440
I wonder whether there is anything that we can do like the in this room after today to help the climate or decrease global warming and vote.

535
00:55:01,440 --> 00:55:10,290
And the next best thing is come up with a really clever way of producing electricity from sunlight and say half the cost of what it currently is.

536
00:55:10,290 --> 00:55:14,680
Then you will also be rich. OK.

537
00:55:14,680 --> 00:55:25,860
Just one last one was. Being greedy on the questions here, the bogeyman in all the climate change seems to be CO2.

538
00:55:25,860 --> 00:55:32,460
We're really worried about CO2. But in the real world, CO2 doesn't just sit there as CO2.

539
00:55:32,460 --> 00:55:40,200
We know, for example, there's ammonia in the atmosphere, and as soon as it mixes CO2 molecule in some water, you get ammonium carbonate.

540
00:55:40,200 --> 00:55:48,940
So in all these models, are you simply modelling it with CO2 or modelling with carbonic acid ammonium carbonate or want to as a proxy?

541
00:55:48,940 --> 00:55:57,090
And what I talked about that took CO2 as a proxy for all greenhouse gases, including methane and to and you name it, in the models.

542
00:55:57,090 --> 00:56:01,140
CO2 is it's very inert, which is part of the problem here.

543
00:56:01,140 --> 00:56:10,410
It stays in the atmosphere for tens of thousands of years. So the models there, we don't tend to model reactions of CO2, but we do model reactions,

544
00:56:10,410 --> 00:56:16,350
for example, of methane, which is much less than our average oxidise is much more easily. And other greenhouse gases as well.

545
00:56:16,350 --> 00:56:26,560
So it's just a convenient proxy and talk like that. You don't want to go through all complexity and take it as a proxy for greenhouse gases broadly.

546
00:56:26,560 --> 00:56:33,040
OK, another one or two more questions, but I think we'll we'll sort of call to halt and just say there will be,

547
00:56:33,040 --> 00:56:36,770
I think, a drinks reception around the corner. So correct.

548
00:56:36,770 --> 00:56:44,740
Yes. So any of those who didn't get a chance to to ask tough here, you can colour him over a glass of wine.

549
00:56:44,740 --> 00:56:59,579
So let's finish by giving a big round of applause to separation, either.