event, namely the 2003 summer heat wave

that struck Europe.

This is a very, very intense heat wave, I'll just point out in a second,

it was responsible for some estimate of 35,000 to 50,000 deaths,

it was most severe, basically over France.

Now this plot, the vertical blue lines

are a mean summer temperatures from

a number of sites in Switzerland and

these data are from 1864 to 2002.

So every summer you've got an average temperature, right, for a number of cities

in Switzerland, and of course that's where they keep excellent records so

that's probably why this is on this plot.

And so that's what all those vertical blue lines are and then there's,

the average, by the way, of all those is 17.2 degrees,

and then there's a normal distribution around that 17.2 degrees.

Now a normal distribution is a distribution of events that

is a consequence of a perfectly random phenomenon.

And so 68% of all those events are in plus or

minus one sigma of the mean, and

95% are within two sigma 99.7 of the events

are within three sigma and so on and so forth.

So what you can see here is that distribution of summer

temperatures is not a perfect normal distribution, but

it's not too far from a reasonable normal distribution.

The red line labeled 2003, of course, is the summer temperature.

It's 5.4 standard deviations from the mean,

which if you know anything about standard deviations,

the likelihood of that happening in [LAUGH] any kind

of a random circumstance is basically zero.

So it was a very, very unusual event.

Now what's gonna happen?

Let's talk more about extremes in the future, so

[COUGH] let's look at this diagram.

The top panel, now, is a model and

what we wanna do is ask ourselves or a given emissions scenario,

what will the summer temperatures in Europe look like?

Okay, and so let's find a model that actually works.

So first thing we want to do is test that model as best we can and

that's what the upper panel shows.

So we run this model between the years 1961 and 1990,

given the atmospheric CO2 contents,

and it returns a summer temperature for every year, and

that's the purple vertical lines.

The average there is 16 point something or other degrees, so

it's producing an average that's a little cooler than the observed, but

it's not too different from the observed.

Okay, now we've established that we have a model

that is giving some reasonable results.

Let's choose an emissions scenario, that is to say,

let's make an assumption about what future emissions are going to be.

This particular assumption is a high mission scenario.

And let's now ask that model to predict,

to calculate what the temperatures are gonna be for the last 30 years of

the 21st century, and that is the second panel from the top.

So, those are the vertical red bars, okay, so the first thing you can see is

the distribution's far, far wider so there's much, much more variability.

The second thing that you can see is that the average summer temperature

in the last 30 years of the 21st century

is now not too distant from that summer temperature in 2003.

And the third thing you can see is the heat waves in the last 30 years

of the century are gonna be far hotter than the heat wave in 2003.

So that's what we mean about extremes.

We've got to consider the extreme rather than the average conditions.

Tomorrow's extreme could be much larger than today's.

And third and very important point is that the extremes are typically

local in these kinds of events.

We've gotta talk about local extremes not global extremes because

you can talk about two degrees centigrade rise above pre-industrial, but

that's a global thing, right?

That averages everything, so

it's not very meaningful in terms of talking about risks.

So now, what we have plotted here, on the ordinate,

is the probability density versus the climate sensitivity to

doubling of the CO2 content Of the atmosphere.

Now there's several curves here.

Let's just consider that green curve for the sake of convenience.

All right?

Now I want to describe what that means.

Okay, so we're using a model.

And we're saying that we're going to double CO2 content and

the model returns a global temperature.

Okay?

And you do that over and over and

over again, and it returns different global temperatures.

So what the probability density means is that the sensitivity of CO2 three degrees,

so what that's saying is that the probability is 50%.

This gives a 50%, let's say it's right there, this gives a 50% probability.

Okay? So 50% of the time,

the climate model is returning three degrees.

Okay.

2% of the time, the climate model is returning about 25.

Excuse me.

25% of the time, the model's returning two degrees.

10% of the time, the model's returning five degrees.

But the reason I point this out is this huge,

huge tail, this long tail on the positive side.