the equator and extends too far into the western part of the
basin [Chap. 8 of Solomon et al. (2007)].
Climate Extremes Relative to the Average
We have seen how it is possible to reconstruct average climate conditions from environmental indicators or simulate
them with climate models of varying levels of complexity.
By average we mean both in terms of space (hemispherically, globally or regionally) and time (seasonal, annual and
decadal averages). Thus, most indicators show that the
average temperature of the planet rose by *0.74 °C in the
course of the last century. It is obvious that behind this
average variation there are significant regional and temporal
disparities: not all the places on the planet heat up at the
same rate (high latitudes seem to be more affected than low
ones), and gradual warming does not eliminate the possibility of local cold spells that can last several days. It is even
statistically possible for local cold records to be beaten in a
climate that is warming globally.
If we examine precipitation patterns, we can observe a
trend towards aridification in southern Europe over the last
few decades. This is reflected in a lower average seasonal
water balance than during the first part of the twentieth
century. On the other hand, observations show that extreme
precipitation events (such as the flooding episodes that occur
during the autumn in southern France or the wet winters in
the UK) may become more frequent and more intense.
Typical extreme climate events can be classified into
several categories:
• Heatwaves and cold snaps, i.e. when the temperature
exceeds certain thresholds for a certain duration. The
definition of a heatwave differs depending on the location: a heatwave in Paris is a typical summer in Seville!
• Intense precipitation and droughts. On a one-month scale,
it is important to distinguish between long episodes of
‘average’ precipitation and those that last only a few
hours. They are responses to distinctly different meteorological conditions.
• Storms, cyclones and other hurricanes.
These examples illustrate the use of statistics and probabilities to describe climate phenomena (von Storch and Zwiers
2001). These are particularly relevant when describing the
interactions between normal climatological magnitudes
(mean temperatures and precipitations) and the occurrence of
extreme or rare events. A typical paradigm for describing
temperatures is to represent them as an average and a standard deviation. These are the statistical parameters necessary
to carry out reconstructions of average temperatures over
recent millennia.
Let’s assume that temperature can be represented by a
Gaussian shape, a debatable assumption but one which
simplifies the subject. If an event is defined as extreme when
it is more than two standard deviations from the mean
(which theoretically happens *2.5 times out of 100) then, at
constant standard deviation, it is obvious that the probability
of reaching a given higher temperature increases as the
average temperature increases.
This hypothesis of Gaussian representation doesn’t hold
for daily precipitation, for which the standard deviation is
ill-defined. One of the crucial parameters to describe precipitation is the one that describes the tail of the distribution,
that is, the way rare values occur. A Gaussian possesses
what is referred to as a ‘light’ tail, so that if big events are
possible, they are exponentially rare.
However, the occurrence of rare major precipitations, for
most locations on the planet, decreases much more slowly
than in a Gaussian graph. The theory of extreme values
makes it possible to statistically describe these phenomena
(Coles 2001). Armed with this statistical theory, it is possible
to model the different ways in which average precipitation
and extreme precipitation behave which do not necessarily
respond in the same way to climate change. This statistical
characterization of climate variability makes it possible to
predict the evolution of the probabilities of intense events, as
well as the uncertainty associated with this prediction.
Outstanding Questions at the Beginning
of the Twenty-First Century
Weather or Climate
Variations in the climate of the Holocene up to the present
time have been punctuated by very short-lived meteorological events that have had a variety of consequences, but were
often disastrous for societies. Since the end of the twentieth
century, there appears to have been an upsurge in extreme
events on the planet (the storms of 1999, record monsoons,
deadly heat waves etc.). Assuming that this increase is not
due to improved communication of information on the planet, the first question to be answered is whether this state of
events has been ‘observed’ in the recent past (such as over
the last millennium). To answer this question, it is essential
to gather as much information as possible on extreme events
(of all types) and put them into perspective. The gathering of
this information is the subject of many national and international programs on historical research and the climate
reconstruction of extreme events.
A parallel approach is the numerical simulation of climates
at a resolution sufficiently high to enable the extremes to be
adequately represented (over the current period). The most
sophisticated models can be used to simulate the reported
440
P. Braconnot and P. Yiou
basin [Chap. 8 of Solomon et al. (2007)].
Climate Extremes Relative to the Average
We have seen how it is possible to reconstruct average climate conditions from environmental indicators or simulate
them with climate models of varying levels of complexity.
By average we mean both in terms of space (hemispherically, globally or regionally) and time (seasonal, annual and
decadal averages). Thus, most indicators show that the
average temperature of the planet rose by *0.74 °C in the
course of the last century. It is obvious that behind this
average variation there are significant regional and temporal
disparities: not all the places on the planet heat up at the
same rate (high latitudes seem to be more affected than low
ones), and gradual warming does not eliminate the possibility of local cold spells that can last several days. It is even
statistically possible for local cold records to be beaten in a
climate that is warming globally.
If we examine precipitation patterns, we can observe a
trend towards aridification in southern Europe over the last
few decades. This is reflected in a lower average seasonal
water balance than during the first part of the twentieth
century. On the other hand, observations show that extreme
precipitation events (such as the flooding episodes that occur
during the autumn in southern France or the wet winters in
the UK) may become more frequent and more intense.
Typical extreme climate events can be classified into
several categories:
• Heatwaves and cold snaps, i.e. when the temperature
exceeds certain thresholds for a certain duration. The
definition of a heatwave differs depending on the location: a heatwave in Paris is a typical summer in Seville!
• Intense precipitation and droughts. On a one-month scale,
it is important to distinguish between long episodes of
‘average’ precipitation and those that last only a few
hours. They are responses to distinctly different meteorological conditions.
• Storms, cyclones and other hurricanes.
These examples illustrate the use of statistics and probabilities to describe climate phenomena (von Storch and Zwiers
2001). These are particularly relevant when describing the
interactions between normal climatological magnitudes
(mean temperatures and precipitations) and the occurrence of
extreme or rare events. A typical paradigm for describing
temperatures is to represent them as an average and a standard deviation. These are the statistical parameters necessary
to carry out reconstructions of average temperatures over
recent millennia.
Let’s assume that temperature can be represented by a
Gaussian shape, a debatable assumption but one which
simplifies the subject. If an event is defined as extreme when
it is more than two standard deviations from the mean
(which theoretically happens *2.5 times out of 100) then, at
constant standard deviation, it is obvious that the probability
of reaching a given higher temperature increases as the
average temperature increases.
This hypothesis of Gaussian representation doesn’t hold
for daily precipitation, for which the standard deviation is
ill-defined. One of the crucial parameters to describe precipitation is the one that describes the tail of the distribution,
that is, the way rare values occur. A Gaussian possesses
what is referred to as a ‘light’ tail, so that if big events are
possible, they are exponentially rare.
However, the occurrence of rare major precipitations, for
most locations on the planet, decreases much more slowly
than in a Gaussian graph. The theory of extreme values
makes it possible to statistically describe these phenomena
(Coles 2001). Armed with this statistical theory, it is possible
to model the different ways in which average precipitation
and extreme precipitation behave which do not necessarily
respond in the same way to climate change. This statistical
characterization of climate variability makes it possible to
predict the evolution of the probabilities of intense events, as
well as the uncertainty associated with this prediction.
Outstanding Questions at the Beginning
of the Twenty-First Century
Weather or Climate
Variations in the climate of the Holocene up to the present
time have been punctuated by very short-lived meteorological events that have had a variety of consequences, but were
often disastrous for societies. Since the end of the twentieth
century, there appears to have been an upsurge in extreme
events on the planet (the storms of 1999, record monsoons,
deadly heat waves etc.). Assuming that this increase is not
due to improved communication of information on the planet, the first question to be answered is whether this state of
events has been ‘observed’ in the recent past (such as over
the last millennium). To answer this question, it is essential
to gather as much information as possible on extreme events
(of all types) and put them into perspective. The gathering of
this information is the subject of many national and international programs on historical research and the climate
reconstruction of extreme events.
A parallel approach is the numerical simulation of climates
at a resolution sufficiently high to enable the extremes to be
adequately represented (over the current period). The most
sophisticated models can be used to simulate the reported
440
P. Braconnot and P. Yiou