untitled

Hazard & Risk Science Review 2009

Foreword

Aon Benﬁeld and PartnerRe are proud to present the Hazard & Risk Science Review 2009,
the sixth edition of this unique publication.

In a year that has seen destruction caused by the L’Aquila earthquake in Italy and the
deadliest wildﬁres in Australia’s history, with insured losses reaching as high as AUD1.5bn,
it is clear that there continues to be a strong requirement for in-depth research into natural
hazards.

The Hazard & Risk Science Review once again assists in bridging the gap between
re/insurance and academia by providing a synopsis of key research papers deemed most
relevant to the industry. The 2009 Review continues to focus on the four main areas of
hazard: atmospheric, geological, hydrological and climate change-related, and summaries
of more than 70 scientiﬁc papers published during the past 12 months are featured in the
publication. Academics from the Aon Benﬁeld UCL Hazard Research Centre worked with
natural catastrophe experts from both Aon Benﬁeld and PartnerRe to source the papers
that would provide most assistance in developing appropriate insurance and reinsurance
solutions.

Our organisations both have a long-standing commitment to research, and we believe that
the Review continues to assist the industry in developing its understanding of emerging risk
science by providing the most up-to-date information in this area.

We hope that you ﬁnd this year’s publication a useful reference point.

Andrew Appel

Patrick Thiele

Chief Executive Ofﬁcer

President and Chief Executive Ofﬁcer

Aon Benﬁeld

PartnerRe

Authors’ note

While the last 12 months have not seen a catastrophe on the scale of the previous year’s
Cyclone Nargis or Sichuan earthquake, devastating wildﬁres in Australia and the Abruzzo
earthquake in Italy have served to remind us that western developed nations are far from
immune to natural disasters. As the world’s population grows and becomes increasingly
concentrated into ever-larger urban centres, so exposure and vulnerability to natural hazards
will inevitably grow – in developing and developed nations alike. Climate change adds to the
uncertainty and may increase the frequency and/or severity of major storm and ﬂood events.
In this year’s Review, we present a digest of peer-reviewed papers that address critical
issues, including hazard characterisation, forecasting and modelling that, together, seek to
reduce vulnerability and exposure and diminish disaster risk. Our aim continues to be to
improve industry awareness and understanding of natural catastrophes and the processes
that drive them, to limit the number of shocks and surprises arising from hazardous events,
and to help drive more informed business decisions on a day-to-day basis.

Professor Bill McGuire
Dr. Steve Edwards
Dr. Simon Day
Dr. Harvey Rodda
Mr. David Smart
Aon Benﬁeld UCL Hazard Research Centre

Aon Benﬁeld
55 Bishopsgate
London
EC2N 3BD
T: +44 (0)20 7088 0044
www.aonbenﬁeld.com

PartnerRe
Wellesley House
90 Pitts Bay Road
Pembroke HM08
Bermuda
T: (1441) 292 0888
F: (1441) 292 7010
www.partnerre.com

Hazard & Risk Science Review 2009

1. EXECUTIVE SUMMARY

Pertinent ﬁndings and conclusions arising from scientiﬁc papers and reports addressed
herein are individually highlighted below. The superscripted numbers following individual
headlines take the reader directly to the relevant paper(s) in Sources and Further Reading at
the end of each section

Geological hazards

•

Potential and challenges of earthquake early-warning systems in Italy, Turkey and
California.

2, 3, 27, 28

•

New studies shed light on attenuation of seismic waves in the Marmara Sea (Turkey)
region and northern Italy, and examine implications for seismic intensity distributions.

17, 23

•

Borehole data suggest lower seismic wave attenuation rates for larger earthquakes,
relative to smaller ones.

•

Model of ground motions produced by large to giant Cascadia Subduction Zone
earthquakes, calibrated for cities of Seattle, Vancouver and Victoria.

•

New earthquake simulations highlight potential for stronger shaking in southern California
and the Seattle area.

9, 11, 20

•

New earthquake hazard map for the Central Apennines region of Italy ﬂags high level of
near-future seismic hazard in L’Aquila area – within days of the April 2009 earthquake.

•

Slow subduction zones ﬁngered as potential sources of major tsunamigenic earthquakes,
as well as their faster-converging relatives.

•

Future tsunami threat from volcanic landslides in the Canary Islands and
Alaska highlighted.

7, 15, 25

•

Tsunami hazard evaluated for the Caribbean, Oregon coast and Portugal.

5, 21, 26

•

New studies address building vulnerability in relation to future eruptions at Vesuvius (Italy)
and Teide (Tenerife, Canary Islands).

16, 29

Hydrological hazards

•

UK extreme rainfall events back to 1866 analysed and complied to form digital archive.

•

Future ﬂood hazard in Rome and Venice evaluated.

32, 37

•

Analysis of ﬂood trends reveals an increase in ﬂood hazard in Germany in the last half
century, particularly in relation to ﬂood frequency.

•

Role of uncertainties highlighted in creation of ﬂood inundation maps.

•

Problem of characterising and predicting urban ﬂood damage addressed in studies of ﬂood
events in the Netherlands and Germany.

30, 39

•

Innovative modelling technique used to evaluate ﬂood hazard at river conﬂuences.

Hazard & Risk Science Review 2009

Atmospheric hazards

•

North Atlantic hurricane activity linked to sea-surface temperature patterns in the Paciﬁc.

50, 53

•

A study including an estimate of “missed” tropical cyclones in the pre-satellite era
suggests a signiﬁcant increase in North Atlantic hurricane numbers reported for the period
1900 to 2006.

•

Atlantic sea-surface temperatures and other parameters linked to frequency and location of
US land-falling hurricanes.

46, 51

•

New study reveals an apparent decline in the annual numbers of Southern Hemisphere
tropical cyclones since 1970, although not for the most powerful storms.

•

No signiﬁcant change observed in tropical cyclone activity across East Asia.

44, 60

•

Problem of the paucity of historic data in evaluating hurricane landfall probability overcome
by creation of synthetic dataset of North Atlantic storm tracks.

•

Multi-season forecast of Atlantic hurricane activity claimed possible out to ﬁve years.

•

Inﬂuences on storminess across the British Isles from 1920 to 2004 suggested.

41, 42

•

Return periods of Europe’s most prominent windstorms estimated.

Climate change

•

Critical role of sea-surface temperatures examined in relation to tropical cyclone activity in
a warmer world.

61, 71, 75, 77

•

Signiﬁcant, general reduction in tropical cyclone frequency projected as the concentration
of atmospheric carbon dioxide is doubled and quadrupled, particularly for the North Atlantic
and tropical western North Paciﬁc.

•

The Atlantic hurricane season appears to be getting longer, but the trend cannot be used to
predict future changes to the season.

•

Satellite wind-speed data conﬁrm a signiﬁcant overall upward trend for wind-speeds in the
most powerful tropical cyclones, across the tropics.

•

A fall in winter cyclone numbers is projected in the Northern Hemisphere, although a rise in
the number of intense cyclones is forecast across the NE Atlantic, the British Isles and the
North Paciﬁc.

•

A warmer world is likely to bring a higher incidence of damaging wind events
across Europe.

•

Jury out on a climate change signal in US tornado activity.

•

Climate change projected to drive an increase in the frequency and magnitude of extreme
discharge levels in many European rivers.

Hazard & Risk Science Review 2009

2. INTRODUCTION

In any single year, thousands of scientiﬁc papers addressing natural hazards, the processes
and mechanisms that drive them, and their impacts and ramiﬁcations, are published in
hundreds of journals and e-journals. Many of these papers contain information that is of
direct relevance and considerable importance to the insurance market, but which can take
several years to ﬁlter down from academia to the business world. The Hazard and Risk
Science Review is designed to accelerate this process by drawing attention to new and
pertinent research results. We present a summary of key publications from 2008 - 2009
with the aim of introducing the reader to current themes in a number of research domains.
The summary text is intended to provide an introduction to the original work and its authors,
by setting the science in a broader context and showing, where appropriate, its potential
relevance to our business. Inevitably the new research addressed in the 2009 issue is only a
tiny sample of the enormous amount of relevant material published over the last 12 months,
and we do not claim that it is entirely representative of published research over the period.
It does, however, highlight studies recommended by researchers and consultants of the Aon
Benﬁeld UCL Hazard Research Centre and natural catastrophe experts at Aon Benﬁeld and
PartnerRe that are considered to be particularly relevant to the interests of those involved in
catastrophe insurance and reinsurance. The success of the review, launched in September
2004 at the Monte Carlo Reinsurance Rendez-Vous, is reﬂected in the ﬁfteen thousand
copies of the ﬁrst ﬁve issues distributed throughout the market. We conﬁdently expect this
sixth Hazard and Risk Science Review to achieve an even wider circulation and further
augment its function as an invaluable resource to the market, through providing a conduit
from the arena of academic research to the business world. The Hazard and Risk Science
Review is published annually in September, and incorporates research published during the
twelve months to the preceding June.

This year’s Review continues to adopt the straightforward fourfold structure developed in
the launch issue in 2004, centring on geological, hydrological and atmospheric hazards,
and on climate change and its hazard implications. With the L’Aquila earthquake disaster in
Italy highlighting the fact that western Europe is far from immune to destructive geological
phenomena, we lead this year with a summary of new research in geological hazard and
risk. In particular, we address earthquake ground-motion studies in Italy, Europe and
elsewhere, and examine the issue of earthquake early-warning. A common theme is that
current event, and so insurance loss, models may understate inherent event uncertainties.
In a change from previous issues of the Review, the papers and articles cited herein are
now listed at the end of each section, providing a resource for the reader who wishes to
pursue follow-up investigations. For ease of use, key themes covered within the Review are
highlighted in the Executive Summary, with numbered links to the relevant publications in the
bibliographies at the end of each section.

Hazard & Risk Science Review 2009

Figure 1

Debris-blocked streets
in the town of L’Aquila
following the magnitude 6.3
earthquake in April 2009.
The quake, the largest in
Italy, since 1980, killed
close to 300 people and
damaged or destroyed up
to 11,000 buildings, mainly
in L’Aquila and surrounding
villages. Reconstruction
costs are estimated at
around USD16 billion.

Courtesy: Wikipedia.

3. GEOLOGICAL HAZARDS

While the resulting loss of life and level of destruction was far from on a par with the 2008
Sichuan (China) earthquake, the Moment Magnitude 6.3 event that struck the Abruzzo
region of Central Italy in April (Figure 1) provided a reminder that western Europe is also
periodically affected by lethal earthquakes. The quake, the largest in Italy, since 1980, killed
close to 300 people and damaged or destroyed up to 11,000 buildings, mainly in the town of
L’Aquila and surrounding villages. Reconstruction costs are estimated at around USD16 bn,
with total economic losses far higher. Without doubt, the L’Aquila event has served to keep
the risks from earthquakes in urban areas at the forefront of catastrophe science.

Earthquake warning and rapid response systems

Whilst a number of research groups have continued to pursue the elusive goal of speciﬁc
prediction of earthquakes, days or even months in advance, another and perhaps more
proﬁtable area of research and development, has focussed on the seconds to tens of
seconds that elapse between initial detection of an earthquake rupture on seismometers
near the source and the peak intensity of shaking in nearby urban areas. The concept
behind these studies is that of providing near-instantaneous automatic alarms to critical
or hazardous facilities that can then be shut down in order to minimise damage from the
earthquake. Papers published in Geophysical Research Letters by A. Zollo

(University of

Naples, Italy) and colleagues; H. Alcik

(KOERI, Bogazici University, Istanbul, Turkey) and

co-authors; and R.M. Allen

and others (University of California, Berkeley, USA), describe

the implementation of such systems in, respectively, the Campania district of Southern Italy;
the greater Istanbul area; and northern California.

All three papers focus on the problems of the automated data networks required to efﬁciently
distribute the warnings and the algorithms needed to reliably analyse the warning signals
(the amplitudes and wave-forms of the initial P-wave vibrations that travel ahead of the more
damaging seismic waves) to provide near-instantaneous estimates of the strength, duration
and frequency spectra of the damaging ground shaking that will follow. Such automatic
systems need to be very carefully implemented, in particular to avoid false alarms: in their
study Allen and colleagues emphasise the need to use records from multiple seismic
stations, which implies dense seismic networks in the area covered by the system in order
to ensure that the alarm is triggered in time to be useful. Similarly, Alcik and co-authors
describe how multiple exceedances of threshold accelerations within ten-second intervals,
as recorded at seismic stations within a network speciﬁcally designed to monitor the sections
of the North Anatolian Fault just south of Istanbul, are to be used as a basis for alarms to
critical facilities in the Istanbul area.

A similar concept to these seismometer-based warning systems, but on a global scale
and on a longer timescale, can provide initial quantitative estimates of damage within an
hour of an earthquake occurring. Writing in Natural Hazards, M. Wyss

World Agency of

Hazard & Risk Science Review 2009

Figure 2

Zones of deﬁned minimum
earthquake magnitude for
which World Agency of
Planetary Monitoring and
Earthquake Risk Reduction
(WAPMERR) conducts
a Seismic Data Review
(SDR) and generates initial
loss estimates (Wyss and
Zibzibadze 2009). In the
blue zone, WAPMERR only
generates loss estimates for
earthquakes with magnitude
over 8.0.

Courtesy: Max Wyss.

Planetary Monitoring and Earthquake Risk Reduction (WAPMERR), Geneva, Switzerland
and M. Zibzibadze (Georgia Seismic Monitoring Center, Tiblisi) show how global data from
different information centres such as the National Earthquake Information Center (NEIC),
USA, provide earthquake alerts and geophysicist-reviewed magnitude estimates that are
used by WAPMERR to provide initial loss estimates for earthquakes exceeding different
threshold magnitudes in different parts of the world, with a median delay of only 41 minutes
after the start of the earthquakes concerned; as they note, a similar service aimed at providing
initial impact assessments for disaster relief agencies is provided through the United States
Geological Survey (USGS) PAGER webpage (Figure 2). Such initial assessments are,
however, only as good as the magnitude estimates provided by the global seismic networks
and the key problem of providing timely identiﬁcation of the very largest earthquakes (as
emphasised by the gradual upgrading of the magnitude of the 2004 Sumatra earthquake over
a period of days) still remains.

Earthquake ground motion estimates and models

The models relating different parts of the earthquake wave-form that are built into the
automated warning system algorithms are just one of many applications of studies that
relate earthquake ground motions to earthquake sources. Central to these studies is the
problem of attenuation of seismic waves with distance from the source fault rupture as
they travel through the Earth. Attenuation varies between different regions of the Earth but
also with the amplitudes and frequencies of seismic waves: the complexities of attenuation
relationships therefore require both model and detailed empirical studies. In the past
year, detailed empirical studies of particular regions include work published in Bulletin
of the Seismological Society of America (BSSA) on the Marmara Sea area of Turkey by
M.B. Sørensen

and co-workers at GFZ – Potsdam, Germany, who show that inclusion

of an estimate of the ﬁnite fault rupture length in the attenuation equation increases the
accuracy of the estimate of felt seismic intensity distributions in this key region; and a
study of attenuation relations in Northern Italy by M. Massa

(INGV-Milano, Italy) and

colleagues, also published in BSSA. The latter considers a whole range of intensity
parameters including acceleration spectra over the ranges of damaging frequencies,
and shows that many of the key measures of seismic intensity attenuate more rapidly for
larger earthquakes in that region than was previously thought to be the case. Their work
implies that such earthquakes may produce smaller losses, over smaller areas, than was
previously thought, but also that accurate assessments of earthquake risk may be more
dependent upon knowledge of the speciﬁc characteristics of faults closest to concentrations
of population and risk exposure.

Hazard & Risk Science Review 2009

Attenuation relationships on wider scales are the subject of studies, both published in
BSSA, by D. Stromeyer

and G. Grünthal (both at GFZ-Potsdam, Germany) and

F. Cotton

(Universite Joseph Fourier, Grenoble, France) and colleagues. Stromeyer

and Grunthal improve on the ﬁt of their empirical attenuation equation to a large number
of intensity data from European earthquakes by describing the data in terms of normal
distributions (with a standard deviation of 0.7 magnitude units) about the mean attenuation
relationship. In contrast to the results of Massa and colleagues for northern Italy, Cotton
and co-workers use data from borehole accelerometers in Japan and elsewhere to indicate
lower attenuation rates for larger earthquakes than smaller ones, but also a decrease in
the sensitivity of attenuation rate to magnitude for larger earthquakes; again underlining
the difﬁculty of extrapolating attenuation models from small, frequent earthquakes to the
larger earthquakes that are of most interest for the understanding of earthquake risk.

Ground motion models and attenuation relationships appropriate to earthquakes in a
different tectonic setting, that of the subduction zones where the largest earthquakes
occur, are the subject of a study (also published in BSSA) by G.M. Atkinson

and

M. Macias of the University of Western Ontario, Canada. They use strong-motion
seismic records from the 2003 M8.1 Tokachi-Oki earthquake and its aftershocks in
Japan to calibrate a model of ground motions in the cities of Vancouver, Victoria and
Seattle produced by large to giant (moment magnitude M7.5 to M9.0) earthquakes on
the Cascadia subduction zone. They emphasise the need to use magnitude-dependent
attenuation relations speciﬁc to particular types of subduction zone and even speciﬁc
regions, rather than attempting to produce a globally – applicable set of attenuation
relationships in subduction zones. Overall, then, the picture to be gained from recent
research into attenuation relationships is that signiﬁcant levels of uncertainty exist even
in the best-studied regions, and that in reducing these uncertainties the models become
ever more region-speciﬁc. Accordingly, there is more interest in simulation of individual
earthquakes in a scenario type approach.

Earthquake simulations

The advantages of the earthquake simulation approach is that it allows investigation
of directivity of earthquake shaking due to ruptures on particular faults (typically with
more intense shaking in the direction of rupture propagation), and the dependence
of shaking patterns on rupture histories (including rupture propagation rates and
displacement histories).

These aspects are particularly important in assessing earthquake risk where exposures
are concentrated in small parts of the area of shaking. K.B. Olsen

(San Diego State

University, California) and co-workers use the simulation approach in a paper published
in BSSA to study a northward-propagating rupture on the southern San Andreas Fault,
extending from near the Mexican border to an area just east of Los Angeles (Figure 3).

Shaking from their model earthquake of moment magnitude Mw=7.7 is particularly intense
in parts of the densely populated Los Angeles basin, greatly exceeding the intensities
predicted by empirical models that do not include rupture direction effects. They also show,
however, that the size of the discrepancy depends on the choice of rupture history and
displacement heterogeneity in the source earthquake, as the directivity effect in this study
is lower than in a previous study by the same authors which assumed a more continuous
(or less “jerky”) fault rupture history and more uniform displacement on the fault in the
earthquake. A similar conclusion is reached by R.W. Graves

from URS Corporation

and co-workers in a study published in Geophysical Research Letters, who simulate a
Mw=7.8 Southern San Andreas Fault earthquake and show that the patterns of ground
accelerations are strongly dependent upon the fault rupture velocity.

Hazard & Risk Science Review 2009

Bakersfield

Lancaster

Los Angeles

Victorville

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Bakersfield

Lancaster

Los Angeles

Victorville

Burst

TS2.3

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Bakersfield

Lancaster

Los Angeles

Victorville

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Bakersfield

Lancaster

Los Angeles

Victorville

Burst

TS2.2

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Bakersfield

Lancaster

Los Angeles

Victorville

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Bakersfield

Lancaster

Los Angeles

Victorville

Burst

TS2.1

Riverside

Palm Springs

San Diego

Mexicali

Yuma

Another application of the simulation approach is provided by a study of the effects of
ampliﬁcation of seismic waves in the Seattle Basin by A. Frankel

and colleagues at

USGS Denver, Colorado. They simulate a series of earthquakes of sizes up to the Mw=6.8
Nisqually earthquake, including deep earthquakes within the subducted slab that produce
shaking frequencies in the resonance range of many building types, and show a directional
effect with earthquake ruptures to the southwest producing stronger shaking in the Seattle
area than earthquakes to the north and the northwest, because of the curvature of the
basin margin which causes focussing of seismic waves within the basin. Their study shows
that detailed crustal structure models need to be included in accurate earthquake risk
assessments both at regional level and for site-speciﬁc risk assessments.

Although earthquake simulations have, as yet, only been carried out in a small number
of areas and for a small number of earthquake sources, they underline the limitations of
more generalised earthquake damage models – particularly global earthquake risk models,
used for example in reinsurance applications – and the need to recognise and allow for the
signiﬁcant uncertainties contained in these more general models.

Investigations of earthquake sources

The increased emphasis upon scenario models, indicated by the studies described in the
previous section, renews the need for detailed understanding of the earthquake sources
themselves. Key problems in this area of research include the size – frequency distribution
of earthquakes; the controls on the maximum size of earthquake that can be expected in

Figure 3

Three simulation models
of peak ground acceleration
from a Mw=7.7 earthquake
generated by a northward-
propagating rupture on the
southern San Andreas Fault
zone (Olsen et al. 2008).
Note the zones of high
acceleration in the
Los Angeles basin in all
three simulations.

Courtesy: Kim Olsen.

Hazard & Risk Science Review 2009

a given region; and the underlying question of whether conventional measures such as
moment magnitude are fully satisfactory as quantiﬁcations of earthquake source size.
Studies published in the past year have addressed all of these problems.

In a short research note published in BSSA, T. Parsons

and E. Geist of the USGS (Menlo

Park, California) question the assumption made in many regional and local earthquake
risk models, that the earthquake hazard produced on individual faults can be described in
terms of repetitions of characteristic earthquakes speciﬁc to the fault concerned. Instead,
they show that the earthquake histories of particular faults can be equally well described by
the Gutenberg-Richter distribution of earthquake sizes used to analyse regional or global
earthquake catalogues containing earthquakes on many faults, with earthquake frequency
decreasing with magnitude according to a logarithmic relationship. Their study lends new
weight to the long-standing idea that all faults (as deﬁned at large map scale) are really fault
zones containing large numbers of linked smaller faults of a wide spread of sizes, that may
rupture together in different combinations or independently.

In such a situation, the upper limit on earthquake magnitude in a given region is governed
by the thickness of the seismogenic zone, which is bounded by the Earth’s surface (or by a
near-surface zone of continuous fault creep) above, and the increase of temperature in the
deeper crust, that ultimately leads to the onset of ductile rock deformation or ﬂow at 350 to
450 degrees centigrade. This implies that regional-scale variation in maximum earthquake
size should correlate with the geothermal temperature gradient, which can be independently
measured. This concept has long been applied to explain why subduction zone thrust
earthquakes, in regions of exceptionally low thermal gradient resulting from the subduction
of cool oceanic lithosphere, are so very large. T. Kudo

(Chubu University, Japan) and

colleagues, in another study published in BSSA, extend the concept to intra-crustal or
upper plate earthquakes in Japan. They show that variation in maximum earthquake size
across Japan correlates with geothermal gradient measurements. Such measurements
may therefore be usefully incorporated in earthquake hazard and risk models, especially in
regions with short earthquake catalogues that are likely to be unrepresentative of the long-
term earthquake hazard from large earthquakes.

A third paper dealing with Japanese earthquakes has however highlighted the question of
whether the standard measure of earthquake size, the moment magnitude Mw (a measure
of the seismic moment or torque or the earthquake), is a fully valid measure of the seismic
energy of the earthquake, which relates more closely to the ground accelerations (and
hence, damage) that it produces. Writing in BSSA, G.L. Choy

and J. Boatwright (USGS,

respectively at Menlo Park and Denver) compare the 1996 Kyushu and 2000 Tottori
earthquakes, events with similar moment magnitudes (Mw=6.7). Using far-ﬁeld seismic
records, they show that the energy magnitudes of the two events were very different (1996
Kyushu Me=6.6; 2000 Tottori Me=7.4) and interpret this in terms of differing fault geology,
with the Kyushu earthquake occurring in weak fault rocks within a mature subduction zone
whilst the Tottori earthquake occurred in stronger, immature fault rocks in the crust of Japan.

Earthquake hazard mapping

The various studies on earthquake sources, simulations and ground motions described
above all raise questions regarding the validity of assumptions built into many existing
methods of constructing earthquake hazard maps. Despite this, hazard maps are key to
current methods of seismic risk assessment, and the past year has seen one striking, if
coincidental, success in hazard mapping. A. Akinci

Instituto Nazionale di Geoﬁsica e

Vulcanologia (INGV), Rome, Italy and colleagues prepared a time-dependent earthquake
hazard map and assessment for crustal faults in the central Apennines region of Italy,
published in BSSA in April 2009 (Figure 4). Their study, which assumed characteristic
earthquakes on the faults, used different levels of aperiodicity (departures from constant
recurrence intervals of earthquakes) to study the extent to which hazard levels on individual
faults vary with the times since the last major earthquakes on these faults. They showed that

Hazard & Risk Science Review 2009

hazard levels in the region were principally controlled by the local faults; and that especially
if aperiodicity was low, the near-future probabilistic seismic hazard was locally much higher
than would be the case if earthquake recurrence was random or aperiodic. One area that
they identiﬁed as having particularly high levels of near-future seismic hazard was the city
of L’Aquila, which by coincidence was severely damaged in an earthquake within days of
publication of the paper.

Great earthquakes and tsunami generation

There is wide acceptance of the hypothesis that the greatest earthquakes are generated
in subduction zones where the rate of plate convergence is high (≥60 mm/year) and the
age of the down-going oceanic lithosphere is young (usually >10 million years), which is a
relationship supported by the huge Chile-Valdivia earthquake of 1960. This simple velocity-
age relationship is, however, being challenged by evidence emerging from the seismic
record and by several recent great earthquakes, such as the 2004 Boxing Day Sumatra-
Andaman event, which was generated in a slow (≤40 mm/year) subduction environment. In
a chapter in the book Subduction Zone Geodynamics, M.-A. Gutscher

, Université de Brest

(France), and G. Westbrook, University of Birmingham (UK), consider these environments,
because their potential to generate great earthquakes and associated tsunami is now
recognised, e.g., Nankai, Cascadia, Gibraltar (southwest Iberia) and Sumatra (Figure 5), but
poorly understood. Giant earthquakes generated in slow subduction zones recur at a lower
rate (of the order of several hundred to a few thousand years) relative to rapidly converging
margins. This low recurrence poses a signiﬁcant problem for earthquake and tsunami risk
assessment because it is difﬁcult to properly assess the long-term seismic hazard presented
by these slow zones. Comparison of the physical characteristics of slow and fast subduction
systems clearly shows that the former have thicker and wider wedges of accreted sediment
and a wider and shallower seismogenic zone. Because the wedge has low rigidity, an
earthquake of a given seismic moment will produce more coseismic slip, and for a longer
duration, than a deeper earthquake in more consolidated material. Consequently, shallow
earthquakes in the wedge are highly efﬁcient at generating strong tsunami.

Figure 4

Predicted probabilistic
peak ground acceleration
(PGA) maps for the central
Apennines constructed
using different values of
aperiodicity coefﬁcient
(low   corresponds to more
periodic earthquakes;  =1
is a random or Poissonian
distribution) by Akinci et al.
(2009). Note: L’Aquila is
located in the central high
PGA probability zone.

Courtesy: Aybige Akinci.

Hazard & Risk Science Review 2009

In stark contrast to slow subduction systems, K. Furlong

of the Pennsylvania State

University, and co-researchers, examine the 1 April 2007 great tsunamigenic earthquake
(moment magnitude 8.1) that ruptured the Solomon Islands subduction zone at the triple
junction where the Paciﬁc plate is underthrust independently by both the Solomon Sea–
Woodlark Basin (135 mm/year) and Australian (97 mm/year) plates. Their paper in the
journal Science demonstrates that in addition to the rupture traversing major geological
boundaries, the earthquake was signiﬁcant in other ways: ﬁrst, it contravened speculation
that subducting extremely young hot oceanic lithosphere will not produce great earthquakes;
and second, it demonstrated that coseismic displacement was concentrated in the overriding
plate, leading to signiﬁcant surface uplift. The latter is particularly important because it
increases the potential of an earthquake to both generate and amplify tsunami.

Mass movement and tsunami generation

Only recently are tools being developed that model the source, propagation and inundation
components of a tsunami event. In terms of sources, earthquakes are the most obvious,
but tsunami may also be generated by mass movement (volcano collapse, landslides
into the sea and huge submarine landslides) and by volcanic eruptions. There is a great
deal of interest in tsunami generated by instabilities on volcanic islands, for which the
Canary Islands present excellent examples. It is now well known that a large-scale failure
of the ﬂank of the Cumbre Vieja volcano on La Palma has the potential to generate large
amplitude waves that could have huge impact on the coastlines of the North Atlantic.
Evidence is now emerging that the same coasts could be affected by a giant landslide
originating in the Icod valley on Tenerife, which is the focus of the paper in Earth and
Planetary Science Letters by P. Coppo

and co-researchers of Switzerland’s University

of Neuchâtel. In modelling the potential tsunami hazard, it is essential to quantify the
volume of material that is likely to move, and the mechanism and speed of mass failure
and transport. The paper considers the ﬁrst parameter and reports on a geophysical
survey to constrain the sliding surface. It estimates the potential volume to be at least
100 km

, which may become displaced during the next strong felsic eruption of the Teide-

Pico Viejo volcanic complex. The authors stress that more studies are urgently required
to understand the structure and behaviour of volcanic islands prone to potential ﬂank
collapse, and that this information needs to be incorporated into models that feed into the
production of a global map of tsunami hazard.

Figure 5

Global distribution of known
slow subduction zones,
superimposed on a strain
map (from Gutscher and
Westbrook 2009). Of the
13 slow zones marked,
ten contain very broad
accretionary wedges (solid
circles) and three do not
(dashed circles).

Courtesy: Marc-Andre
Gutscher.

Hazard & Risk Science Review 2009

The potential failure volume in the Icod valley correlates well with the typical estimates of 50–
200 km

from Canary Island ﬂank failures, but it is a failure of the Cumbre Vieja volcano in

La Palma that is most likely, as this volcano is growing more rapidly than any other volcano
in the Canary Islands. Previous studies of this potential failure and its consequences have
been newsworthy, particularly the most extreme event involving simultaneous failure of up to
500 km

. Some scientists have argued that such an event is unlikely, but in their paper in the

Journal of Geophysical Research, F. Løvholt

of the Norwegian Geotechnical Institute, and

colleagues, have re-modelled the consequences of a Cumbre Vieja collapse. By combining
a multimaterial model for the wave generation with Boussinesq models for the far-ﬁeld
propagation, they generate an initial wave of several hundred metres height that is directed
southwestwards, which would have catastrophic consequences for the coastal regions of the
Canary Islands. The oceanic propagation of the tsunami would affect the whole of the central
Atlantic, but the largest waves are smaller than the most pessimistic previously published
estimates; for example, surface elevations are 2–3 times smaller than the worst estimates of
25 m along the coastline of Florida. However, these differences largely reﬂect uncertainties
in the landslide source, as the rates of attenuation of the waves during their transoceanic
propagation are very similar to the results of the previous studies.

Continuing with the theme of mass movement and tsunami generation, C. Waythomas

of the USGS Alaska Volcano Observatory, and co-researchers, consider the hazard
originating in the Aleutian volcanic island arc, which forms the northern rim of the Paciﬁc
Ocean basin. Writing in Quaternary Science Reviews, they numerically simulate previously
unrecognised tsunami hazards generated from three sources: submarine mass ﬂows
originating in submarine canyons, mass ﬂows that evolve from submarine landslides on
the trench slope, and failures of the ﬂanks of volcanoes. All simulations produce run-ups
of 10–20 m above sea level along coastlines nearest to the tsunami source. They also
produce transoceanic tsunami, with the largest generated by submarine ﬂows. Some of
these tsunami produce waves that are several metres in elevation at far-ﬁeld locations,
such as Japan, Hawaii, and along the North and South American coastlines, where they
may present a signiﬁcant hazard.

Tsunami hazard assessment: probability and exposure

We begin this section in the Caribbean by referring to the paper in Pure and Applied
Geophysics by T. Parsons

and E. Geist of the USGS at Menlo Park, who have calculated

tsunami run-up (>0.5 m) probabilities at coastal sites in the region. They constructed an
empirical tsunami probability map from the exceptional written records of events, which
date back to 1498, and tested this against a second map they generated from numerically
calculated run-up rates. Interestingly, this comparison demonstrated that the empirical model
may under-represent tsunamis sourced by plate boundary earthquakes (reﬂecting, perhaps,
the large interval between such events even compared to the ~500 year long historical
earthquake catalogue for the Caribbean, as noted by Gutscher & Westbrook). In contrast,
the synthetic catalogue derived from the numerical model lacks sources linked to landslides
and back-arc faulting. In order to overcome these respective shortfalls, a third probability
map was produced using a Bayesian method, which combined tsunami probabilities
calculated from the synthetic catalogue with results of the empirical model.

The best-estimate results from this last method are portrayed in (Figure 6), which shows that
the highest probabilities (typically 10–20% in 30 years for run-up ≥0.5 m) are for the eastern
Lesser Antilles, including the islands of Antigua, Barbuda, Dominica, Guadeloupe, Martinique,
Grenada, St. Kitts, Nevis, St. Lucia, St. Vincent and the Grenadines. There is, however, a need
to test the validity of these alternative probability maps with independent data, for example
from palaeoseismic studies and investigation of prehistoric tsunami deposits.

Hazard & Risk Science Review 2009

-DPDLFD

3XHUWR 5LFR

/HVVHU $QWLOOHV

%DKDPDV

3DQDPD

&RVWD 5LFD

1LFDUDJXD

+RQGXUDV

%HOL]H

&D\PDQ ,VODQGV

0H[LFR

+LVSDQLROD

&XED

&ROXPELD

9HQH]XHOD

%HVW HVWLPDWH \U SUREDELOLW\

.LQJVWRQ

5RVHDX

&DVWULHV

)RUWGH)UDQFH

.LQJVWRZQ

6W *HRUJH¶V

6DQ -XDQ

5RDG 7RZQ

&RFNEXUQ 7RZQ

&DUWDJHQD

3XHUWR /LPRQ

%OXHILHOGV

3XHUWR
%DUULRV

%HOL]H &LW\

*HRUJH 7RZQ

6DQWR
'RPLQJR

3RUW$X3ULQFH

6DQWLDJR

GH &XED

%DVVHWHUUH

%DVVH7HUUH

:LOOHPVWDG

&XQDQD

/D &HLED

&RORQ

6W
-RKQ¶V

3RUW RI
6SDLQ

Having identiﬁed, modelled and mapped regions likely to be affected by a tsunami, it is
necessary to combine this information with the type of land use within deﬁned tsunami-hazard
zones. To this end, N. Wood

of the USGS in Washington State estimates tsunami exposure

along the Oregon coast, linked to a tsunamigenic earthquake along the Cascadia subduction
zone. The method is presented in Applied Geography and uses mid-resolution (30 m) Landsat
Thematic Mapper land-cover data overlain on tsunami hazard zone data. This provides a
simple ﬁrst step for visualising and estimating potential exposure and it may also be adapted
to allow for broad-scale assessments of critical infrastructure and economic impacts.

Superimposition of a past tsunami event on present-day land-use and economic conditions is
another way of gleaning important information for hazard assessment. In this context, writing
in the journal Geography, D. Chester

of the University of Liverpool revisits the 1755 Lisbon

earthquake and tsunami. Together these cost Portugal around 32-48% of its gross domestic
product, which probably makes the 1755 event the greatest natural disaster in ﬁnancial
terms to have impacted western Europe. If a similar event were to occur today (the worst-
case scenario has a minimum estimated recurrence interval of 614±105 years), it would
still have catastrophic consequences, as the Algarve region is essential to the Portuguese
economy, particularly as it represents one of Europe’s main tourist destinations. The resident
population of this region is over 400,000, but the inﬂux of tourists during the summer months
more than doubles this number. The potential hazard is clear, when one considers that the

Figure 6

Thirty-year probability of
tsunami run-up (≥0.5 m) in
20 x 20 km cells at coastal
sites in the Caribbean region
(from Parsons and Geist
2008). The probability is
derived from combined rate
estimates from empirical
and numerical models.
Many of the important cities
in the region are identiﬁed
for reference.

Courtesy: Tom Parsons,
United States
Geological Survey.

Hazard & Risk Science Review 2009

1755 tsunami took about 16–30 minutes to reach different parts of the Algarve coast, and
it is estimated to have penetrated up to 2.5 km inland and had run-up heights reaching 20,
perhaps even 30, metres. Chester concludes by calling for more hazard assessment and
mapping in the Algarve, which is already underway in many areas. It must not be forgotten
that strict building codes, which have been applied across the whole country and periodically
updated, were pioneered in Portugal following the 1755 event.

Risk from explosive volcanic eruptions in Europe

Despite the passing of another twelve months without a major volcanic disaster, research on
the impacts of volcanic eruptions continues to grow. Of particular note is a special volume of
the Journal of Volcanology and Geothermal Research dedicated to Explosive eruption risk
and decision support for European Union populations threatened by volcanoes (EXPLORIS).
This innovative multi-disciplinary project was funded by the European Union and ran from
2002 to 2006, and during this time it developed an analytical approach to volcanic risk
management that speciﬁes and quantiﬁes all the processes and effects of volcanic hazards
that inﬂuence risk. Of the 16 papers published, relating to Vesuvius (Italy), Teide (Tenerife,
Canary Islands), La Soufrière (Guadeloupe, French Antilles) and Sete Cidades (San Miguel,
Azores), three are highlighted here.

We begin with Vesuvius, as a future explosive eruption here, which affects the huge urban
population and infrastructure around the Bay of Naples (Figure 7), should be considered
amongst the most serious potential natural disasters in Europe. A. Neri

of the Italian INGV,

and colleagues, have formulated an Event Tree that provides a logical pathway connecting
generic probabilistic hazard assessment to quantitative risk evaluation at the volcano. The
Event Tree was created to provide a formal structure in which diverse strands of information
could be uniﬁed and linked together by using expert knowledge. The end product depicts
the various eruption categories, with their likely probability of occurrence and associated
uncertainty, as well as the generic hazards and timeline evolution associated with them. The
paper highlights the importance of expert elicitation and presents a ﬁrst example of how this
can be used to integrate numerical simulations and ﬁeld evidence to generate probabilistic
maps of the hazard posed by pyroclastic ﬂows. The Event Tree is organised to enable easy
updating as new information becomes available, and this will be particularly important for
enabling regular updating of risk assessments of buildings and infrastructure at different
locations around the volcano.

With reference to this last point, G. Zuccaro

of the University of Naples, and colleagues,

present a ﬁrst attempt to develop a probabilistic dynamic model that allows detailed
analysis of the vulnerability of built structures around Vesuvius, which is made possible and
meaningful because of the comprehensive inventory of ﬁeld data on buildings in the area.

Figure 7

Image of Mt. Vesuvius
(central Italy), acquired
on 26 September 2000 by
the Advanced Spaceborne
Thermal Emission and
Reﬂection Radiometer
(ASTER). The false-colour
image covers an area of 36
x 45 km, with most of the
developed land in Naples
and the surrounding area
showing up green.

Courtesy: NASA/GSFC/
MITI/ERSDAC/JAROS
and U.S./Japan ASTER
Science Team.

Hazard & Risk Science Review 2009

The model considers three hazardous events, namely volcanogenic earthquakes, tephra
falls and pyroclastic ﬂows, and allows for consideration of single events or any sequence of
these events. In the case of the latter, a timeline has been constructed to show the change
in building vulnerability that occurs at each phase of a reference eruption, starting with
volcanogenic earthquakes and loading of roofs with tephra prior to a pyroclastic ﬂow.
Results of the model are presented as maps showing either the extent of building damage
or the number of casualties.

The theme of building vulnerability is also the subject of the paper by J. Martí

of Spain’s

Institute of Earth Sciences Jaume Almera, Consejo Superior de Investigaciones Cientíﬁcas
in Barcelona, and colleagues. These authors focus on the Icod valley of Tenerife, already
considered earlier for its potential to generate tsunami, which is one of the areas most
at risk in any future explosive event from the huge Teide-Pico Viejo complex. The study
considers the same three hazards investigated by Zuccaro and colleagues, but emphasises
the importance of surveying building stock and modelling humanitarian losses. Marti and
colleagues point out that their work could easily be adapted to modelling economic losses
under various hazard scenarios, simply by additional consideration of building quality and
use in the survey. By assigning a construction cost per square metre it would be easy to
estimate a replacement value per building. The studies of both Martí and Zuccaro, and
their colleagues, could be extended to estimate economic losses associated with business
interruption brought about by building and infrastructure damage.

EXPLORIS highlights the importance of constraining the past and future potential behaviour
of volcanoes, but this still presents a huge challenge for volcanic hazard and risk estimation.
In an effort to improve this situation, A. Mendoza-Rosas

and S. De la Cruz-Reyna of the

Universidad Nacional Autónoma de México have published a statistical method in the Journal
of Volcanology and Geothermal Research, which links geological and historical eruption time
series in order to calculate probabilities of future explosive eruptions. Their ﬁrst step in the
method is to characterise eruptions by their magnitudes (volcanic explosivity index) and then
to test the eruptive time series for independence between successive events and for the time
dependence or stationarity of the process. Weibull analysis is then used to study the distribution
of repose periods between successive eruptions, and a non-homogeneous generalised
Pareto–Poisson process is used to obtain volcanic hazard estimations. When tested on
Mexican volcanoes, this method produces higher exceedance probabilities for larger magnitude
eruptions when compared to Poisson and Binomial distribution-based hazard estimates.

Towards an integration of sensor information

This section of the Review ends by considering how risk managers and geohazard analysts
access sensor information, such as high-resolution airborne or space imagery,
hyper-spectral data, borehole data, seismometer data, global positioning system networks,
and other geophysical in-situ and space instrumentation. This is the subject of the paper by
G. Le Cozannet

of the French Bureau de Recherches Géologiques et Minières (BRGM),

and colleagues, which is published in Sensors. They argue that information about sensor
networks is difﬁcult to obtain and is available in too many formats to facilitate effective
access. In order to promote decision-making based on greater access to information,
interoperable catalogues are being developed as part of the Group on Earth Observations
workplan. One example is GeoHazData, which is a prototype homogeneous inventory of
hazard maps and their underlying data for a speciﬁc region. Currently this tool is open to
critical review and time will tell if the approach is appropriate. Regardless, it does begin to
address the need for homogeneous and systemic approaches to the presentation of data
needed for hazard and risk analysis.

Hazard & Risk Science Review 2009

Sources and further reading

1. Akinci, A. F. et al. 2009 Effect of Time Dependence on Probabilistic Seismic-Hazard Maps
and Deaggregation for the Central Apennines, Italy. Bulletin of the Seismological Society of
America 99, 585–610. doi: 10.1785/0120080053.

2. Alcik, H. et al. 2009. A study on warning algorithms for Istanbul earthquake early warning
system. Geophysical Research Letters 36. doi:10.1029/2008GL036659.

3. Allen, R. M. et al. 2009. Real-time earthquake detection and hazard assessment by
ElarmS across California. Geophysical Research Letters 36. doi:10.1029/2008GL036766.

4. Atkinson, G. M. & Macias, M. 2009 Predicted Ground Motions for Great Interface
Earthquakes in the Cascadia Subduction Zone. Bulletin of the Seismological Society of
America 99, 1552–1578. doi: 10.1785/0120080147.

5. Chester, D. K. 2008 The effects of the 1755 Lisbon earthquake and tsunami on the
Algarve region, southern Portugal. Geography 93, 78-90.

6. Choy, G. L. & Boatwright, J. 2009 Differential Energy Radiation from Two Earthquakes
in Japan with Identical Mw: The Kyushu 1996 and Tottori 2000 Earthquakes. Bulletin of the
Seismological Society of America 99, 1815–1826. doi: 10.1785/0120080078.

7. Coppo, N. P. et al. 2009 A deep scar in the ﬂank of Tenerife (Canary Islands): geophysical
contribution to tsunami hazard assessment. Earth and Planetary Science Letters 282, 65-68.
doi:10.1016/j.epsl.2009.03.017.

8. Cotton, F. et al. 2008 On the Discrepancy of Recent European Ground-Motion
Observations and Predictions from Empirical Models: Analysis of KiK-net Accelerometric
Data and Point-Sources Stochastic Simulations. Bulletin of the Seismological Society of
America 98, 2244–2261. doi: 10.1785/0120060084.

9. Frankel, A., Stephenson, W. and Carver, D. 2009 Sedimentary Basin Effects in Seattle,
Washington: Ground-Motion Observations and 3D Simulations. Bulletin of the Seismological
Society of America 99, 1579–1611. doi: 10.1785/0120080203.

10. Furlong, K. P. et al. 2009 A great earthquake rupture across a rapidly evolving three-plate
boundary. Science 324, 226-229. doi: 10.1126/science.1167476.

11. Graves, R. W. et al. 2008 Broadband simulations for Mw 7.8 southern San Andreas
earthquakes: Ground motion sensitivity to rupture speed. Geophysical Research Letters 35.
doi:10.1029/2008GL035750.

12. Gutscher, M. -A. and Westbrook, G. K. 2009 Great earthquakes in slow-subduction,
low-taper margins. In: Lallemand S. and Funiciello F. (eds) Subduction Zone Geodynamics,
Springer-Verlag p119-133. doi: 10.1007/978-3-540-87974-9.

13. Kudo, T., Tanaka, T. and Furumoto, M. 2009 Estimation of the Maximum Earthquake
Magnitude from the Geothermal Gradient, Bulletin of the Seismological Society of America
99, 396–399. doi: 10.1785/0120080946.

14. Le Cozannet, G. et al. 2008 Connecting hazard analysts and risk managers to sensor
information. Sensors 8, 3932-3937. doi: 10.3390/s8063932.

15. Løvholt, F. et al. 2008 Oceanic propagation of a potential tsunami from the La Palma
Island. Journal of Geophysical Research 113, C09026, doi:10.1029/2007JC004603.

Hazard & Risk Science Review 2009

16. Marti, J. et al. 2008 Estimating building exposure and impact to volcanic hazards in Icod
de los Vinos, Tenerife (Canary Islands). Journal of Volcanology and Geothermal Research
178, 553-561. doi:10.1016/j.jvolgeores.2008.07.010.

17. Massa, P. et al. 2008 Empirical Ground-Motion Prediction Equations for Northern Italy
Using Weak- and Strong-Motion Amplitudes, Frequency Content, and Duration Parameters.
Bulletin of the Seismological Society of America 98,1319–1342. doi: 10.1785/0120070164.

18. Mendoza-Rosas, A. T. and De la Cruz-Reyna, S. 2008 A statistical method linking
geological and historical eruption time series for volcanic hazard estimations: applications to
active polygenetic volcanoes. Journal of Volcanology and Geothermal Research 176, 277-
290. doi:10.1016/j.jvolgeores.2008.04.005.

19. Neri, A. et al. 2008 Developing an Event Tree for probabilistic hazard and risk
assessment at Vesuvius. Journal of Volcanology and Geothermal Research 178, 397-415.
doi:10.1016/j.jvolgeores.2008.05.014.

20. Olsen, K. B. et al. 2008 TeraShake2: Spontaneous Rupture Simulations of Mw 7:7
Earthquakes on the Southern San Andreas Fault, Bulletin of the Seismological Society of
America 98, 1162–1185. doi: 10.1785/0120070148.

21. Parsons, T. and Geist, E. L. 2008 Tsunami probability in the Caribbean region. Pure and
Applied Geophysics 165, 2089-2116. doi: 10.1007/s00024-008-0416-7.

22. Parsons, T. and Geist, E. L. 2009 Is There a Basis for Preferring Characteristic
Earthquakes over a Gutenberg–Richter Distribution in Probabilistic Earthquake Forecasting?
Bulletin of the Seismological Society of America 99, 2012–2019. doi: 10.1785/0120080069.

23. Sørensen, M. B. et al. 2009 Attenuation of Macroseismic Intensity: A New Relation for the
Marmara Sea Region, Northwest Turkey, Bulletin of the Seismological Society of America 99,
538–553. doi: 10.1785/0120080299.

24. Stromeyer, D. & Grünthal, G. 2009 Attenuation Relationship of Macroseismic Intensities
in Central Europe, Bulletin of the Seismological Society of America 99, 554–565. doi:
10.1785/0120080011.

25. Waythomas, C. F. et al. 2009 Paciﬁc Basin tsunami hazards associated with mass ﬂows
in the Aleutian arc of Alaska. Quaternary Science Reviews 28, 1006-1019. doi:10.1016/j.
quascirev.2009.02.019.

26. Wood, N. 2009 Tsunami exposure estimation with land-cover data: Oregon and
the Cascadia subduction zone. Applied Geography 29, 158-170. doi:10.1016/j.
apgeog.2008.08.009.

27. Wyss, M. and Zibzibadze, M. 2009 Delay times of worldwide global earthquake alerts.
Natural Hazards. 50, 379-387, DOI - 10.1007/s11069-009-9344-9.

28. Zollo, A., et al. 2009 Earthquake early warning system in southern Italy: Methodologies
and performance evaluation. Geophysical Research Letters 36. doi:10.1029/2008GL036689.

29. Zuccaro, G. et al. 2008 Impact of explosive eruption scenarios at Vesuvius. Journal of
Volcanology and Geothermal Research 178, 416-453. doi:10.1016/j.jvolgeores.2008.01.005.

Hazard & Risk Science Review 2009

4. HYDROLOGICAL HAZARDS

Hydrological hazards over the past year have included severe snow-melt ﬂoods in the
mid-western USA and Canada in March and April 2009, ﬂash ﬂooding in the Czech Republic
and Poland in June, ﬂooding in central and southern China in July and earlier in the year
(January), ﬂoods in Queensland and the Northern Territory following tropical storm Dominic.
One of the worst natural disasters this year was also a result of the hydrological conditions
when bush ﬁres in southern Australia ensued during the continuation of a severe drought.

UK precipitation and ﬂood impacts

Following the 2007 summer ﬂoods, there has been a general concern about surface water
management in the United Kingdom. Governmental bodies such as the Environment Agency
and the Meteorological Ofﬁce have been working together to provide a surface ﬂood warning
from extreme rainfall. A paper in Weather by Aon Benﬁeld UCL Hazard Research Centre
associate H. Rodda

together with researchers at Hydro-GIS Ltd and Oxford University,

investigated extreme rainfall observations in the British Isles over the period 1866-1968.
Data for this study was taken from the annual publication British Rainfall, which included
listings of all daily rainfalls over a speciﬁc depth threshold together with detailed descriptions
of the events from observers, rainfall maps and photographs (Figure 8).

The research showed that extreme rainfalls of 100 mm or more in 24 hours are not
uncommon. In fact over the study period there were only 3 years (1866, 1870 and 1873)
where an annual maximum of over 100mm was not recorded and this was largely due to the
sparse network of gauges at this time. Another particularly useful aspect of the study was
that all the data was compiled into a digital archive with over 28,000 records. This archive,
even as far back as the year 1900, includes records from over 4,000 gauges. This ﬁgure far
exceeds the digital archives held by the Met Ofﬁce where only a few hundred rain gauges
have archived data going back over 100 years. This archive is now available to use for
further research by interested parties.

Another aspect of the UK 2007 ﬂoods which was not widely reported in the media was
the impact on agriculture in England. This topic is covered in a paper by H. Posthumus

and researchers from Cranﬁeld University, UK, in the Journal of Flood Risk Management.
Approximately 42,000 ha of agricultural land was ﬂooded mostly in north-eastern and central
southern England. This particular study looked at the damage experienced by farmers in
three areas namely Yorkshire and Humberside, Worcester and Gloucestershire,
and Oxfordshire. The ﬂood impacts were assessed at the farm and land unit scale and
ﬁnancial losses calculated based on the physical damage and the unit price for crops
although the damage was often indirect. The average loss per farm was calculated as nearly
£90,000 (£1207/ha) with the highest losses in the horticultural sector of nearly £7,000/ha.
The estimated total agricultural loss, given the area of farmland ﬂooded was over £50 million
and the study estimated a 40% decrease in the yield of cereals in ﬂood affected areas.

Figure 8

Numbers of extreme
24-hour rainfall
observations per year from
British Rainfall 1866-1968.

Courtesy: Harvey Rodda.

Hazard & Risk Science Review 2009

European ﬂooding: past and future

Going back a little further, the Central European ﬂoods of 2002 are still the subject of
research in the ﬁeld of ﬂood management. In a paper also published in the Journal of Flood
Risk Management, M. Socher

and G. Böhme-Korn, from the Saxon State Ministry for

Environment and Agriculture, report on the lessons learned in the German state of Saxony in
the years following these ﬂoods. About two-thirds of the state was affected in August 2002,
with 21 deaths, 25,000 buildings damaged and a total economic loss estimated at EUR6.2
bn. The state government’s review of the ﬂoods in 2003 initiated a revised ﬂood protection
strategy including a detailed assessment of the hydrology and ﬂooding potential of the rivers
in the region, proposing more than 1600 ﬂood protection measures, and the production of
ﬂood maps for the communities at risk. The main thrust of the strategy was the rebuilding of
ﬂood defences and other hydraulic structures as these were damaged in a total of 18,000
locations. Rebuilding was prioritised on a points-based scoring system using both cost-beneﬁt
and more qualitative values. Highest priority was given to the construction of new ﬂood
detention reservoirs, dyke improvements, protection of settlements and infrastructure with a
high qualitative value and measures with international and national cross border relevance.

The use of historical records to provide better estimates of design ﬂoods is explored in a
paper in the Journal of Hydrology by G. Calenda

and researchers from the Universita Degli

Studi Roma. The study on the Rome’s River Tiber makes use of river-level records going
back to the 15th Century, primarily to improve estimates of 200-500 year ﬂoods, which are
commonly based on less than 100 years of systematic ﬂow monitoring data. They advocate
making use of “non-systematic” information whereby ﬂoods are described in historical
accounts or recorded as physical marks on buildings and bridges etc. By combining historical
observations of water level with information on the physical properties of the river including
dates of construction of hydraulic structures and even channel surveys, historical ﬂow-level
rating curves were derived to calculate the ﬂow associated with observed historical levels.
This provided a much longer ﬂow time series (1422-1989) from which the design ﬂows could
be calculated.

Also focusing on Italy, a paper entitled: Sea level rise, hydrologic runoff and the ﬂooding
of Venice was published in Water Resources Research by A. Rinaldo

and a team of

researchers from Italy, Switzerland and the USA. The study was a response to proposed
hydraulic engineering works designed to allow the temporary closure of the lagoon
surrounding the city of Venice during storm-surge events in the Adriatic. The theme of the
research was to consider what changes in the lagoon water levels may occur when it is
closed to the open sea, largely from the input of freshwater from the surrounding terrestrial
drainage and also from small scale meteorological effects on the lagoon water itself. To
achieve this, detailed hydrological and hydrodynamic models were set up to predict the
water levels from the contributing 2000 km

drainage basin with 27 channels ﬂowing into the

lagoon. This input was coupled with a ﬁnite element (i.e. triangular interpolation network)
tidal model of the Venice Lagoon. The combined model was simulated using tidal and
meteorological information from 2000-2007 and assuming climate change projections of a
30 cm and 50 cm rise in sea level. Concerns that under the barrier closure Venice may ﬂood
from the river input into the lagoon were proven to be false, even under the 50cm sea level
rise scenario.

Heading north to Germany, a study of ﬂood trends in the country was undertaken by
T. Petrow

and M. Merz from the Geoforschungszentrum, Potsdam. Their paper in the

Journal of Hydrology used data from 145 gauging stations over the period 1951-2002 to
assess whether there are any signiﬁcant trends in ﬂooding in Germany. Nine previous
studies either worldwide or for speciﬁc countries have found wide ranging trends in the
frequency and magnitude of ﬂooding over time. Such changes are not just the result of
climate change – other environmental factors such as land use change and river engineering
have signiﬁcant effects on ﬂooding. Eight different ﬂood indicators were used based on the
annual and seasonal (summer and winter) maxima and a statistical test (the Mann-Kendall

Hazard & Risk Science Review 2009

test) was applied to the time series of each of these indicators. Overall, they found that the
ﬂood hazard in Germany increased over the study period, particularly in relation to ﬂood
frequency. However, clear differences were found when looking at the spatial patterns
and the different ﬂood indicators. Upward trends were evident in the rivers of southern
and western Germany whereas almost no upward trends were found in eastern German
rivers. Western rivers (the Rhine and Weser) were dominated by an increase in winter
ﬂoods whereas the rivers of the Danube basin in the south were dominated by an increase
in summer ﬂoods. There was no signiﬁcant difference in the trends for different catchment
areas. The seasonal and spatial coherence of the trends suggests that the changes are
climate driven. Changes based on river engineering would show evidence in larger rivers but
not for small rivers where few structures have been implemented.

Mapping ﬂood risk and impacts

Following the programmes of ﬂood risk mapping that are now established in many
countries, a number of studies have been made which investigate the accuracy of the
methods used for ﬂood risk mapping. J. D. Bales

and C. R. Wagner from the USGS

in North Carolina published a paper on the sources of uncertainty in ﬂood inundation
maps in the Journal of Flood Risk Management. Their study considered the widespread
river ﬂooding experienced in North Carolina following extreme rainfall associated with
Hurricane Floyd in 1999 (Figure 9). In particular they considered the sources of uncertainty
associated with producing maps from a Light Detection and Ranging (LiDAR) derived
Digital Terrain Model (DTM) and a 1-dimensional hydrodynamic model. Generating maps
using such techniques gives a good representation of the maximum ﬂood extent although
the study identiﬁed that a change of elevation of 15cm from the DTM can give up to a 150m
change in the horizontal position of the maximum ﬂood extent. Another weakness of the 1-d
hydrodynamic modelling is that the maximum ﬂood extent is associated with the peak ﬂow
in the river but due to the time lag of the ﬂow onto and back from the ﬂood plain and local
storage effects it was shown that different water levels can be associated with similar ﬂows
according to whether they occur on the rising and falling limbs of the ﬂood hydrograph. In
conclusion the authors warn that the creators and users of ﬂood inundation maps need to
be mindful of the uncertainties and sources of uncertainties associated with these maps.

In a paper published in the Journal of Flood Risk Management, W. Veerbeek

and

C. Zevenbergen, from Dura Vermeer Business Development in the Netherlands, consider
the issue of urban ﬂood damage. In ﬂood models, damage is often not given the same
level of attention as the calculation of the depth and extent of water, and in this paper the
authors consider the distribution of ﬂood damage in the Netherlands city of Dordrecht.
The city is at risk of ﬂooding both from the sea and a series of rivers and canals, but is
defended by dykes designed to withstand events up to the 1 in 4000 year ﬂood. Perhaps

Figure 9

Inland ﬂooding in North
Carolina associated with the
1999 landfall of Hurricane
Floyd. Inland ﬂooding is
responsible for more than
half the deaths associated
with tropical cyclones
striking the United States.

Courtesy: U.S. Army Corps
of Engineers, J. Jordan.

Hazard & Risk Science Review 2009

unsurprisingly, estimates of damage show a large increase when this design ﬂood level is
exceeded. Potential damage following a 1 in 4000 year ﬂood is considered in terms of its
spatial distribution, the age of buildings affected, the actual components of the damage, and
the effect of climate change. The results show damage is characterised by clusters of high
damage mostly in post-war neighbourhoods but also with the city’s historical building stock
being affected. The largest damage component is to household interiors and for a scenario
associated with the year 2100 ﬂood damage could increase by a factor of 4.

The results of a similar study by H. Apel

and others from the Geoforschungszentrum,

Potsdam (Germany) were reported in Natural Hazards. Titled Flood risk analysis - how
detailed do we need to be? This study considered the different levels of complexity in ﬂood
risk modelling from estimates of ﬂooded areas through to different methods of damage
calculations using information from the August 2002 ﬂood on the River Mulde in the city of
Eilenburg in Saxony. The prediction of the ﬂooded area was made using a linear interpolation
of observed ﬂood levels, a more complex 1-d/2-d hydrodynamic model, and in the most
detail using a fully 2-d hydrodynamic model. Likewise the estimation of damage was over
three levels of complexity: a simple damage function through to a meso-scale damage model
and at the most detailed a micro-scale damage model where the individual buildings are
taken into account. The performance of the different combinations of hazard and vulnerability
predictions was measured against the actual ﬂood extent and damage from the 2002 ﬂood
where water extents were available from satellite images and ﬂood depth and the cost
of damage were available for 380 buildings. Despite the varying levels of complexity for
the ﬂooded area, each technique produced almost identical ﬂood outlines although some
errors were apparent for the calculation of ﬂood depths at an individual building level. The
damage estimates ranged from just below EUR10 million to over EUR100 million with the
full combination of techniques compared to the ofﬁcial cost of EUR77 million. The closest
estimate was through a combination of the meso-scale damage model with the 1d/2d
hydrodynamic model, although a much greater variation was evident with the choice of
damage model as opposed to using different hazard models.

Another study, published in the International Journal of River Basin Management, also
addressed the level of detail required for evaluating ﬂood risk. Y. Huang

, of the Water

Resources Ministry in China, considered the use of different methods to predict ﬂooded
areas and damage following a dyke breach on the River Elbe, Germany. The rationale for
the study was to ﬁnd more rapid ways of assessing potential damage during a ﬂood event.
Flood forecasting methods traditionally predict a ﬂow and associated water levels, but
estimating the area ﬂooded and the depth of ﬂood water is not usually practical during the
short (2-3 hour) lead times available to the forecasters. This is because the complex 1- and
2-d hydrodynamic models can take in the order of many hours to run and require a large
amount of preparatory work. Instead the author describes a more rapid GIS based approach
to estimate the maximum depth and extent of ﬂooding following a dyke breach based largely
on the DTM and grid cell based hydrological functions available in Arc GIS. These results
are compared to those generated by the SOBEK1D2D model. Estimates of damage are also
made using CORINE land use data. Both models predict very similar ﬂood extents with a
slightly greater water depth apparent from the SOBEK1D2D estimates. This greater depths
means the damage estimates are also greater with the SOBEK1D2D model. Despite this
the GIS method is able to provide an initial rapid assessment (within minutes) of the ﬂooded
area which can aid the authorities in terms of emergency management such as evacuations
and strengthening ﬂood defences.

Hazard & Risk Science Review 2009

Advances in ﬂood modelling

A recent development in the ﬁeld of ﬂood modelling is the use of the so-called Copula
Method to identify relationships between measured ﬂows or water levels for river gauging
stations covering a wide area. C. Wang

and researchers at the University of Central Florida

have undertaken a study on Copula-based ﬂood frequency analysis which was published in
Hydrological Processes. They consider the ﬂood hazard at the conﬂuences of rivers where
building or structures in the vicinity can be at risk of ﬂooding from each river individually or
both rivers together. If ﬂows are high in one river the water levels in the adjoining river can
also be affected through backwater effects. The joint probability of such ﬂooding is reported
in this paper through using the copula method. The study used three river basins (the Des
Moines in Iowa; the Suwannee in Florida and the Calcasieu in Louisianna), each with two
or more upstream tributary gauges and a further gauge downstream of the conﬂuence.
The paper reports the theory behind the Copula method in detail and tests four different
types of the method against standard extreme value methods for deriving the design ﬂood
magnitudes. The authors advocate using the method for estimating ﬂoods at river basin
conﬂuences. Given that the methods do require a specialist understanding of complex
mathematics, however, their application may be limited.

Sources and further reading

30. Apel, H. et al. 2009 Flood risk analyses—how detailed do we need to be? Natural
Hazards 49, 79–98.

31. Bales, J.D. & Wagner, C.R. 2009 Sources of uncertainty in ﬂood inundation maps
Journal of Flood Risk Management 2, 139–147.

32. Calenda, G., Mancini, C.P. and Volpi, E. 2009 Selection of the probabilistic model of
extreme ﬂoods: The case of the River Tiber in Rome Journal of Hydrology 371, 1-11.

33. Huang, Y. 2009 Rapid ﬂood risk assessment using GIS technology. Journal of River
Basin Management 7, 1, 3-14.

34. Petrow, T. & Merz, M. 2009 Trends in ﬂood magnitude, frequency and seasonality in
Germany in the period 1951–2002. Journal of Hydrology 371, 129-141.

35. Posthumus, H. et al. 2009 Impacts of the summer 2007 Floods on agriculture in England.
Journal of Flood Risk Management 2, 1–8.

36. Rodda, H.J.E. et al. 2009 A digital archive of extreme rainfalls in the British Isles from
1866 to 1968 based on British Rainfall. Weather 64, 3, 71-75.

37. Rinaldo, A. et al. 2008 Sea level rise, hydrologic runoff, and the ﬂooding of Venice.
Water Resources Research 44. doi:10.1029/2008WR007195.

38. Socher, M. & Bohme-Korn, G. 2008 Central European Floods 2002: lessons learned in
Saxony. Journal of Flood Risk Management 1, 123–129.

39. Veerbeek, W. & Zevenbergen, C. 2009 Deconstructing urban ﬂood damages: increasing
the expressiveness of ﬂood damage models combining a high level of detail with a broad
attribute set. Journal of Flood Risk Management 2, 45–57.

40. Wang, C., Chang, N. B. and Yeh, G.T. 2009 Copula-based ﬂood frequency (COFF)
analysis at the conﬂuences of river systems. Hydrological Processes 23, 1471 – 1486.

Hazard & Risk Science Review 2009

5. ATMOSPHERIC HAZARDS

At the time of writing (late July, 2009) tropical cyclone activity across the world’s ocean basins
remains quiet (see the Seasonal Tropical Cyclone Activity Update page at:
http://www.coaps.fsu.edu/~maue/tropical/). This continues a downward trend in measures
of integrated cyclone energy which is reaching historic low-levels of tropical cyclone activity
on a global scale for the past 30 years. Despite North Atlantic hurricane activity being above
normal in the 2008 season, producing the long lived Tropical Storm Bertha and record four-
times landfalling Tropical Storm Fay, the Paciﬁc basin season produced far fewer and far
weaker storms than normal. This level of tropical cyclone activity is consistent with cooler sea
surface temperatures (SSTs) in the Paciﬁc for the past few years, associated with La Niña
conditions, and relatively favourable atmospheric conditions (weak vertical wind shear) aloft
for hurricane formation in the Atlantic basin. However, currently, a return to a positive ENSO
(El Niño-Southern Oscillation) index and warmer conditions in the Paciﬁc seems likely.

Tropical cyclone variability

In a paper published in Geophysical Research Letters, R. Maue

, of the Center for Ocean-

Atmospheric Prediction Studies at Florida State University, draws attention to this remarkable
low-level of tropical cyclone activity in the Northern Hemisphere (NH) in recent years and
the large observed variability of the past three decades, as measured by the Accumulated
Cyclone Energy (ACE) index (Figure 10).

Here he uses ACE as a metric of the combined total activity of each ocean basin during
a calendar year, integrating frequency, duration and intensity from historical datasets. He
ﬁnds the strong inter-annual variability of ACE is highly correlated with Paciﬁc SST structure
and variability, and that North Paciﬁc boreal spring SST patterns may have considerable
predictive value for subsequent NH tropical cyclone activity. Such a ﬁnding underlines how
strongly tropical cyclone activity is modulated by the large scale modes of global climate
variability, but the relevance to the controversy over the inﬂuence of Anthropogenic Global
Warming (AGW) may be contested: the issue centers on his deﬁnition of ACE as an estimate
of the actual kinetic energy of tropical storms.

Figure 10

24-month running sum
of tropical cyclone
Accumulated Cyclone
Energy, as deﬁned by Maue
(2009), for the entire globe
(top black squares/time
series) and the Northern
Hemisphere only (bottom
green squares/time series).
The difference between
the two time series is the
Southern Hemisphere
(SH) total. Data are shown
from January 1979 to July
19, 2009 mainly because
intensity estimates of SH
cyclones are often missing
in the Joint Typhoon
Warning Centre best-tracks
prior to 1980.

Courtesy: Ryan Maue,
Florida State University.

Hazard & Risk Science Review 2009

On the subject of ENSO inﬂuence on NH tropical cyclone activity H-M. Kim

and colleagues,

of the Georgia Institute of Technology, in a landmark paper published in Science, identify
two distinct structures, hitherto unrecognised, in tropical Paciﬁc Ocean warming and their
different impacts on NH tropical cyclone activity (Figure 11). Episodes of Eastern Paciﬁc
Warming (EPW) are identical to El Niño warming, whereas Central Paciﬁc Warming (CPW)
is centred mid-basin, on the international dateline, and associated with increased potential
for landfall of storms along the Gulf of Mexico and Central America. The authors show
that modulation of the vertical wind shear forced by differential Paciﬁc teleconnections is
responsible and the CPW mode is more predictable than EPW thus producing increased
predictability of cyclones earlier in the season. An increase in the prevalence of CPW events
in recent decades is also detected (within the limitations of the available datasets), but the
reason for this increase is unclear.

Writing in the Journal of Climate and looking at the longer, historical time scale G. Vecchi

and T. Knutson, of the Geophysical Fluid Dynamics Laboratory, Princeton, examine
the relationships between North Atlantic (NA) tropical cyclone activity and SSTs. Vecchi
and Knutson attempt to produce a historical metric for tropical cyclone activity for the
period 1878-2006 by estimating the number of storms missed in the era before satellite
observations. They ﬁnd a highly signiﬁcant increase (+~4.2 storms century

-1

) in storm

numbers for 1900-2006. However, the estimate for storm frequency for 1878-2006 is weakly
positive, not statistically signiﬁcant and dependant on highly uncertain numbers for the late
19

century. An unexpected ﬁnding is a highly signiﬁcant decrease in storm duration for

1878-2006, but this is subject to greater uncertainties. The relationship of these ﬁndings to
increased SSTs in the development region of North Atlantic cyclones and other factors at
work remain unclear.

In a note, also in Journal of Climate, P. Klotzbach

and W. Gray, of Colorado State

University, also report results of a historical analysis back to 1978. They ﬁnd a ‘remarkable
agreement’ between NA SST, sea level pressure (SLP) anomalies and multi-decadal
variability in both Atlantic storm activity and the frequency of landfall in the United States.
The frequency of landfalling storms along the US East Coast and Florida peninsula is
correlated with phases of the Atlantic Multidecadal Oscillation (AMO) in SSTs.

Figure 11

Composites of sea
surface temperature
(SST) anomalies
(contours interval is
0.5°C) according to Kim
et al. (2009) during the
August to October period
for (A) Eastern Paciﬁc
Warming (EPW), (B)
Central Paciﬁc Warming
(CPW), and (C) Eastern
Paciﬁc Cooling (EPC). The
average number of North
Atlantic tropical cyclones
per month from June to
November is shown in
(D) for climatology (grey
bar), EPW (red), CPW
(green), and EPC (blue).
The time series has been
de-trended to eliminate
the effects of decadal
variability or climate trends.

Courtesy: Science.

Hazard & Risk Science Review 2009

Further to the last point P. Dailey

and co-workers from AIR Worldwide Corporation, Boston,

Massachusetts, writing in the Journal of Applied Meteorology and Climatology, focus on
where and how SSTs inﬂuence the location of US landfall of storms. The results of their two
part, statistical and physical, study indicate East Coast land-falling storms have a different
genesis region and intensiﬁcation characteristics compared to Gulf Coast land-falling storms
(Figure 12). Warming SSTs and shifting genesis regions and intensiﬁcation patterns may
strongly inﬂuence regional landfall risk, with profound implications for the distribution of the
risk to life and property.

The above papers focus their analyses on the Northern Hemisphere tropical cyclone
season and we now turn to a paper on Southern Hemisphere storms, in which similar
themes are apparent. H. Ramsey

, and co-workers from the University of Oklahoma,

report in the Journal of Climate the results of a study of various large-scale environmental
factors on the inter-annual variability of tropical cyclones in the Australasian region.
Ramsey and colleagues provide a brief but useful review of the previous literature
on tropical cyclone climatology for Australasia and then go on to analyse a 1970-
2006 “best-track” dataset for the satellite observation era. They ﬁnd large correlations
between Niño-3.4 and Niño-4 region SSTs in the Paciﬁc and seasonal numbers of
tropical cyclones. The correlations are greatest during August to October, immediately
preceding the Australasian tropical cyclone season, and are potentially useful predictors
as early as July. Correlations with local SSTs in the region where these cyclones are
born north of Australia are found to be much weaker. However, low-level vorticity and
tropospheric vertical wind shear in this genesis region are found to be strongly inﬂuenced
by central Paciﬁc SSTs via an ‘atmospheric bridge’- an idea for which the authors provide
a conceptual model (see Figure 13). Intriguingly, correlations between SSTs and cyclone
activity are increased by combining tropical Paciﬁc SST and sub-tropical North Atlantic
SSTs even several months prior to the cyclone season, hinting at teleconnections from
far aﬁeld. Finally, the authors ﬁnd an apparent decline in annual numbers of all tropical
cyclones since 1970, but not for storms with central SLPs of below 965 hPa.

Figure 12

Climatological mean
tracks for tropical cyclones
originating in two select
genesis regions according
to Dailey et al. (2009). The
thick black line indicates
the mean trajectory of
storms originating from the
(top) western Atlantic and
(bottom) eastern Atlantic
regions discussed in the
text. The thin black (grey)
line is based on historical
tracks originating from each
genesis region in warm
(cool) years.

Hazard & Risk Science Review 2009

Compared to the North Atlantic, North West Paciﬁc and Australasian regions, studies of
East Asian tropical cyclone climatology have received less attention. The two following
studies seek to address this imbalance. Z. Zuki

, and A. Lupo, University of Missouri,

Columbia, in the Journal of Geophysical Research (Atmospheres) study tropical storm
activity impacting Malaysia for the period 1960-2006. Motivated by two particularly costly
and deadly storms- ‘Greg’ and ‘Vamei’, that struck in December 1996 and December
2001- they show that indeed November and December are the most active months in this
region. Their dataset from the Joint Typhoon Warning Center (JWTC) shows a slight but
statistically insigniﬁcant increasing trend in activity for the period 1960-2006. Study of the
inter-annual variability shows there was more cyclone activity in La Niña years compared
to El Niño years. La Niña years were associated with warmer SSTs, relatively weak
tropospheric vertical wind shear, greater moisture, and stronger low-level cyclonic vorticity
in the study region. Regardless of whether SSTs were warmer or cooler non-active years
tended to have more low-level anti-cyclonicity, even weaker wind shear and less moisture
and these were mostly El Niño years. Longer term variability, such as linkage to the Paciﬁc
Decadal Oscillation (PDO), was not found.

J. Chan

and M. Xu, in the International Journal of Climatology, perform one the ﬁrst

comprehensive studies of landfalling tropical cyclones in the East Asian region, i.e. China,
Vietnam, the Philippines, the Korean peninsula and Japan. Their dataset from the JTWC
covers 1945-2004 and again they ﬁnd no signiﬁcant overall trend in cyclone numbers using
a wavelet analysis. They do ﬁnd signiﬁcant inter-annual and decadal variability which is well
correlated with the total number of cyclones in the North West Paciﬁc region on a
multi-decadal time scale. Detailed analysis is left to a second paper, not yet published.

Tropical Cyclones: hazards, prediction and warning

The authors of many of the above papers are only too aware of the limitations of their
historical datasets. The record available for risk assessment in, for instance, the North
Atlantic basin only goes back ~150 years whereas reinsurers need to consider far longer
return periods in their calculations. In a paper in Natural Hazards J. Rumpf

and

V. Schmidt, of the Institute of Stochastics at Ulm University, and co-workers from Munich Re
attempt a different approach to assess the probability and intensity of landfalling storms in
the North Atlantic. Using a basin-wide Monte-Carlo simulation they create a synthetic dataset
of storm tracks which they claim is more comprehensive than the available historical data
and, they argue, provides a better basis for assessing landfall probability, especially in areas
of low landfall frequency. In this approach the ﬂuctuations of inter-annual, decadal and multi-
decadal variability are neglected in order to capture the statistics of the historical dataset
as a whole, in this case for the period 1900-2005, and enable the calculation of the return
periods of a large number of ‘wind impacts’ at any one location affected by tropical cyclones.

Figure 13

Schematic from Ramsey
et al. (2008) showing
the connection between
anomalously warm Sea
Surface Temperature
in the Niño-3.4 and
Niño-4 regions (pink
shading) associated with
El Niño events and the
corresponding atmospheric
response: increased 200-
hPa zonal westerly winds
(blue horizontal arrows)
around 15°S resulting in
increased vertical shear of
the zonal wind, co-located
with decreased 850-hPa
relative cyclonic vorticity
(green) associated with
a weakened monsoon
trough over the Australian
region. Light blue vertical
arrows indicate anomalous
subsidence over the
northern Australian region
and anomalous ascent
over the central
equatorial Paciﬁc.

Hazard & Risk Science Review 2009

The concern underlying all these recent studies is that human-induced climate change is
inﬂuencing the course of weather and climate variability and related trends in insurance
losses. R. Crompton

and K.J. McAneney, of the Risk Frontiers Natural Hazards Research

Centre at Macquarie University, Sydney address this point directly in a paper in the journal
Environmental Science & Policy. Using the Natural Disaster Event List from The Insurance
Council of Australia they normalise the weather related losses to estimate the insured loss
that would be incurred if these events were to occur under 2006 societal conditions. For
instance, they include a factor adjusting for the inﬂuence of improved building standards in
cyclone prone areas since the early 1980s- this is a marked advance on previous studies. In
summary they suggest their evidence shows that societal factors- for instance the increased
number and value of dwellings- are the reason for increased insurance losses in Australia
and ﬁnd no impact of human-induced climate change during the time period of their study.
They ﬁnd improved building standards have effectively reduced risk and cyclone related
losses in recent years and conclude, on an optimistic note, that “Employing both mitigation
and adaptation contemporaneously will beneﬁt society now and into the future”.

An important aspect of hazard mitigation is accurate and timely forecasting of tropical
cyclones. The leading centre for tropical cyclone prediction in the Northern Hemisphere is the
National Hurricane Center (NHC), based in Miami, Florida. In a paper published in Weather
and Forecasting E. Rappaport

and colleagues from the NHC comprehensively review the

current status of the United States’ hurricane warning program highlighting recent progress,
current limitations and the challenges ahead- all this in the light of the recent extremely active
hurricane seasons of 2004 and 2005. Looking forward to the next 10 years Rappaport and
co-authors identify key areas for progress, namely 1) improvements in global and regional
forecast systems to reduce errors in storm track and intensity guidance; 2) optimisation and
enhancement of observing capabilities not only for operational, forecasting purposes but also
for advancing research; 3) expanding and enhancing the forecast tools required to apply and
add value to the guidance from the numerical models and the observing systems.

Multi-season forecasting is addressed by J. Elsner

of The Florida State University, and

co-authors, in a paper in the Journal of Climate. They have developed an improved statistical
algorithm to predict North Atlantic hurricane activity out to 5 yrs. The results indicate that
forecast skill over the period 1997-2005 is improved over persistence and previous multi-
season forecasts. A major caveat is that the technique does not, at present, account for
ﬂuctuations such as ENSO.

Finally, in this section, we conclude with a paper which tackles a grey area in our
fundamental understanding of hurricanes and the extreme winds and damage they produce-
the Hurricane Boundary Layer (HBL). P. Zhu

, of the Florida International University, Miami,

writing in the Journal of Geophysical Research (Atmospheres) proposes a conceptual model
for the structure of the updrafts and downdrafts that bring the hurricane winds to the surface.
Operating the National Centers for Atmospheric Research (NCEP) Weather Research
and Forecasting (WRF) model in Large Eddy Simulation (LES) mode he investigates the
structure of Hurricane Ivan’s boundary layer on landfall along the Gulf Coast in 2004. The
use of extremely high resolution, real data, so-called ‘large eddy resolving’ simulations (grid
point spacing of order ~100m) is a new area of research which will test our understanding of
how hurricanes interact with the changing surface roughness and heat and moisture ﬂuxes
on making landfall, increase our understanding of how localised wind damage is produced
and will lead to better parameterizations of the HBL in numerical models thus improving the
prediction of storm intensity.

Extratropical Storms

A major windstorm, ‘Klaus’, struck Western Europe during the winter of 2009 (see Aon
Benﬁeld 2009

). On the night of 23 January and into the 24 January sustained winds up

to 170 kph and gusts over 200 kph were recorded across southern France and northern
Spain leaving at least 27 dead, millions of homes without electricity and extensive damage to

Hazard & Risk Science Review 2009

property and infrastructure. Estimated losses range up to EUR2.5 billion. A few weeks later
a second strong, but less damaging storm, ‘Quinten’, struck north-western France. ‘Klaus’
is likely to be the most damaging windstorm to strike Western Europe since ‘Lothar’ and
‘Martin’ in 1999.

Windstorms in Europe in recent decades have been extremely costly to insurers and two
papers in a special issue of the International Journal of Climatology (Alexander, L.V.

et al.

Eds.), 2009) devoted to “Climate extremes: progress and future directions” will be of interest
to the industry. R. Allan

of the Hadley Centre, Exeter (UK), and co-authors, examine

extreme storms across the British Isles during the autumn and winter months from 1920
to 2004. They ﬁnd a signiﬁcant correlation of storminess in the autumn months (October,
November, December) with the Azores-Iceland North Atlantic Oscillation (NAO) index, but not
with the Gibraltar-SW Iceland NAO index. A weaker linkage to ENSO is detected. In the winter
months (January, February, March) signiﬁcant correlation with both NAO indices is found but
this ﬂuctuates over the period of the dataset. Interestingly, and importantly, extreme storms
in the autumn and winter months appear to be linked to different physical mechanisms. Such
a study helps put recent western European storms in perspective and emphasises the role
played by inter-decadal to multi-decadal ﬂuctuations in natural climate variability.

In another study of historical storminess in the European region, E. Hanna

of the University

of Shefﬁeld, and colleagues, examine trends in surface pressure back to 1830, as a proxy for
Atlantic and Northwest European storminess and climate variation over the period. Writing
in the Journal of Climate, they use these records to demonstrate the existence of periods of
increased storminess around 1900 and during the early 1990s, with a relatively quiet period
between the 1930s and 1960s. They also note that there is little evidence that the mid- to
late-nineteenth century was any less stormy than at present, and observe that there is no
sign – as yet – of a sustained enhanced storminess signal associated with global warming.

Of direct interest and importance to the reinsurance industry is the paper by P.M. Della-
Marta

and colleagues from MeteoSwiss, Zurich. Using the European Centre for Medium

Range Forecasting (ECMWF) ERA-40 reanalysis they estimate, using Extreme Value
Analysis techniques, the return periods of a catalogue of 200 of the most prominent
European windstorms. This seems to be a signiﬁcant step forward; however the authors
conclude with remarks on the challenges of further developing the statistical methodology
and overcoming the inherent limitations of current historical datasets.

Severe local storms- tornado formation

The ﬁrst part of a major ﬁeld program, ‘Vortex-2’, designed to investigate severe local
storms and tornado formation has taken place this spring in the Plains of the American
Mid-West. However, 2009 was a quiet tornado season after two very active years, including
the devastating Greensburg, Kansas tornado in May, 2007. The researchers may collect
further data during the second leg of Vortex-2 in spring 2010.

A key aim of ﬁeld experiments like Vortex-2 is to obtain data which will throw light on why
some, but by no means all, rotating thunderstorms – ‘supercells’ – produce tornadoes and
how the near surface storm environment inﬂuences tornadogenesis. Finding a canonical
theory for tornado formation has proved elusive but two papers published in Geophysical
Research Letters may well prove to be important pieces of the jigsaw puzzle. Both focus
on the inﬂuence of cloud microphysics on the rear ﬂank downdraft (RFD) and cold pool
generation in supercell storms. RFDs in supercell storms have long been thought to be
crucial to the evolution of low-level storm rotation. N. Snook

and M. Xue, of the Center for

Analysis and Prediction of Storms at the University of Oklahoma, use idealized supercell
simulations to investigate inﬂuences on cold pool (the rain cooled storm downdraft) intensity
and the low-level storm dynamics. Tornadic circulations occur in their simulations only when
the cold pool strength is relatively weak and just so in order to provide the dynamic lift and
vertical stretching of low-level air parcels that support tornadogenesis.

Hazard & Risk Science Review 2009

Reporting the results of a complementary study in Geophysical Research Letters,
D. Lerach

of Colorado State University, and co-researchers, also describe the use of

idealised simulations to show that storm environments polluted by aerosols produced
stronger RFDs, weaker cold pools and tornadoes. Clean environments were less favourable
for tornadogenesis. Further study of these processes is likely to be of importance in
improving the cloud microphysical parameterisations in the next generation of storm scale
forecast models.

Sources and further reading

41. Alexander, L.V. et al. (Eds.) 2009 Special Issue: Climate Extremes: progress and future
directions. International Journal of Climatology 29, 317 – 459.

42. Allan, R., Tett, S. & Alexander, L.V. 2009 Fluctuations in autumn-winter severe storms
over the British Isles: 1920 to present. International Journal of Climatology 29, 357-371.

43. AON Benﬁeld 2009 Event Recap Report: Windstorm Klaus – Jan. 23-24, 2009. Available
at: http://www.aon.com/about-aon/intellectual-capital/attachments/reinsurance/Event_
Recap_Report_Windstorm_Klaus.pdf.

44. Chan, J. C. L. & Xu, M. 2009 Inter-annual and inter-decadal variations of landfalling
tropical cyclones in East Asia. Part I: time series analysis. International Journal of
Climatology 29, 285-1293. doi: 10.1002/joc.1782.

45. Crompton, R. P. & McAneney, K. J. 2008 Normalised Australian insured losses from
meteorological hazards: 1967-2006. Environmental Science and Policy 11, 371-378.

46. Dailey, P. S. et al. 2009 On the Relationship between North Atlantic Sea Surface
Temperatures and U.S. Hurricane Landfall Risk. Journal of Applied Meteorology and
Climatology 48, 111-129.

47. Della-Marta, P.M. et al. 2009 The return period of wind storms over Europe. International
Journal of Climatology 29, 437-459.

48. Elsner, J. B. et al. 2008 Improving multiseason forecasts of north Atlantic hurricane
activity. Journal Of Climate 21, 1209—1219.

49. Hanna, E. et al. 2008 New insights into North European and North Atlantic surface
pressure variability, storminess and related climatic change since 1830. Journal of Climate
21, 6739 – 6766.

50. Kim, H-M, Webster, P.J. and Curry, J. A. 2009 Impact of Shifting Patterns of Paciﬁc
Ocean Warming on North Atlantic Tropical Cyclones. Science 325, 77-80.

51. Klotzbach, P. J. and Gray, W. M. 2008 Multidecadal Variability in North Atlantic Tropical
Cyclone Activity. Journal of Climate 21, 3929-3935.

52. Lerach, D.G., Gaudet, B.J and Cotton, W.R. 2008 Idealized simulations of aerosol
inﬂuences on tornadogenesis. Geophysical Research Letters 35. doi:10.1029/2008GL035617.

53. Maue, R. N. 2009 Northern Hemisphere tropical cyclone activity. Geophysical Research
Letters 36. doi:10.1029/2008GL035946.

54. Ramsay, H. A. et al. 2008 Interannual variability of tropical cyclones in the Australian
region: Role of large-scale environment. Journal Of Climate 21, 1083-1103.

Hazard & Risk Science Review 2009

55. Rappaport, E. N. et al. 2009 Advances and Challenges at the National Hurricane Center.
Weather and Forecasting 24, 395-419.

56. Rumpf, J. et al. 2008 Tropical Cyclone Hazard Assessment Using Model-based Track
Simulation. Natural Hazards 48, 383-398.

57. Snook, N. and Xue, M. 2008 Effects of microphysical drop size distribution on
tornadogenesis in supercell thunderstorms. Geophysical Research Letters 35.
doi:10.1029/2008GL035866.

58. Vecchi, G. A. & Knutson, T. R. (2008), ‘On estimates of historical north Atlantic tropical
cyclone activity’, Journal Of Climate 21, 3580--3600.

59. Zhu, P. 2008 Simulation and parameterization of the turbulent transport in the hurricane
boundary layer by large eddies. Journal of Geophysical Research-Atmospheres 113.
doi:10.1029/2007JD009643.

60. Zuki, Z. M. and Lupo, A. R. 2008 Interannual variability of tropical cyclone activity
in the southern South China Sea. Journal of Geophysical Research-Atmospheres 113.
doi:10.1029/2007JD009218.

Hazard & Risk Science Review 2009

6. CLIMATE CHANGE

Over the past three years the climate change issue has risen to the very top of the political
agenda, and is now widely recognised as the greatest threat to society at large and to a
fully-functioning and sustainable global economy. Both the US and UK have published new,
detailed, studies addressing climate change impacts on the respective nations (see at:
http://www.globalchange.gov and http://www.ukcip.org.uk), while the United States
Geological Survey has issued a report on abrupt climate change and its potential impacts
that certainly helps to concentrate the mind (see at: http://www.usgs.gov/global_change/).
Ahead of the critical COP15 UN Climate Change Conference, to be convened in
Copenhagen in December to address the issue of a post-Kyoto greenhouse gas emissions
agreement, the Climate Science Congress, held in the same city in March, produced a
particularly sobering synthesis report on future prospects (see at: http://climatecongress.
ku.dk/). Notwithstanding, however, the enormous research effort focusing on our changing
climate and its implications, there is still a great deal that we cannot be certain of, or are
unable to quantify. One area that still triggers high emotions within certain elements of the
climate science community relates to how, or even whether, a warmer world will affect the
number and intensity of tropical cyclones, and this contentious area is addressed in the
following section.

Climate change and tropical cyclone activity

Despite very quiet Atlantic hurricane seasons over the last two years, the debate continues
about whether episodes of recent elevated hurricane activity in the North Atlantic are
driven by climate change or are simply a reﬂection of a natural cycle. Advocates of both
sides of the argument present their results in a series of papers published in a new book
on Hurricanes and Climate Change, edited by J. Elsner and T. Jagger

of Florida State

University. The papers are the result of an international summit on hurricanes and climate
change that attracted more than 70 scientists to the Greek island of Crete in 2007 to debate
the issue. Elsner

, an advocate of a role for climate change in recent tropical cyclone

activity, addresses the often turbulent discussions at the summit in a paper in the Bulletin of
the American Meteorological Society. Here he succinctly gets to the heart of the problem:
studies have shown that tropical cyclones are becoming more powerful, most notably in the
North Atlantic, with the increase correlated with a rise in sea-surface temperature (SST). Is
this temperature rise, however, a consequence of anthropogenic climate change resulting in
increases in radiative forcing due to accumulating greenhouse gases in the atmosphere, or
to natural climate variations? While a warmer ocean might be expected to spawn and nurture
tropical cyclones more easily, increasing wind shear in a warmer world would be expected
to breaking up storms more effectively. Critically, Elsner observes that although the question
of whether changes in tropical cyclone intensity can be attributed to anthropogenic climate
change remains open, based upon data models of extreme winds, the difference in hurricane
intensity for storms near the US coast between globally warm and cool years is consistent in
both sign and magnitude with theory and simulations. The corollary of this is that apparent
discrepancies between numerical model results and observations probably reﬂect a reliance
on data analysis rather than data models. Elsner also reports the results of storm deposit
studies presented at the summit that suggest there were fewer Atlantic hurricanes when the
northern hemisphere was cooler during the so-called Little Ice Age (15th – 19th centuries).
This he suggests, however, could reasonably be explained – at a speciﬁc location - by a
change in hurricane tracks rather than abundance. Finally, and perhaps most signiﬁcantly,
Elsner notes that results presented from a number of high-resolution models were consistent
with the occurrence of stronger tropical cyclones in a warmer world, while storm numbers
were predicted to fall overall.

In an important and insightful paper in the journal Science, G. Vecchi

of the US National

Oceanic and Atmospheric Administration, and co-authors, zero-in on future prospects for
Atlantic hurricanes in a carbon-enriched world, noting that alternative interpretations of the

Hazard & Risk Science Review 2009

relationship between SST and hurricane activity result in very different pictures. On the one
hand, a simple link between absolute SST and Atlantic hurricane activity, would result in
a situation, by 2100, in which even a quiet year would be comparable to 2005, when four
hurricanes made US landfall resulting in losses in excess of USD100 billion (Figure 14 top).
On the other, if Atlantic hurricane activity is modulated by the Sea Surface Temperature in
the tropical Atlantic main development region (where most hurricanes grow) relative to the
tropical mean SST, the picture would be far less dramatic. Between 1946 and 2007, this
‘relative SST’ is as well correlated with Atlantic hurricane activity as the absolute SST, but
looking ahead relative SST is not projected to experience a signiﬁcant upward trend as the
tropics as a whole warm (Figure 14 bottom). A future wherein Atlantic hurricane activity is
controlled by relative SST would, therefore, be comparable to the recent past, with periods
of more and less intense activity relative to present-day conditions – due to natural climate
variability – but with little or no longer-term trend.

Sea-surface temperatures in tropical cyclone spawning grounds are also addressed by
N. P. Gillett

of the University of East Anglia’s Climatic Research Unit, and co-researchers.

In a paper in Geophysical Research Letters, Gillett and colleagues provide evidence from
modelling in support of a clear anthropogenic inﬂuence on ocean warming in both the
North Atlantic and Western North Paciﬁc cyclogenesis regions. The authors conclude that
greenhouse gas increases due to human activities are probably the main cause of warming
in these regions. In the case of the Atlantic, they note that their ﬁndings disagree with the
idea that warming in the Atlantic Cyclogenesis Region is driven by natural cycles such as the
Atlantic Multi-decadal Oscillation. Writing in Theoretical & Applied Climatology,
P. A. Steenhof

and W. A. Gough of the University of Toronto also examine Atlantic SST,

this time in the context of the resurgence of hurricane activity that started in 1995.

Figure 14

Past and extrapolated
changes in Atlantic
hurricane activity. Observed
hurricane power dissipation
index (PDI) anomalies are
regressed onto observed
absolute and relative Sea
Surface Temperature
over the period from
1946 to 2007, and these
regression models are
used to build estimates
of PDI from output of
global climate models
for historical and future
conditions. Anomalies are
shown relative to the 1981
to 2000 average (2.13 ×
10

-2

). The green bar

denotes the approximate
range of PDI anomaly
predicted by the statistical/
dynamical calculations.
The other green symbols
denote the approximate
values suggested by
high-resolution dynamical
models. Sea Surface
Temperature indices are
computed over the region
70°W-20°W, 7.5°N-22.5°N,
and the zero-line indicates
the average over the period
from 1981 to 2000.

Courtesy: Science.

Hazard & Risk Science Review 2009

The authors use a range of different measures of hurricane activity to examine the inﬂuence
of SST in the main development region of Atlantic hurricanes. In one approach, they take
the 10 warmest years and the 10 coolest years between 1941 and 2006, and identify a
signiﬁcant statistical relationship between SST and hurricane activity. In a second element
of the study the authors identify a signiﬁcant correlation between SST and all measures
of hurricane activity (Figure 15). For reasons addressed in the aforementioned paper by
Vecchi and colleagues, however, Steenhof and Gough are ambivalent about whether or not
the established link between recent Atlantic SST and hurricane activity will translate into
increased hurricane activity in a warmer world.

Staying with the inﬂuence of sea-surface temperature, K. Arpe

of Germany’s Max

Planck Institute for Meteorology and S. Leroy of Brunel University in the UK, examine, in
Quaternary International, the impact of local SST on Atlantic hurricanes. They also address
the potential effects of other phenomena affecting hurricane development in the North
Atlantic, including ENSO (El Niño – Southern Oscillation) and the stratospheric QBO (Quasi-
Biennial Oscillation). Arpe and Leroy note that elevated local SST, increases in the latent
heat ﬂux from the oceans to the atmosphere, a westerly airstream in the tropical stratosphere
(which reduced the occurrence of strong easterly phases of the QBO), and a more
moist, unstable stratiﬁcation of the atmosphere, all lead to more frequent or more intense
hurricanes. In contrast, increased vertical wind shear or a drier atmosphere will lead to
fewer North Atlantic hurricanes. Arpe and Leroy pick out wind shear as a critical determinant
that they conclude has a dominant inﬂuence, even over tropical SST. Looking ahead, they
speculate that increased wind shear in a warmer world could counteract the tendency for
higher SSTs to trigger increased Atlantic hurricane activity.

Looking more broadly at tropical cyclone activity across the globe, S. Gualdi

of the Istituto

Nazionale di Geoﬁsica e Vulcanologia in Bologna, Italy, and co-workers, use the results of
a study using a high-resolution, coupled-circulation model, to predict how global warming
will affect tropical cyclone activity in the major ocean basins. Writing in the Journal of
Climate, Gualdi and colleagues demonstrate that the study predicts a substantial general
reduction in tropical cyclone frequency as the concentration of atmospheric carbon dioxide
is doubled and quadrupled, particularly for the North Atlantic and tropical western North
Paciﬁc regions. In the former, hurricane activity appears to be suppressed by a more stable
atmosphere and by – as suggested in the previous paper – stronger vertical wind shear.

Figure 15

Steenhof and Gough (2008)
demonstrate a statistically-
signiﬁcant relationship
between Sea Surface
Temperature in the Main
Development Region and
hurricane activity in the
Atlantic Basin. This positive
correlation is illustrated by
a change in total hurricane
activity in the Atlantic Basin
in association with changing
SSTs in the MDR.

Courtesy: Theoretical &
Applied Climatology.

Hazard & Risk Science Review 2009

The study also predicts that tropical cyclones will generate more rainfall and – despite the
expansion pole-ward of warmer SSTs – will remain conﬁned to the tropics.

Whether or not tropical cyclones will eventually expand their traditional ‘hunting grounds’
– bearing in mind that Atlantic hurricanes have struck both Brazil and Spain in recent
seasons (Figure 16) – there may already be evidence that the Atlantic hurricane season is
getting longer. Writing in Geophysical Research Letters, J. P. Kossin

, of the University of

Wisconsin, looks at the timing of Atlantic hurricanes in recent years. He notes that in 2007,
storms occurred in both May and December, while the following year, four tropical storms
– two of hurricane status – occurred in July. Similarly, 2005 saw anomalous activity both
early and late in the season. Examining observed historical trends in the annual distribution
of North Atlantic tropical storm formation events, and their relationship with tropical
Atlantic SST, Kossin detects an apparent tendency towards more common early and late
season storms that correlates with higher SST. Currently, however, the uncertainties in the
correlation are high and the author makes the point that although storm formation dates have
been changing over the historical record, the observed trends cannot be used to reliably
predict future changes to the season.

In addition to the season getting longer, it also seems that Atlantic hurricanes are getting
stronger, along with tropical cyclones in other ocean basins. Writing in Nature, J. Elsner

of Florida State University, and colleagues, conﬁrm a 30-year trend that shows tropical
cyclones in the Atlantic getting progressively stronger, in line with SST rises in the Atlantic
region and elsewhere. By analysing homogeneous satellite wind-speed data they also show
that a comparable trend can be seen in the rest of the tropics. The authors demonstrate
a signiﬁcant overall upward trend for wind-speeds in the most powerful tropical cyclones,
which they also recognise in the very strongest storms over each ocean basin, and most
noticeably in the North Atlantic. As explanation, Elsner and his co-researchers, note that their
results are consistent with the simple and straightforward idea that as the oceans warm they
have more energy to convert to tropical cyclone wind.

While the prospect of more powerful tropical cyclones is a concern, this only becomes a real
problem when and if such storms make landfall. For the US, this is an issue tackled by
K. Coughlin

and colleagues at Risk Management Solutions and Germany’s Alfred

Wegener Institute in Bremerhaven. The authors note that the fraction of Atlantic hurricanes

Figure 16

Hurricane Vince close to
Madeira in October 2005.

Courtesy: EUMETSAT.

Hazard & Risk Science Review 2009

that make landfall in the US is a measure of the relationship between landfalling and basis
hurricane numbers. Using statistical tests, they go on to establish that there is a signiﬁcant
change in the proportion of Atlantic hurricanes that struck the US coastline between the
ﬁrst and second halves of the 20

century, relative to hurricane activity in the Atlantic Basin

as a whole; supposedly an artefact of a paucity of basin observations in the earlier period.
After 1948, however, Coughlin and co-workers analysis reveals no change in the proportion
of Atlantic hurricanes striking the US coast, despite signiﬁcant changes in the total number
of hurricanes. They note, however, that there probably are changes – to some degree –
in both the proportion and number of landfalling hurricanes, but that both are masked by
the fact that the scarcity of landfalling hurricanes means that there are insufﬁcient data to
resolve potential changes. Notwithstanding this, the authors conclude that by assuming
that the proportion of landfalling hurricanes is constant, information about the total number
of hurricanes in the Atlantic Basin can be used to make more accurate forecasts of US
landfalling hurricane activity in the future.

Climate change and extra-tropical cyclones

As with tropical cyclones, a lively debate is ongoing about whether climate change will
mean more and/or more intense windstorms outside the tropics, and in particular across
the European region. The issue of extra-tropical cyclones in a warmer world is reviewed in
Theoretical & Applied Climatology by U. Ulbrich

of the Freie Universität Berlin, and co-

authors, who acknowledge at the start that it is not easy to present a common viewpoint of
extra-tropical cyclones, their variability and trends, either in the real world or in global climate
model (GCM) studies. Nevertheless, the authors arrive at conclusions that broadly support
the ﬁndings of the 2007 IPCC (Intergovernmental Panel on Climate Change) 4th Assessment
report, namely that global warming will shift storm tracks towards the poles in both
hemispheres (most noticeably in the southern hemisphere) and drive greater storm activity
at higher latitudes. More speciﬁcally, two regions of high cyclonic activity are predicted in
the northern hemisphere; one over the North Paciﬁc and the other over the North Atlantic,
with a secondary region of elevated activity over the Mediterranean. Broadly-speaking, a fall
in winter cyclone numbers is projected, although a rise in the number of intense cyclones
is forecast across the NE Atlantic, the British Isles and the North Paciﬁc. In the southern
hemisphere, the existing southward band of cyclonic activity is predicted to move further
south, leading to reduced cyclonic activity around 50º S and increased activity around 60º S.

In a paper in Geophysical Research Letters, P. M. Della-Marta

of MeteoSwiss and

J. G. Pinto of the University of Cologne, look ahead to the potential impact of a warmer
climate on winter storms over the North Atlantic and Europe. For two IPCC emissions
scenarios (A1B – more optimistic and A2 – less optimistic), the authors quantify changes in
storm frequency and intensity across the region, in terms of return periods. By 2100, under
both scenarios, they show that both the minimum central pressure (CP) and the maximum
vorticity (VOR) of North Atlantic storms are projected to remain unchanged compared to the
present day. For the British Isles – North Sea – western Europe area, however, shorter return
periods for VOR of all intensities are apparent as early as 2040. Although they speculate
that the changes may be unrealistically large, with a 50 (20) year event projected to become
a 9 (5.5) year event under both scenarios, Della-Marta and Pinto suggest that the likely
consequences are a higher incidence of damaging wind events across Europe.

Climate change, extreme wind events and tornadoes

It might seem intuitively reasonable that a warmer, more dynamic atmosphere will be
characterised by more extreme winds, but the true situation is far more complex. Writing in
Geophysical Research Letters, G. Gastineau

and B. J. Soden of the Rosenstiel School

for Marine and Atmospheric Science in Miami, use a multi-model ensemble of coupled
climate model simulations to investigate changes in the frequency of occurrence of extreme
windstorm events in response to anthropogenic global warming. The authors conclude that
a weakening of large-scale atmospheric circulation results in a reduction in the frequency

Hazard & Risk Science Review 2009

of the most extreme wind events over the tropics. Conversely, however, they found that
the strongest near-surface wind events increased in frequency at higher latitudes, as the
mid-latitude storm tracks migrated towards the poles. As a consequence of higher moisture
contents in the lower troposphere, the frequency of the heaviest precipitation events was
also projected to increase.

It has often been speculated that climate change may drive more of the severe
thunderstorms that spawn destructive tornadoes, but is this the case, and if so can a climate
change signal already be recognised? This issue is examined by N. S. Diffenbaugh

Purdue University, and colleagues, in a paper in Eos, who look at trends in US tornado
activity in an attempt to unravel a climate change signal (Figure 17). The authors note that
the number of reported tornadoes in the US has been rising, on average, by about 14 a year
over the past century, although this may well reﬂect an inconsistent reporting record rather
than a true, progressive rise. Surprisingly, the numbers of the most damaging tornadoes
(F2 – F5) have actually fallen over the past 50 years, although again this may be an
artefact. The authors note that tornado activity may be affected by natural climate variations
associated, for example, with ENSO or shifts in the jet stream. A climate change effect
could also make itself felt, although this could go either way – resulting in more or fewer
tornadoes. Looking ahead, Diffenbaugh and his co-authors note that higher greenhouse
gas concentrations in the atmosphere are projected to increase the incidence of severe
thunderstorms in the US, and therefore opportunities for tornado formation, but this is not a
robust forecast. In conclusion, the jury remains out in relation to whether global warming will
bring more tornadoes or suppress their formation.

Climate change, precipitation and ﬂooding

One of the broadly accepted tenets of climate change is that it will result in more extreme
precipitation events and increased ﬂood risk and, indeed, this trend is already detectable in
some parts of the world. The detail remains, however, to be determined, and the resolution of
predictive models – although improving all the time – remains less than desirable. Here, the
ﬁndings of three papers relating to future European precipitation and ﬂood risk in a warmer
world are presented. In the ﬁrst, J. Kysely

and R. Beranova, of the Institute of Atmospheric

Physics in Prague, examine the effects of climate change on extreme precipitation events

Figure 17

Average yearly patterns
of tornado occurrence in
the United States based
on 1954–2007 data. The
annual totals were ﬁt with
a linear regression in time
and adjusted to 2007 values
to account for changes in
tornado monitoring and
other factors through time.
The progression of each
year was then scaled to
ﬁt the adjusted annual
total, and quantiles (the
grey, blue, thin black, and
thick black curves) were
calculated for each day
of the year from those
adjusted series. 2008’s
tornado count (red) is
shown for comparison.

Courtesy: American
Geophysical Union.

Hazard & Risk Science Review 2009

in central Europe. Writing in Theoretical & Applied Physics, the authors note that although
the hydrological cycle is predicted to intensify and becomes increasingly volatile, large
uncertainties remain, in particular in relation to the seasonal and spatial variability of future
precipitation changes. To address this, Kysely and Beranova evaluate scenarios of changes
in extreme precipitation events in 24 future climate runs of 10 regional climate models,
focusing on a part of the Czech Republic where local conditions make future projections
difﬁcult using global models. The authors conclude that heavy precipitation events are likely
to increase in winter, with summers generally drier. Uncertainties are large, but the broad
pattern matches recently observed trends, lending weight to the conclusions. Unsurprisingly,
Kysely and Beranova conclude that their ﬁndings raise expectations of signiﬁcantly elevated
peak river discharges and increased ﬂood risk.

Looking at the broader picture, R. Dankers

and L. Feyen of the Joint Research Centre at

Ispra (Italy) use high-resolution climate simulations to examine the impact of climate change
on ﬂood hazard across the continent. In the Journal of Geophysical Research, the authors
note that by 2100, under IPCC emissions scenario A2, extreme discharge levels in many
European rivers may increase in both frequency and magnitude (Figure 18). In some rivers
in the west, and in parts of eastern Europe, for example, 100 year ﬂoods could occur every
50 years or less. In contrast, ﬂood hazard appeared to be reduced in rivers in central and
southern Europe, and in the north-east.

Continuing the pan-European ﬂood hazard theme, J. L. Barredo

, also of the Joint

Research Centre at Ispra, evaluates ﬂood losses in Europe between 1970 and 2006. Writing
in Natural Hazards & Earth Systems Sciences, Barredo presents an assessment of ﬂood
losses in 31 European nations, normalised to current societal conditions and taking account
of a range of factors at country level, including population changes, wealth and inﬂation.
Despite some perceptions to the contrary, the author concludes that no climate-change
signal is detectable in ﬂood loss trends, with an observed increase in ﬂood losses over the
period explainable purely by changes in societal factors.

Figure 18

Change in recurrence of
a 100-year ﬂood in the
control run of the H12A2
Intergovernmental Panel
on Climate Change (IPCC)
emissions scenario. In
some rivers in the west, and
in parts of eastern Europe,
for example, 100 year
ﬂoods could occur every 50
years or less. In contrast,
ﬂood hazard appeared to be
reduced in rivers in central
and southern Europe, and
in the north-east.

Courtesy: American
Geophysical Union.

Hazard & Risk Science Review 2009

Sources and further reading

61. Arpe, K. et al. 2009 Atlantic hurricanes – testing impacts of local SSTs, ENSO,
stratospheric QBO – implications for global warming. Quaternary International 195, 4 – 14.

62. Barredo, J. L. 2009 Normalised ﬂood losses in Europe 1970 – 2006. Natural Hazards &
Earth Systems Sciences 9, 97 – 104.

63. Coughlin, K. et al. 2009 A relationship between all Atlantic hurricanes and those that
make landfall in the USA. Q. Journal of the Royal Meteorological Society 135, 371 – 379.

64. Dankers, R. & Feyen, L. 2008 Climate change impact on ﬂood hazard in Europe: an
assessment based on high-resolution climate simulations. Journal of Geophysical Research
113. doi:10.1029/2007JD009719.

65. Della-Marta, P. M. & Pinto, J. G. 2009 Statistical uncertainty of the changes in winter
storms over the North Atlantic and Europe in an ensemble of transient climate simulations.
Geophysical Research Letters 36. doi:10.1029/2009GL038557.

66. Diffenbaugh, N. S., Trapp, R. J. and Brooks, H. 2008 Does global warming inﬂuence
tornado activity? Eos 89, 553 – 554.

67. Elsner, J. B. 2008 Hurricanes and Climate Change. Bulletin of the American
Meteorological Society 89, 677--679.

68. Elsner, J. B.; Kossin, J. P. & Jagger, T. H. 2008 The increasing intensity of the strongest
tropical cyclones. Nature 455, 92--95.

69. Elsner, J. B. & Jagger, T. H. 2008 Hurricanes and Climate Change. Springer. Berlin.
380pp.

70. Gastineau, G. & Soden, B. J. 2009 Model projected changes of extreme wind events in
response to global warming. Geophysical Research Letters 36. doi:10.1029/2009GL037500.

71. Gillet, N. P. et al. 2008 Attribution of cyclogenesis region sea surface temperature change
to anthropogenic inﬂuence. Geophysical Research Letters 35. doi:10.1029/2008GL033670.

72. Gualdi, S. et al. 2008 Changes in tropical cyclone activity due to global warming: results
from a high-resolution coupled general circulation model. Journal of Climate 21, 5204 – 28.

73. Kossin, J. P. 2008 Is the North Atlantic hurricane season getting longer? Geophysical
Research Letters 35. doi:10.1029/2008GL036012.

74. Kysely, J. et al. 2009 Climate-change effects on extreme precipitation in central
Europe: uncertainties of scenarios based on regional climate models. Theoretical & Applied
Climatology 95, 361 – 374.

75. Steenhof, P. A. 2008 The impact of tropical sea-surface temperatures on various
measures of Atlantic tropical cyclone activity. Theoretical and Applied Climatology 92, 249 –
255.

76. Ulbrich, U., Leckebusch, G. C. & Pinto, J. G. 2009 Extra-tropical cyclones in the present
and future climate: a review. Theoretical and Applied Climatology 96, 117–131.

77. Vecchi, G. A. et al. 2008. Climate Change: whither hurricane activity? Science 322,
687 – 89.