See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/255989247

Basic principles of multi-risk assessment: A case study in Italy

Article  in  Natural Hazards · June 2012

DOI: 10.1007/s11069-012-0092-x

CITATIONS READS

112 1,560

5 authors, including:

Alexander Garcia-Aristizabal Paolo Gasparini

National Institute of Geophysics and Volcanology INFN - Istituto Nazionale di Fisica Nucleare
61 PUBLICATIONS   354 CITATIONS    144 PUBLICATIONS   3,399 CITATIONS   

SEE PROFILE SEE PROFILE

Maria Laura Mastellone

Università degli Studi della Campania "Luigi Vanvitelli
71 PUBLICATIONS   1,526 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Production of high quality diesel from plastic waste from MRF View project

Performance indicators of a material recovery facility View project

All content following this page was uploaded by Alexander Garcia-Aristizabal on 23 May 2014.

The user has requested enhancement of the downloaded file.

 Basic principles of multi-risk assessment: a case
study in Italy
Warner Marzocchi(1) Alexander Garcia-Aristizabal(2)
Paolo Gasparini(2) Maria Laura Mastellone(3)
Angela Di Ruocco(2)
Reference: W. Marzocchi, A. Garcia-Aristizabal, P. Gasparini,
M. L. Mastellone, and A. Di Ruocco (2012). Basic principles of
multi-risk assessment: a case study in Italy. Nat. Hazards, 62(2),
551-573. DOI: 10.1007/s11069-012-0092-x

Abstract
The assessment of the impact of different catastrophic events in a given
area require innovative approaches that allow risks comparison and that
account for all the possible risk interactions. In the common practice,
the risk evaluation related to different sources is generally done through
independent analyses, adopting disparate procedures and time-space reso-
lutions. Such a strategy of risks evaluation has some evident major draw-
backs as, for example, it is difficult (if not impossible) to compare the risk
of different origins, and the implicit assumption of independence of the
risk sources leads to neglect possible interactions among threats and/or
cascade effects. The latter may amplify the overall risk and potentially
the multi-risk index could be higher than the simple aggregation of single
risk indexes calculated considering each source as independent from the
others. In this paper we put forward some basic principles for multi-risk
assessment, and we consider a real application to Casalnuovo municipality
(Southern Italy), in which we face the problem to make different hazards
comparable, and we highlight when and how possible interactions among
different threats may become important.

1 Introduction
Several risks emerge within a complex relationship that exist between human
and environmental factors, and impend on the environment, on public and pri-
vate properties, on strategic and priceless infrastructures in any inhabited area
of our planet. They are generated by different sources, both natural and an-
thropogenic, and have different relevance depending on the synergy between
∗(1) Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Rome,

Italy; (2) Center for the Analysis and Monitoring of Entironmental Risk (AMRA), Via Nuova
Agnano 11, 80123 Napoli, Italy; (3) Dipartimento di Scienze Ambientali, Seconda Università
di Napoli

1
the generated events and on the peculiar interaction with the region where they
break out. A joint analysis, quantification and comparison of all risks (both
natural and anthropogenic) that potentially can affect a territory is a basic fac-
tor for development of a sustainable environment and land use planning, as well
as for competent emergency management before and during catastrophic events
(Durham, 2003).

Objective difficulties in the quantification of (possibly interacting) risks and

their comparison exist, which make of it a complex discipline. The first source
of difficulty comes from methodological issues; for instance, to date the risk
evaluation related to different sources is commonly done through independent
analyses, adopting disparate procedures and time-space resolutions. In most of
cases, only qualitative estimates of the risk level are available. Such a strategy
of risk evaluation has some evident major drawbacks: (1) it is difficult, if not
impossible, to compare risks of different origins; (2) the implicit assumption of
independence of the risk sources leads to neglect possible interactions among
threats and/or the so-called ‘cascade’ effects. In practice, this means that a po-
tential ‘multi-risk’ index could be higher than the simple aggregation of single
risk indexes calculated considering each source as independent from the others.
Multi-risk assessment (MRA hereinafter) is meant to overcome the two draw-
backs listed above. This requires the use of the same space-time window, the
same metric for losses evaluation, and to account for the possible interactions
among the risks, both in terms of probabilistic hazard and vulnerability.
Another difficulty source for MRA, shared with more classical risk analysis,
comes from epistemological issues that are related to the nature of scientific
knowledge; for instance, risk assessment is based on the convolution of dif-
ferent complex concepts as hazard source identification and assessment, and
vulnerability evaluation; each concept bears implicit assumptions and modeling
methodologies that can propagate uncertainties that should not be ignored, but
they have to be incorporated into the final assessment in a coherent way. Fi-
nally, another obstacle for MRA is that scientists of various disciplines do not
use a common terminology; in some cases scientists dealing with different types
of risks assign different definitions even to the same term.
All these factors, and many other that may arise if we consider other cultural
and human factors, have probably determined the slow development of multi-
risk methodologies. As a matter of fact multi-risk evaluation is a relatively new
field, until now developed only partially by experts with different backgrounds
(engineering, statistics, seismology, toxicology etc.). Among the few works on
this field we quote the UNDRO study (UNDRO, 1977), the KATANOS report
(BZS, 1995), Granger et al (1999), Van Westen et al (2002), Ferrier and Haque
(2003), Blong (2003), Grunthal et al (2006), ?, and Schmidt et al (2011). How-
ever the specific problem of possible interactions among different threats and/or
cascade effects has been approached qualitatively only by ?.
In this paper, we discuss some basic principles for MRA, emphasizing the two
main issues that are risks comparability and risks interaction. Through a real
example considered in the EC FP6 NaRaS project, we rank the different risks
and we show how hazards interaction may lead to the underestimation of some
risks. Before discussing the main subject of this paper, a clarification of the
used terminology and a homogenization of the concepts used by the scientists
and practitioners in the different risks areas is imperative. This effort does not

2
aim at providing ‘the solution’ of the lack of homogeneity in terminology, but
just at being a useful reference to clarify the meaning of the terms used here.
Different terminology and definitions of the same terms are used in the prac-
tice of risk evaluation for natural and anthropogenic risks. In our work we define
the meaning of each term to make easier a comparison with the use made in
other disciplines. The definitions we have adopted are the following (a more
detailed list may be found, e.g. in Marzocchi et al, 2009):

• Risk source or hazard: anything that can potentially generate adverse

events and consequently create damage to the population and/or environ-
ment. It is therefore a concept related only to the intrinsic characteristics
of a substance, a plant or of a physical/geological status of a site.

• Probabilistic hazard (H): the probability that a certain adverse event

generating a phenomenon of a given intensity will occur in a given area and
in a given time interval. Consequently, probabilistic hazard is evaluated
by taking into account the characteristics of the risk source, the location
to which it refers and the physical process of intensity diffusion from the
risk source location to the investigated area.

• Value at risk (L): it measures the total potential loss due to an adverse
event in a given area. It can be expressed in human casualties, either in
economic or conventional terms (since it is difficult to express heritage or
environmental losses monetarily). It depends on the various activities (hu-
man, cultural, economic) as well as on the environmental characteristics
of the referred area.

• Vulnerability (V) : considered here just in terms of physical vulnera-

bility, it is the fraction of L that could be lost after a specific adverse
event.

• Damage: the amount of destruction or damage, either in health, finan-

cial, environmental functional and/or other terms as a consequence of an
occurred hazard; its value can be obtained by multiplying the value at risk
(or exposed value) by the vulnerability.

• Risk: the combination of the consequences of an event (damage) and the

associated probability of its occurrence (probabilistic hazard); it may be
understood as the expected loss due to a particular hazard for a given
area and reference period. In some cases (e.g. when the exposed value is
not explicitly considered), risk is understood as the probability (or rate)
that a negative consequence (e.g. loss) can occur in a given period of time
following a specific adverse event.

2 The notion of Risk in Natural and Environ-

mental Sciences
For environmental and natural issues, risk factors can be conveniently defined
as a function of the probability that a certain event will occur and of the extent

3
of the damage caused to man, environment and objects. Thus, it is often quan-
tified by

R=H×L×V (1)
where H is the probabilistic hazard, L the value at risk, and V the vulnerability,
as defined in section 1. This definition basically coincides with that provided
by the European Community (EN 1050, 1996) which indicates risk related to
a specific source (or hazard) as a function of the magnitude of the potential
damage (defined as D = L × V) that may result from the considered hazard
and from the probability that it will occur (also a function of the frequency and
duration of the exposure, of the probability it will occur and of the possibility to
avoid or limit the damage). Figure 1 summarizes the main aspects to be consid-
ered for the evaluation and how they combine for a generic, single-source, risk
assessment. It follows that risk is a non-normalized probability because it has a
lower limit (which is zero when there is no appreciable possibility of an adverse
event occurring or when the degree of resulting damage is practically null), but
not necessarily an upper limit. The presence of an upper limit (equal to one) is
possible only if the value at risk in equation 1 is provided as percentages of the
area’s total value (normalized risk).

Equation 1 is an obvious simplification of the reality because usually prob-

abilistic hazard, vulnerability and value are not single values; more generalized
formulations can be found in literature, as for example that reported by the
Pacific Earthquake Engineering Research (PEER) Center, commonly used for
seismic risk assessment. In a simplified and more general version (Cornell and
Krawinkler, Spring 2000), the PEER formula reads
Z Z
λ(`) = G(`|D)dG(D|H)dλ(H) (2)
D H

where λ stands for a generic rate (i.e., λ(`) is referred to a loss rate, while
λ(H) is referred to a rate of assessed probabilistic hazard), G is a cumulative
distribution function, ` is the loss in one specific metric, D is the damage, and
H is the probabilistic hazard. Despite equation 2 is commonly used for seismic
risk assessment, we argue that it may be used as a general formulation for
any kind of natural risks, since it simply describes how the probabilistic hazard
analysis (λ(H)), the damage analysis (dG(D|H)), and the loss analysis (G(`|D))
are interconnected. From a formal point of view, equation 2 holds for hazards
that can be considered time-independent in the forecasting time window. The
inclusion of possible time-dependency requires some changes of equation 2, but
it does not pose any conceptual difficulty (e.g., Der Kiureghian, 2005).
The introduction of the concept of probability in risk analysis is useful for
many important reasons; for example: (1) it allows an immediate gradation
of environmental risks and may even facilitate comparison and management
of events of a different nature (including, e.g., terrorist attacks); (2) it can be
used as supporting tool during the decision-making phase and, in particular, to
perform the cost/benefits analyses of management strategies; (3) it allows the
definition of the fundamental concept of societal “acceptable risk”, i.e. the level
of risk that a community is willing to accept. In this way, it establishes an ob-
jective reference both for land use planning and for a selection of risk mitigation

4
actions based on a cost/benefit analysis. An example of risk mitigation through
cost/benefit analysis in volcanology is given by Marzocchi and Woo (2007, 2009)
and in seismology by Van Stiphout et al (2010).

3 The multi-risk assessment (MRA)

3.1 Basic principles
The purpose of MRA is to establish a ranking of the different types of risk taking
into account possible interactions among the single hazards/risks, because such
interactions may amplify the overall risk. The procedure for MRA is illustrated
here by identifying the main steps to be followed.
1. Definition of the space-time window for the risk assessment and the metric
for evaluating the risks;
2. identification of the risks impending on the selected area;

3. identification of selected hazard scenarios covering all possible intensities

and relevant hazard interactions;

4. probabilistic assessment of each scenario;

5. vulnerability and exposure assessment for each scenario, taking into ac-
count the vulnerability of combined hazards;
6. loss estimation and multi-risk assessment;

The definition of the metric and target should precede the identification of
the risks, because, for example, some risks can create a significant amount of eco-
nomic losses without threatening the life of the persons. Therefore, depending
on the metric adopted, some risk may have a large or small interest. These steps
are those usually used to perform a single environmental risk assessment with
some important differences: first, the estimation of the multi-risk index has to
take into account possible cascade and/or triggered related adverse events (see
step 3 above); second, a common space-time window and metric for the risk
evaluation have to be defined a priori (step 1). The latter derive directly from
the end-users needs, imposing a bottom-up risk assessment scheme. For exam-
ple, emergency management and land use planning require different space-time
windows and different metrics for risk assessment.

3.2 Hazards/risks interaction

Different hazards may interact one to each other, and also the vulnerability may
change dramatically if two adverse events occur close in time (e.g., Zuccaro et al,
2008). Here, we introduce some concepts of the interaction between hazards,
but we argue that similar ideas may be straightforwardly extrapolated also for
vulnerability assessment. Let us introduce this topic with a simplified example.
We consider two different threatening events, whose occurrence is E1 and E2 .
In general, the probability of E1 occurrence (H1 ) can be written as

H1 = p(E1 ) = p(E1 |E2 )p(E2 ) + p(E1 |Ē2 )p(Ē2 ) (3)

5
where p represents a probability or a distribution of probability (see Marzocchi
et al, 2004), and Ē2 means that the event E2 does not occur. The generalization
of equation 3 to more than two events does not pose any particular conceptual
problem even though it may require cumbersome calculations. Hence, we think
that this two-events case highlights the most interesting features of the interac-
tion among events. Below we report three main cases where hazards interaction
may play an important role.

1. In common practice, the probabilistic hazard assessment of E1 is usually

estimated through the analysis of existing databases that may account
or not in a proper way the occurrence of the event E2 . Let us make an
example of particular relevance in the practical case that we describe in
the following. The probability of failure of a specific component of an
industrial site (E1 ) is usually taken from a database of failures of similar
components all around the world. Anyway, close to the specific industrial
site under analysis, there may be a natural threat (E2 ) that can amplify
many orders of magnitude the probability of occurrence of E1 (through
eq. 3). In this case, the database does not provide p(E1 ), but p(E1 |Ē2 ).
This could be a reasonable estimation of p(E1 ) only when p(E2 ) is close to
zero, but this may not hold in specific areas. It is worth noting that some
existing good practice in probabilistic hazard assessment implicitly ac-
knowledge this problem; a remarkable example is the nuclear power plant
risk assessment that include the possible damages caused by earthquakes
(Budnitz et al, 1997). Moreover, in some cases, databases may already
include an unbiased estimation of H1 , because the database accounts for
the occurrence of E2 ; an example is given by the tsunami databases that
already included the possibility that a tsunami is caused by an earthquake.

2. In the long-term hazard/risk assessment, the databases provide p(E1 ) as-

suming that p(E2 ) has been constant through time, but this may be vi-
olated by the time changing properties of the boundary conditions. A
typical example is the hurricane hazards that has to account the time-
varying conditions imposed by the global warming.
3. Hazard/risk interactions play a basic role in short-term (e.g., of the order
of days) probabilistic hazard/risk assessment. For example, the occur-
rence of heavy rain (E2 ) changes significantly the landslide occurrence
(E1 ) probability in a time horizon of few days. Whether forecast tracks
the time evolution of p(E2 ) that leads to a time evolution of p(E1 ) through
equation 3. In this case, the most challenging scientific problem is the re-
liable estimation of p(E1 |E2 ) that may be obtained by theoretical models,
empirical laws, or expert opinion. It is worth remarking the differences
with the long-term assessment; in the long-term perspective, a local land-
slides database already accounts for the fact that most of the landslides
are due to heavy rain or storms. In other words, it is expected that the
database provides a reliable estimation of p(E1 ) directly. The same conjec-
tures may be applied for earthquakes and fire, earthquakes and eruptions,
and many other couples of hazards. Notably, while interactions among
hazards are ubiquitous, their impact on MRA is not always relevant. The
scenarios of the step 2 have to include the most important ones.

6
3.3 A practical procedure for MRA
In this subsection, we introduce one practical procedure for MRA following the
steps listed above. A schematic view of this process is summarized on Figure 2.
Note that the goal here is not to write the ’golden rules’ for MRA, but simply
to illustrate how the MRA principles of subsections 3.1 and 3.2 may be applied
in real cases.

3.3.1 Definition of the space-time window and the risk metric

In practical cases, the request for risk assessment comes from the end users
or stakeholders imposing a bottom-up procedure. The starting point may be
nicely summarized by some typical questions/problems posed by stakeholders:
(i) I am the mayor of a municipality that is threated by different risks; how
can I prioritize them in order to spend at best funds for their mitigation? (ii)
I am the responsible of an emergency team; in order to take mitigation actions
during an emergency I need real-time forecast for each different risk scenario.
(iii) I am responsible of land-use planning; is this area safe enough to build
an important industrial plant? Of course a large variety of similar questions
may be posed. Each one of them implicitly contains the basic input parameters
for MRA, that is, the risk metric of interest (e.g. human or economic losses),
the spatial scale and resolution of interest, and the time window. For example,
the three questions/problems reported above require completely different time
windows; the risk comparison requested by question (i) requires a typical time
frame that can facilitate the comparison, e.g., one year. Problem (ii) requires
much shorter time windows, i.e., days or weeks. Conversely, the last problem
(iii) involves a much longer time frame, typically decades or centuries. To sum,
the practical problem constrains the space-time window and the risk metric that
have to be clarified as a first step of MRA.

3.3.2 Identification of hazard/risk sources

Once the space-time window of interest has been defined, the hazards/risks
threatening that area have to be identified. The basic strategy is to look at what
happened in the past in the selected area; then, MRA requires also a specific
attention to hazards that do not pose direct threats but that can trigger more
impacting events. For example, in a short-term risk assessment, the occurrence
of heavy rain may not impact directly the territory but it causes an increase in
floods and landslides probabilistic hazard.

3.3.3 Identification of the scenarios and their probabilistic assess-

ment
In many cases, it can be unwieldy to consider all kind of hazards and their
possible interactions in MRA. The problem becomes tractable if we simplify the
problem using some selected scenarios. They may be few or many depending
on the peculiarities of the problem under consideration and the availability of
data and information. The basic point is that the set of scenarios should cover
all possible intensities of the hazards. The use of single“ worst case scenarios”
may provide strongly biased estimations (e.g. Selva et al, 2010) and make the
different risks incomparable.

7
 The scenarios for hazards interaction depend on the time scale under consid-
eration. As mentioned above, short-term (few days) and long-term (years) risk
assessment may involve completely different scenarios of risks interaction. The
event tree (ET) structure (e.g., Cooke, 1991; Newhall and Hoblitt, 2002; Mar-
zocchi et al, 2004, 2008, 2010) is particularly suitable for describing scenarios
composed by different chains of events. The ET is a tree graph representation
of events in which individual branches are alternative steps from a general prior
event, state or condition through increasingly specific subsequent events (inter-
mediate outcomes) to final outcomes. In this way, ET should show all relevant
possible outcomes of a risk process at progressively higher degrees of detail. ET
is basically described by its general structure that is case-dependent (see for
instance, Cooke, 1991; Marzocchi et al, 2004, 2008), and by the probability
at each node. The combination of these probabilities provides the probability
of any event at which we may be interested in. Notably, if the ET contains
vulnerability and exposure, the output may be directly the MRA.
A recent Bayesian evolution of ET includes a probability density function
(PDF) instead of a single value for the probability at each node (Marzocchi et al,
2004, 2008, 2010). In this way, both aleatory and epistemic uncertainties may
be accommodated and assessed. This procedure may be generalized to any kind
of probabilistic assessment (including or not the definition of ET) and it has the
paramount advantage to properly account for the different pieces of available
information, like data, models and expert opinions, weighting and merging them
into a final probabilistic assessment (e.g., Grezio et al, 2010).

3.3.4 Vulnerability and exposure assessment

A complete vulnerability analysis, as defined at the beginning of this paper, has
to start from a careful identification of the assets under threat. They generally
include: (a) man made structures, infrastructures and buildings; (b) cultural
heritage; (c) life-lines; (d) humans and animals; (e) agricultural and forest areas;
and (f) ecosystems. The evaluation of vulnerability of the first three categories
is generally made through the computation of vulnerability functions enabling
to assess the distribution of the expected damage in each typology of buildings
for each expected intensity of hazards. With respect to MRA, vulnerability
functions and exposure have to be estimated for each single risk scenario, where
the interaction among hazards and vulnerabilities have to be carefully accounted
for. In fact, the vulnerability to two different concomitants threats may be much
larger than the simple ”sum” of the two single vulnerabilities. For example,
the seismic vulnerability of a building with some ash load on the roof may be
dramatically higher than the vulnerability in regular condition. Therefore, even
though few centimeters of ash does not represent a direct threat for the building,
they may significantly increase its seismic vulnerability and the risk as well (e.g.
Baratta et al, 2004; Spence et al, 2007; Zuccaro et al, 2008).
Humans, animals, and ecosystems have a vulnerability due to direct con-
tact with the phenomenon (exposure to toxic substance, direct contact with a
flood surge or a tsunami wave) and a vulnerability due the impact of damaged
structures (collapses of parts or whole buildings, breakdown of bridges, lifelines,
industrial components, etc.). A complete evaluation of population vulnerability
requires a knowledge of the percentage of the total inhabitants residing in build-
ings of different vulnerability, in industries, working in dangerous situation, and

8
the time distribution during the day. Statistical models of people distribution
during each day of the week and in different seasons are a way to get reliable
information on the exposure of the population. Each additional information
(age distribution, percentage of unable population, etc) increases the reliability
of the model.

3.3.5 Loss assessment and MRA

The probabilistic assessment of each hazard scenario and the exposure and vul-
nerability assessment are two components of equation 2. The last factor in
equation 2 to be calculated is the common metric for the losses of all risks or,
alternatively, the identification of a selected kind of loss (reference loss), for
instance the risk of having a number of casualties or introducing the related
concept of ’micromort’ (Howard, 1980), defined as a unit of risk measuring a
one-in-a-million probability of death. The use of different reference loss instead
of an economic metric may overcome the problem of assigning a monetary value
to human life (e.g., Marzocchi and Woo, 2009) that is necessary if we want to
compare risks to damage structures or infrastructures with risks for human life.

4 Multi-risk assessment: a case-study

The case-study of the Casalnuovo municipality, developed during the NaRaS
project (e.g., Marzocchi et al, 2009), is used to describe a possible procedure
for MRA. The emphasis here is not on how single risks have been calculated,
but on how they have made comparable and on the effects of the interactions
among them. Noteworthy, in this application we compare only the best guess of
the risks assessment without considering their uncertainties. We argue that this
is a major issue that must be considered in future analysis in order to improve
the capability to compare different risks. One possible strategy for uncertainty
assessment to probabilisitic estimation has been suggested by Marzocchi et al
(2004, 2008).
Casalnuovo is a municipality of the Campania region located 12-13 km NE
of the crater of Mt. Vesuvius volcano. The resident population is 47940 and
the town consists of 3615 buildings, most of them being apartment houses.
Several public (e.g. schools) and commercial buildings are located mainly at
the outskirts of the town. Because of its location, Casalnuovo is exposed to
adverse events from different sources: Mt. Vesuvius volcano, the Irpinia tectonic
earthquake source, a small river passing through the municipality. The presence
of industries and illegal landfills widens the number of possible threats (Figure
3).
A sound MRA is fundamental to give answers to typical questions posed by
local authorities. What is the most dangerous hazard that threats the munic-
ipality? If we have some resources for mitigating risks, where we have to put
our money? A direct comparison of single risks threatening Casalnuovo is hard
to do with the available data since they are referred to quite different time win-
dows and typologies of damages. Moreover, possible triggering effects were not
considered at all. Risk ranking and risk assessment including possible triggered
(cascade) events are basic issues to be approached by MRA.

9
4.1 Ranking the single risk assessments
The ranking of risks requires that each typology is calculated under the same
conditions. For the Casalnuovo case-study, we use a common time frame of
one year, and the risk analysis is focused on a damage consisting of human
life loss. Table 1 summarizes the hazard sources identified for the Casalnuovo
area, and the estimates of hazard, vulnerability, and exposed value for each case
(all estimated for a time window of one year). We explore only some selected
risks and we combine them in a MRA perspective using available probabilistic
hazards and vulnerability assessment. It is worth remarking that the main goal
here is not to provide the ”real” risk assessment for that area, but to give some
hints about how MRA should be applied. Using these information (Table 1,
for details on calculations see Appendix A, and Marzocchi et al (2009)), it is
possible to rank the annual risk for human life as follows:

Volcanic risk: RVolc = 1.37 · 10 0

Seismic risk: RSeis = 5.00 · 10−3
Flooding risk: RFlood = 4.20 · 10−5 (4)
Landslides risk: RLand = 6.00 · 10−7
Industrial risk: RInd = 1.83 · 10−7

4.2 MRA due to triggering effects

The inclusion of interaction among adverse events may be very important to
obtain a reliable MRA. Following the practical procedure delineated before, we
have identified for the Casalnuovo case one possible scenario where threats may
interact in a long-term perspective.
Specifically, in this example we assume that, from a long-term perspective,
the seismic, volcanic, flooding and landslide probabilistic hazards have been cal-
culated by databases that already take properly into account the simultaneous
occurrence of different hazards. This is a reasonable assumption since we have,
for that area, long historical catalogs. For example, the landslides catalog al-
ready accounts for the landslides that have been triggered by earthquakes. The
industrial risk has been calculated through a database that accounts for natural
failures due to micro-damages that may evolve to a major failure of the system.
The potential interaction with seismic activity has been already included trough
the building code derived from the seismic probabilistic hazard map. In prac-
tice, it is expected that the moderate seismicity of Casalnuovo does not lead
to major amplification of the failure probability. Conversely, the database does
not account for the fact that the industrial site is close to an active volcano
that may produce ash fall loads and create a significant increase of a major
failure probability of the system. In this case, it is reasonable to think that the
industrial risk may be significantly underestimated if the probability of having
a significant ash fall on the industry site is comparable if not higher than the
probability to have a natural major failure of the system.
The long-term scenario for risks interaction considered for Casalnuovo can be
described as follows: while a 10 cm thickness of ash does not have virtually any
direct effect on human beings because it does not lead to building collapses, it
originates the collapse of a pipe-bridge inside the industry, causing an explosion
and the subsequent air and water contamination. We assume that the vulnera-

10
bility function of the pipe-bridge has a Heaviside function, 0 for ash thicknesses
less than 10 cm, and 1 for thicknesses of 10 cm or more. The annual proba-
bility of a pipe-bridge collapse damage is 2.09 · 10−4 (see also Marzocchi et al,
2009), while the annual probability to have a 10 cm thickness or more on the
industry is 3.6 · 10−3 , i.e., more than one order of magnitude larger (ibid.). This
means that, keeping the probability to all other nodes equal, the industrial risk
in Casalnuovo is underestimated one order of magnitude if considered alone. In
terms of equation 3, if E1 is a major pipe-bridge failure and E2 is occurrence of
ash fall loading of 10 cm or more, the probability of a pipe-bridge collapse is

p(E1 ) = 1 × 3.60 10−3 + 2.0910−4 × 0.9964 ' 3.8010−3 (5)

Here, P (E1 |E2 ) = 1 (we are sure of a pipe-bridge collapse if the event E2 oc-
curs; see above), P (E2 ) is the probability to have more than 10 cm of ashes,
P (Ē2 ) = 1 − P (E2 ), and P (E1 |Ē2 ) is the pipe-bridge collapse probability
(P (ΘB ) in equation 15 of Appendix A). Since P (E1 ) is much larger than P (ΘB )
(see equation 15 of Appendix A), this means that there is a significant increase
of the industrial risk.
From the volcanic risk point of view, although a small accumulation of ash
may not lead to any building collapse, it could produce casualties through an
industrial accident, therefore increasing also the volcanic risk. In this case, the
increase of the volcanic risk is not significant because it already overwhelms all
the others. The new risk prioritization becomes

Volcanic risk: RVolc = 1.37 · 10 0

Seismic risk: RSeis = 5.00 · 10−3
Flooding risk: RFlood = 4.20 · 10−5
(6)
Industrial risk affected by volcanic activity RI&V = 3.20 · 10−6
Landslides risk: RLand = 6.00 · 10−7
Industrial risk: RInd = 1.83 · 10−7

5 Discussion and final remarks

A joint analysis and quantification of all the natural and anthropogenic risks
that can affect a territory (MRA) is a basic factor for the development of a sus-
tainable environment, land use planning, and risk mitigation strategies. To date,
risks assessment is generally achieved through independent analyses, adopting
non-uniform procedures and different time-space resolutions. In most of cases,
only qualitative estimates of the risk level are available. Such a strategy of
risks evaluation has some evident drawbacks, as for example the difficulty to
compare risks of different origins, and the implicit assumption of independent
and disjointed risk sources. In fact, the assumption of independence leads to
neglect possible interactions among threats and/or cascade effects, which may
underestimate the risk; in other words, the multi-risk index could be higher than
the simple aggregation of single risk indexes calculated considering each source
as independent from the others.
In this paper, we have put forward some basic principles for MRA that are
often overlooked. Specifically, we have discussed some general rules to facilitate
the risks comparison. Then, we have shown how interactions among threatening

11
events may affect the final risk estimation and how these interactions can be
included in MRA. In some cases interactions may have a negligible effect, or
they have been already accounted for through classical single risks assessment.
Anyway, we argue that in a significant number of cases, it is necessary to verify
if an interaction is relevant or not through a careful analysis. All these concepts
have been applied to a practical case-study of the Casalnuovo municipality that
was developed during the NaRaS project (e.g., Marzocchi et al, 2009). The goal
is not to present a deeply detailed and comprehensive description of the multi-
risk assessment for this area, but to highlight how the interaction among hazards
may cause a significant amplification and new ranking of the risks. Specifically,
in the long-term risk assessment for Casalnuovo, we have shown that the volcanic
risk overwhelms significantly all the others. In this first approach to consider
interactions between events, we have also shown that volcanic and in particular
industrial risks are underestimated if we do not consider the interaction among
them. In this specific case, while the industrial risk is severely underestimated if
we do not account for the interaction with the hazard posed by the Vesuvius, the
rank of risks is not drastically changed. Of course this result is case-dependent
and it easy to foresee other real cases where the interactions may lead to a
significant change into the ranking. For example, remaining in the volcano-
industry interaction, if the industrial site is a nuclear plant or a critical facility
for the economy of a large part of the area, an amplification of the industrial risk
of one order of magnitude could change significantly the ranking. Alternatively,
if the site is at large distances from the volcano, the overall ash fall hazard
will be significantly lower (i.e., the volcanic risk is much lower), but again few
centimeters may still produce major failures in industrial sites. In the case of
Casalnuovo, the municipality is about 12-13 km from the volcano; even though a
direct impact of the highly dangerous pyroclastic flows is very small, heavy ash-
fall loading remain highly probable. At larger distances, the probability of heavy
ash fall diminishes drastically, but the likelihood to have few centimeters of
accumulation is significant for a very wide area where the industrial risk may be
significantly underestimated. The recent impact of the Iceland Eyiafjallajokull
eruption clearly shows how a remote volcano may change completely, among
many other examples, the probability of a plane crash (if no mitigation actions
are taken). In general, it may be argued that interactions play a major role in
changing the ranking of the risks in most of the cases where one single risk does
not dominate the ranking.

acknowledgements
This research has been funded by the FP6 European project NaRaS and by the
FP7 European project MATRIX.

A Appendix: Analysis of risks induced by a sin-

gle source
The procedures for probabilistic hazard and risk assessment from different sources
are hereafter described. The discussion is focused on the preliminary single risk

12
assessments made for the case-study of Casalnuovo. For more details on specific
methodologies applied, see Marzocchi et al (2009).

A.1 Seismic Risk in Casalnuovo

The most recent evaluation of the probabilistic seismic hazard in the Casalnuovo
area, as provided by the National Seismic Hazard Map (http://www.INGV-
zonesismiche.mi.ingv.it), is represented in figure 4. The 90th percentile of the
peak ground acceleration in the next 50 years is in the range 0.15-0.175 g. The
same map has been also expressed in terms of seismic intensity (Gomez Caprera
et al, 2007), where Casalnuovo shows a 10% of probability to overcome an in-
tensity V III in the next 50 years. Since the process is time-independent, the
annual probability is pV III = 0.002.
In order to estimate the expected damage for such a kind of event, we have to
assume the vulnerability of the buildings, the level of damages, and how these
parameters reflect on the probability of having casualties. The relationship
between intensity (I) and degree of damage has been defined for some munici-
palities close to Casalnuovo by Zuccaro et al (2008) for the 2007 version of the
Italian Civil Protection Emergency Plan of Mount Vesuvius. Buildings have
been gathered in four categories (from A to D) according to their structural
characteristics, and five levels of increasing damage have been defined. The
probability of damage (y-axis) of an earthquake of intensity VIII as a function
of possible damage level (x-axis) for four different building typologies is reported
in figure 5. Casalnuovo has prevalent D type buildings, as the surrounding areas
(see for example, Figure 15 in Zuccaro et al (2008)).

Using the plot in figure 5, the expected damage corresponding to I = VIII

will be:
Damage 4 = 0.1%
(7)
Damage ≤ 3 = 99.9%
while earthquakes of I = VII will produce Damage level less than 4 in the Casal-
nuovo area.

It can be assumed that only Damage level 4 (partial collapses) or more can
lead to casualties of people living inside the buildings. Therefore, no risk for
people can be assumed for I=VII. Specifically, a person living in a building
suffering Damage 4, has 5% of probability to be killed (Zuccaro et al, 2008).
The risk for human life can be estimated assuming that the number of people
staying on average inside one building is given by the total number of citizens
(47940) divided by the number of edifices (3615), i.e., 13.26. This is a very
rough assumption, since we assume that all citizens will be inside a building at
the time of the earthquake. Actually, this number will be lower, but in this way
we accommodate partially the possible death of people on the street caused by
a building collapse. The risk for human life RSeis is composed by three fac-
tors: (1) the probability of occurrence of a I=VIII earthquake, (2) the average
number of edifices with expected damage level 4, and (3) the average number of
people killed inside a building with expected damage level 4. Such a probability
is:
RSeis = 0.002(0.001 × 3615)(0.05 × 13.26) = 0.005 (8)

13
A.2 Volcanic risk assessment: the ash fall hazard
Historical data indicate that ash fall is the main adverse event due to erup-
tions of Mt. Vesuvius in the Casalnuovo area. Due to volcano morphology
and the distance from the volcano, other possible adverse events, such as pyro-
clastic flows, lava flow, and lahars, have been very rare in this area. Actually,
Casalnuovo is out of the red zone of danger defined by the Mt. Vesuvius Civil
Protection Emergency Plan, but it is well within the zone under threat of ash
fall. However, we remark that the inclusion of such events do not pose further
conceptual problems. The probabilistic ash fall hazard assessment is obtained
through the use of Bayesian Event Tree (BET) (see figure 6). The full descrip-
tion of the method is described in Sect. 2.2 of Marzocchi et al (2009). The set
up of the conditional probabilities at each node were taken from Marzocchi et al
(2004).
The annual probability of Ash Fall in Casalnuovo can be estimated using the
probabilities at the nodes at the Bayesian event tree, as indicated in Marzocchi
et al (2004, 2009):
X h i h i h i h i
(VEI) (TF) (Casalnuovo) (threshold)
[π] = 12 [Θ1 ] [Θ2 ] [Θ3 ] Θ5 Θ6,VEI Θ7,VEI Θ8,VEI
VEI
(9)
where the summation is for eruptions with Volcanic Explosion Index (VEI =
III, IV, and V+), and the factor 12 transforms the month probability (node
1) into annual probability. We stress that this approximation holds when the
probabilities are small as in the present case. The multiplication is performed
through 1000 values randomly selected for each node. At the end, we have four
distributions, each one relative to a specific thickness. These distributions are
reported in Figs. 7, 8, 9, and 10.

The Figs. 7 to 10 represent the exceedance probabilities calculated in a time

frame of 1 month. The annual probability for any specific thickness can be
approximated as:

P10 = P(10<thickness≤20) = 12 ∗ (0.0003 − 0.0002) = 0.0012

P20 = P(20<thickness≤40) = 12 ∗ (0.0002 − 0.0001) = 0.0012
(10)
P40 = P(40<thickness≤60) = 12 ∗ (0.0001 − 0.00007) = 0.00036
P60 = P(60>thickness) = 12 ∗ (0.00007) = 0.00084

In order to retrieve the individual risk we need to incorporate the vulnerability.

Using a density of 900 Kg/m3 for dry ash deposits and the probability of collapse
as a function of the load reported by Zuccaro et al (2008), the probability of
collapse Pc for different thickness of the ash deposits on the roofs of a building
are:
Pc (10 cm) = 0.00
Pc (20 cm) = 0.01
(11)
Pc (40 cm) = 0.15
Pc (60 cm) = 0.60
The collapse from ash fall can be compared to a Damage 4 for seismic risk. If we
assume, that the probability for human beings living inside the building to be
killed by a collapse is 0.05 (the same as in the case of earthquakes), the annual

14
volcanic risk for human life is:
RVolc = 0.0012 (0.01 × 3615) (0.05 × 13.26) +
= 0.00036 (0.15 × 3615) (0.05 × 13.26) +
(12)
= 0.00084 (0.60 × 3615) (0.05 × 13.26)
= 1.37

A.3 Hydrogeological risk assessment

The geological nature, the topography, the climatic conditions and particularly
the carefree use of land have made Campania one of the Italian regions with
the highest hydrogeological risk. Exposure to hydrogeological risk is therefore
a problem of great social relevance, both for the number of potential casualties
and for the high exposed value, consisting of homes, industries, infrastructures,
life-lines and cultural heritage. Possible change of scenarios in relation to cli-
mate change are of further concern for the near future.

In Italy, the identification, delimitation and outlining of the boundaries of the

areas that present flood and landslide probabilistic hazards and risks are under
the competence of the River Basin Authorities, established by the Law 183/89
and identified as the responsible authority for the basin planning. Each River
Basin Authority used different criteria for the definition of the areas at landslide
or flooding risk. Figure 11 represents the flood hazard map for the area and is
the result of the harmonization for the Campania territory of the information of
hydrogeological asset plans given by each local River Basin Authority of Cam-
pania region (e.g. Autoritá di Bacino Nord-Occidentale della Campania, 2002).

The Casalnuovo territory does not have significant landslide probabilistic haz-
ard. Flooding risk is moderate (R1 o R2) in some localized areas of the munici-
pality due to a hazard of level P1 (see Figs. 11 and 12). The annual probability
of inhabitants in the Casalnuovo municipality to be killed by a hydrogeologi-
cal event has been evaluated using the AVI catalogue and average number of
inhabitants. The AVI catalogue reports the number of casualties in Campania
Region for each event and the historical sequence of event. The result has been
downscaled to the Casalnuovo number of residents. The results are:
Pflooding = 1.4 · 10−7
(13)
Plandslide = 2.0 · 10−8

From these numbers the annual risk for flooding (Rflooding ) and for landslide
(Rlandslide ) can be calculated. Looking at similar past events in Campania
Region, we can assume that a landslide impacts few tens of inhabitants, while
a flood can impact one order of magnitude larger. Therefore, we can estimate:

Rflooding = 1.4 · 10−7 × 300 = 4.2 · 10−5

(14)
Rlandslide = 2.0 · 10−8 × 30 = 6.0 · 10−7

A.4 Industrial risk

The map of the Casalnuovo municipality, reported in figure 3, shows the local-
ization of the industrial risk sources (A: industrial site for LPG storage) that

15
can generate adverse events in the area. The most likely risk source is the pipe-
bridge collapse. In the classical risk assessment, the pipe-bridge can undergo
a catastrophic collapse due to the macroscopic growth of microscopic material
defects. Partial collapses and the subsequent breakage of the whole pipe-bridge
has been taken into account in Sect. 2.4 in Marzocchi et al (2009). This infor-
mation will be used to evaluate the conditioned probability to have casualties
death due to the exposure to the heat flux generated by the fire of the LPG
released by the pipelines after its collapse.
In the following the probability of each section of the event tree is reported (for
details see Marzocchi et al (2009)):

P (ΘB ) = 2.09 · 10−4

P (ΘE ) = 6.50 · 10−2
P (ΘF ) = 1.25 · 10−1
(15)
P (ΘG ) = 5.00 · 10−1
P (ΘH ) = 2.16 · 10−1
P (ΘI ) = [1 : 0.01]

where: P (ΘB ) is the probability of release of a specific substance due to a pipe-

bridge collapse; P (ΘE ) is the probability that the event occurs under the form
of a specific phenomenon (e.g. toxic release, explosion fire, etc.); P (ΘF ) is the
probability that a specific phenomenon goes toward a specific direction; P (ΘG )
is the probability that the phenomenon has a certain intensity distribution in
an area where the consequences are to be investigated; P (ΘH ) is the probability
of the presence of people where the phenomenon occurs, and P (ΘI ) is the prob-
ability of death of people in the area, due to the exposure to the phenomena
generated by the event. The individual risk is then:

Y
I
ri = ΘJ,i (16)
J=A

By applying the product of the probability values for each event, the following
individual risk index range is obtained for the death of population in the risk
area (117m from the risk source localization):

1.83 · 10−9 < IR < 1.83 · 10−7 (17)

Assuming that 10 people will be in that area of influence (117 m from the
explosion), we can estimate the direct annual industrial risk as

1.83 · 10−8 < RInd < 1.83 · 10−6 (18)

In our calculation, and for the sake of example, we consider the average of this
distribution, i.e., RInd = 1.83 · 10−7

16
Table 1: Risk source, probabilistic hazard, vulnerability, and exposed value (in
human life loss per year) for the Casalnuovo municipality case-study
Risk Source Hazard Vulnerability Exposed value
(H) (V) (L)
Seismic 0.0007 (0.001 × 3615) (0.05 × 13.26)
Volcanic: Ash fall
10 < thickness ≤ 20 0.0012 (0.00 × 3615) (0.05 × 13.26)
20 < thickness ≤ 40 0.0012 (0.01 × 3615) (0.05 × 13.26)
40 < thickness ≤ 60 0.00036 (0.15 × 3615) (0.05 × 13.26)
thickness ≥ 60 0.00084 (0.60 × 3615) (0.05 × 13.26)
Hydrogeological
Flooding (H) × (V) = 1.4 · 10−7 300
Landslide (H) × (V) = 2.0 · 10−8 30
Industrial 1.83 · 10−9 < H×V < 1.83 · 10−7 10

17
 Hazard source Evaluation of elements
identification at risk: Exposed values
Probabilistic model
characterizing the
process

Vulnerability
Hazard assessment
assessment

Loss estimation and

Risk assessment
(for a single hazard)

Figure 1: Schematic description of the general procedure followed for the single
risk evaluation

18
 (a) Definition of:
- space/time assessment window
(target area, time window)
- Metric for expected loss
(e.g. economic, fatalities, etc.)

simultaneous
(b) occurrence triggering or
Source 1 (independent) Source 2 cascade effects Source n
Hazard Source
(S1) (S2) ... (Sn)
identification:
(e.g. volcano,
landslide,
meteorological Source 3
events, etc.) (S3)

Hazard Hazard Hazard Hazard

(c)
Single and multi-
assessment assessment assessment ... assessment
Hazard H1 = f1(S1) H2 = f2(S2) H3 = f3(S3) Hn = fn(Sn)
assessment:

e.g. rate, pathway, cascade hazard

assessment
intensity measure, H(1,2) = f1 * f2 H’3 = f’(S3|S2)
etc.

(d)
Vulnerability Vulnerability Vulnerability Vulnerability
Vulnerability of
assessment assessment assessment ... assessment
exposed elements
V1 = g1(H1) V2 = g2(H2) V3 = g3(H3) Vn = gn(Hn)
(receptors):

e.g. people, Vulnerability

assessment Vulnerability
buildings, assessment
environment V’1 = g’1[H(1U
2)]
V’3 = g’3(H’3)
vulnerabilities V’2 = g’2[H(1U
2)]

(e)
Consequences: Risk Risk
assessment
assessment (for source 3,
(for source 2)
risk assessment in triggered by source 2)
Risk Risk
terms of, e.g., loss of assessment Multi-risk ... assessment
life, economic losses, (for source 1) (for source n)
environmental Risk Risk
degradation,etc. assessment assessment
(for source 1 and 2) (for source 3)

(Ranking or integration in a
single risk index

Figure 2: Schematic description of the general procedure followed for the multi-
risk evaluation: (a) the first step is the definition of the space/time assessment
window, and metric for expected loss; (b) hazard sources or triggering hazardous
events are identified for a given interest area; different sources may interact or
present cascades of events; (c) from each hazard or multi-hazard source, the
process of hazard assessment is performed as a function f() of the stochastic
characteristics of the source(s), hazard intensity, and the diffusion processes
(i.e. the pathway between the source and the receptor); (d) assessment of the
vulnerability of the different exposed elements (receptors), which is a function
g() of the hazard intensity and may be modified in the case of interactions or
cascading of hazards; (e) risk assessment, in terms of expected loss estimation
for the different scenarios considered: multi-risk assessment.

19
Figure 3: The Casalnuovo municipality, the main anthropogenic risk sources
and distribution of different buildings and infrastructures.

Figure 4: Probabilistic seismic hazard in the Casalnuovo area taken from the
National Seismic hazard map (amax are in units of g).

20
 A
B
0.6 C
D
Probability of damage

0.4

0.2

0
0 1 2 3 4 5
Damage level

Figure 5: Probability of damage as a function of damage level for an earthquake

of I = V III. The curves are relative to four typologies of buildings Zuccaro
et al (2008).

21
 Loop n

P(! A-IND
) P(! B-IND
) P(! C-IND
) P(! D-IND
) P(! E-IND
) P(! F-IND
) P(! G-IND
) P(! H-IND
) P(! I-IND
)
Magnitude
Plant Substance Source RAI (incident) Incident Sector People Vulnerability
distribution

Clone Clone Clone

Clone Tank Vapour cloud N (D1, I1) Clone Residents Clone

Clone explosion
RAI=5
Pipe Clone Bleve Clone Clone
NE Clone
Clone
RAI=4 Clone
Clone
Tanker Fireball E
Lorry Clone Clone Clone

Loss of GPL
containment Cylinder RAI=3 Jet Fire SE (Di, Ii) Workers
Flash S
Clone Clone
Compressor Clone
RAI=2 Pool
Clone SW
Fire
Clone
Clone Pool Clone W Clone Clone

22
Clone
Clone Clone NW Clone

Pipebridge Toxic
(DN, In) Tourists
Clone Clone Clone
HAZARD Clone
IDENTIFICATION EXPOSURE

Figure 6: Event Tree scheme for probabilistic volcanic hazard assessment

Figure 7: Probabilistic tephra fall hazard map for Mount Vesuvius. The dialog
box reports the 10th, 50th, and 90th percentiles of the annual probability of
exceedance of a thickness of 10 cm.

Figure 8: The same as Figure 7, but relative to a thickness of 20 cm.

23
Figure 9: The same as Figure 7, but relative to a thickness of 40 cm.

Figure 10: The same as Figure 7, but relative to a thickness of 60 cm.

24
Figure 11: Map of the flood hazard in Campania.

Figure 12: Map of the flood hazard in Casalnuovo.

25
References
Autoritá di Bacino Nord-Occidentale della Campania (2002) Relazione idraulica
del piano di assetto idrogelogico dell’autoritá di bacino nord-occidentale della
campania. Tech. rep., Autoritá di Bacino Nord-Occidentale della Campania
Baratta A, Zuccaro G, Binetti A (2004) Strength capacity of a no-tension portal
arch-frame under combined seismic and ash loads. J Volcanol Geotherm Res
133(1-4):369–376, DOI DOI: 10.1016/S0377-0273(03)00408-6
Blong R (2003) A new damage index. Nat Hazards 30(1):1–23
Budnitz RJ, Apostolakis G, Boore DM, Cluff LS, Coppersmith KJ, Cornell CA,
Morris PA (1997) Recommendations for probabilistic seismic hazard analysis:
Guidance on uncertainty and use of experts. Tech. Rep. NUREG/CR-6372
UCRL-ID- 122160, Senior Seismic Hazard Analysis Committee (SSHAC)
BZS (1995) KATANOS Katastrophen und Notlagen in der Schweiz - Eine ver-
gleichende Übersicht. Tech. rep., BZS (Bundesamt für Zivilschutz), Bern
Cooke RM (1991) Experts in Uncertainty. Oxford University Press
Cornell CA, Krawinkler H (Spring 2000) Progress and challenges in
seismic performance assessmentcornell. PEER Center News 3(2), URL
http://peer.berkeley.edu/news/2000spring/index.html
Der Kiureghian A (2005) Non-ergodicity and peers framework formula. Earth-
quake Enging Struct Dyn 34:1643–1652
Durham K (2003) Treating the risks in Cairns. Nat Hazards 30(2):251–261
Ferrier N, Haque CE (2003) Hazards risk assessment methodology for emer-
gency managers: A standardized framework for application. Nat Hazards
28(2/3):271–290
Gomez Caprera A, Meletti C, Rebez A, Stucchi M (2007) Mappe di peri-
colosita’ sismica in termini di intensita’ macrosismica utilizzando lo stesso
impianto metodologico di mps04. Progetto DPC-INGV S1, Deliverable D7,
http://esse1.mi.ingv.it/d7.html
Granger K, Jones T, Leiba M, Scott G (1999) Community Risks in Cairns: A
Multi-Hazard Risk Assessment. Australian Geological Survey Organisation,
Canberra
Grezio A, Marzocchi W, Sandri L, Gasparini P (2010) A bayesian procedure for
probabilistic tsunami hazard assessment. Nat Hazards DOI 10.1007/s11069-
009-9418-8
Grunthal G, Thieken AH, Schwarz J, Radtke KS, Smolka A, Merz B (2006)
Comparative risk assessments for the City of Cologne - Storms, Floods, Earth-
quakes. Nat Hazards 38:21–44
Howard RA (1980) On making life and death decisions. In: Richard J, Schwing
C, Albers WA (eds) Societal risk assessment: how safe is safe enough?, Plenum
Press

26
 Marzocchi W, Woo G (2007) Probabilistic eruption forecasting and the call for
an evacuation. Geophys Res Lett 34(L22310)
Marzocchi W, Woo G (2009) Principles of volcanic risk metrics: theory and
the case study of Mt. Vesuvius and Campi Flegrei (Italy). J Geophys Res
114(B03213)
Marzocchi W, Sandri L, Gasparini P, Newhall C, Boschi E (2004) Quantifying
probabilities of volcanic events: The example of volcanic hazard at Mount
Vesuvius. J Geophys Res 109(B11201), DOI 10.1029/2004JB003155
Marzocchi W, Sandri L, Selva J (2008) BET EF: a probabilistic tool for long-
and short-term eruption forecasting. Bull Volcanol 70(5):623–632
Marzocchi W, Mastellone ML, Di Ruocco A, Novelli P, Romeo E, Gasparini
P (2009) Principles of multi-risk assessment: Interaction amongst natural
and man-induced risks (Project report). Office for Official Publications of the
European Communities, Luxembourg
Marzocchi W, Sandri L, Selva J (2010) BET VH: a probabilistic tool for long-
term volcanic hazard assessment. Bull Volcanol DOI 10.1007/s00445-010-
0357-8
Newhall CG, Hoblitt RP (2002) Constructing event trees for volcanic crises.
Bull Volcanol 64:3–20, DOI 10.1007/s004450100173
Schmidt J, Matcham I, Reese S, King A, Bell R, Henderson R, Smart G, Cousins
J, Smith W, Heron D (2011) Quantitative multi-risk analysis for natural
hazards: a framework for multi-risk modelling. Natural Hazards 58:1169–
1192, URL http://dx.doi.org/10.1007/s11069-011-9721-z, 10.1007/s11069-
011-9721-z
Selva J, Costa A, Marzocchi W, Sandri L (2010) BET VH: exploring the in-
fluence of natural uncertainties on long-term hazard from tephra fallout at
campi flegrei (italy). Bull Volcanol DOI 10.1007/s00445-010-0358-7
Spence R, Kelman I, Brown A, Toyos G, Purser D, Baxter P (2007) Residential
building and occupant vulnerability to pyroclastic density currents in ex-
plosive eruptions. Natural Hazards and Earth System Science 7(2):219–230,
DOI 10.5194/nhess-7-219-2007
UNDRO (1977) Composite vulnerability analysis - A methodology and case
study of the Metro Manila area. Tech. rep., Office of the U.N. Disaster Relief
Coordinator, Geneva
Van Stiphout T, Wiener S, Marzocchi W (2010) Are short-term evacuation
warranted? Case of the 2009 L’Aquila earthquake. Geophys Res Lett 37,
DOI 10.1029/2009GL042352
Van Westen CJ, Montoya L, Boerboom L (2002) Multi-Hazard risk assessment
using GIS in urban areas: A case study for the city of Turrialba, Costa Rica.
In: Proc. Regional workshop on Best Practise in Disaster Mitigation, Bali,
pp 120–136
Zuccaro G, Cacace F, Spence RJS, Baxter PJ (2008) Impact of explosive erup-
tion scenarios at vesuvius. J Volcanol Geotherm Res 178(3):416–453

27

View publication stats