Probabilistic E-tools for HazardAssessment and Risk Management

Stefania Bartolini, Joan Martí, Rosa Sobradeloand Laura Becerril

AbstractThe impact of a natural event can significantly affect human life and theenvironment.Although fascinating, a volcanic eruptioncreates similar or evengreater problems than more frequent natural events due to its multi-hazardnature and the intensity and extent of its potential impact. It is possible to livenear a volcanic area and take advantage of the benefits that volcanoes offer, butit is also important to be aware of the existing threats and to know how tominimise risks. In this chapter, we present an integrated approach usinge-tools for assessing volcanic hazard and risk management. These tools havebeen especially designed to assess andmanage volcanic risk, to evaluate long-and short-term volcanic hazards, to conduct vulnerability analysis, and toassist decision-makers during the management of a volcanic crisis. Themethodology proposed here can be implemented before an emergency inorder to identify optimum mitigating actions and how these may have to beadapted as new information is obtained. These tools also allow us identifyingthemost appropriate probabilistic and statistical techniques for volcanologicaldata analysis and treatment in the context of quantitative hazard and riskassessments. Understanding volcanic unrest, forecasting volcanic eruptions,and predicting themost probable scenarios, all imply a high degree of inherentuncertainty, which needs to be quantified and clearly explained whentransmitting scientific information to decision-makers.

ResumenEl impacto de un evento natural puede afectar significativamente la vidahumana y el medio ambiente. Aunque fascinante, una erupción volcánica

S. Bartolini (&) � J. Martí � L. BecerrilGroup of Volcanology, SIMGEO (UB-CSIC),Institute of Earth Sciences Jaume Almera,ICTJA-CSIC, Lluís Solé I Sabarís S/N, Barcelona08028, Spaine-mail: sbartolini.1984@gmail.com

J. Martíe-mail: joan.marti@ictja.csic.es

L. Becerrile-mail: lbecerril@ictja.csic.es

R. SobradeloWRN Earth Hub Leader, Analytics Technology &Willis Research Network, London, UKe-mail: rosa.sobradelo@willistowerswatson.com

Advs in Volcanology (2019) 47–61DOI 10.1007/11157_2017_14© The Author(s) 2017Published Online: 15 July 2017

puede generar problemas parecidos o incluso mayores que otros eventosnaturales más frecuentes, ya que se trata de un fenómeno multipeligro ycon un impacto potencial de gran intensidad y magnitud. Vivir cerca deuna zona volcánica es posible considerando sus innumerables beneficios,pero sin embargo debemos ser conscientes de la amenaza existente ytomar las medidas oportunas para minimizar el riesgo. En este capítulo, sepresenta un enfoque integrado que utiliza herramientas para la evaluaciónde la peligrosidad y del riesgo volcánico, el análisis de vulnerabilidad, ypara ayudar en la gestión de una crisis volcánica. La metodología aquípropuesta puede ser implementada antes de una emergencia con el fin dedeterminar las medidas óptimas de mitigación y como pueden adaptarse amedida que se obtiene nueva información. Además, estas herramientas sebasan en las técnicas probabilísticas y estadísticas más adecuadas para elanálisis de datos vulcanológicos y su implementación en el contexto de lasevaluaciones cuantitativas del peligro y riesgo. La interpretación delunrest, la predicción de las erupciones volcánicas, y la identificación delos escenarios más probables, implican un alto grado de incertidumbre,que debe ser cuantificado y claramente explicado para transmitircorrectamente la información científica a los gestores de las crisisvolcánicas.

KeywordsVolcanic risk � Volcanic hazard � Vulnerability � Risk management �Decision-making � E-tools

Introduction

One of the most important tasks in modern vol-canology is to manage volcanic risk and, conse-quently, to minimise it. Forecasting volcaniceruptions and predicting the most probable sce-narios are tasks that are subject to high degrees ofuncertainty but which need to be quantified andclearly explained when transmitting scientificinformation to decision-makers. Assessing erup-tion risk scenarios in probabilistic ways hasbecome one of the main challenges tackled bymodern volcanology (Newhall and Hoblitt 2002;Marzocchi et al. 2004; Aspinall 2006; Neri et al.2008; Sobradelo et al. 2014).

The volcanic management cycle consists offour phases (Sobradelo et al. 2015): (i) the

pre-unrest phase that includes long-term assess-ment, hazard and risk mapping, and estimationregarding expected scenarios, volcano monitor-ing, and emergency planning; (ii) the unrestphase, which includes short-term assessment,alert, communication, and information proce-dures, the implementation of emergency mea-sures, and the interpretation of eruptionprecursors; (iii) the volcanic event itself, repre-sented by a major change in the state of thevolcano; decisions can be revised as new infor-mation is obtained and the evolution of the vol-canic event is updated; (iv) the post-event phasecharacterized by rescue and recovery.

Previous studies, focusing attention on thefirst phase of volcanic crisis management, havedeveloped different methodologies for evaluating

48 S. Bartolini et al.

hazards. Most of these studies are based on theuse of simulation models and GeographicInformation Systems (GIS) that enable volcanichazards such as lava flows, pyroclastic densitycurrents (PDCs), and ash fallout to be modeledand visualised (Felpeto et al. 2007; Toyos et al.2007; Cappello et al. 2012; Martí et al. 2012;Becerril et al. 2014; Bartolini et al. 2014a, b).The use of spatio-temporal data and GIS hasnow become an essential part of integratedapproaches to disaster risk management in lightof the development of new analysis/modellingtechniques. These spatial information systemsare used for storage, situation analysis, mod-elling, and visualisation (Twigg 2004). Nor-mally, these studies present a systematicapproach based on the estimation of spatialprobability, that is, a susceptibility analysis(Martí and Felpeto 2010), a temporal analysisbased on Bayesian inference (Marzocchi et al.2004; Sobradelo et al. 2014), or the evaluation ofhazards (Felpeto et al. 2007; Bartolini et al.2014a, b; Becerril et al. 2014) and vulnerability(Marti et al. 2008; Scaini et al. 2014). Thesestudies have been applied in a number of vol-canic areas such as Etna, Sicily (Cappello et al.2013), Tenerife, Spain (Martí et al. 2012), Peru(Sandri et al. 2014), the island of El Hierro,Spain (Becerril et al. 2014), and DeceptionIsland, Antarctica (Bartolini et al. 2014a). Otherprocedures have been employed to assess vol-canic hazards in Campi Flegrei, Italy (Lirer et al.2001), Furnas (São Miguel, Azores), Vesuvius,Italy (Chester et al. 2002), and Auckland, NewZealand (Sandri et al. 2012).

These studies underline the fact that scientistsare aware of the relationship between volcanichazards and socio-economic impacts and are inthe process of developing new approaches andmodels to assess its importance. One positiveaspect is that, as a result, there is now a choice offreely available models; on the other hand, thesemodels are not integrated into a single platformand have been developed in a variety of differentprogramming languages.

Despite the fact that these tools have beencreated for application in real situations and havebeen successfully tested in different volcanic

areas and retrospectively for a number of vol-canic crisis (Martí et al. 2012; Sobradelo et al.2014; Bartolini et al. 2014a, b; Becerril et al.2014; Scaini et al. 2014), to date they only existas academic exercises. To convert them intopractical tools ready to be used by Civil Protec-tion managers and decision-makers, they must bechecked and tested, and then adapted to the realneeds of end users.

Probabilities are still the best outcome ofscientific forecasting. However, they are noteasily understood. Understanding the potentialevolution of a volcanic crisis is crucial fordesigning effective mitigation strategies. One ofthe main issues when managing a volcanic crisisis how to make scientific information under-standable for decision-makers and Civil Protec-tion managers. Thus, we need quantitativerisk-based methods for decision-making underconditions of uncertainty that can be developedand applied to volcanology. In order to resolvethis problem and to take a step forward in min-imising risks, we have defined an integratedapproach using user-friendly e-tools, which canbe run on personal computers. They are specifi-cally useful for long- and short-term hazardassessment, vulnerability analysis, decision-making, and volcanic risk management. In thischapter, we describe the e-tools designed tomanage and to minimise volcanic risk.

Volcanic Risk: Hazard, Vulnerability,and Value

In general, risk is defined as the probability orlikely magnitude of a loss (Blong 2000). Involcanic risk assessment, risk depends on theadverse effects of volcanic hazards and can bedefined as the product of three main factors:volcanic hazard, vulnerability to those hazards,and the value of what is at risk.

Volcanic hazard is defined as the probabilityof any particular area being affected by adestructive volcanic event within a given periodof time (Blong 2000). The quantification andevaluation of the volcanic hazard allow usdetermining which areas will be affected by a

Probabilistic E-tools for Hazard Assessment … 49

given volcanic event, and to design appropriateemergency plans and territorial planning.

A vulnerability assessment uses indicators toquantify the physical vulnerability of the elementsof each sub-system (buildings, transportationsystem, urban services, and population) and is ameasurement of the proportion of the value likelyto be lost as a result of a given event (Blong 2000).In fact, the values of elements-at-risk that can bedirectly or indirectly damaged by a given hazardwill vary. Each hazardous phenomenon affectselements and infrastructures in different ways interms of their specific physical vulnerability; thisin turn will be depend on their number (of build-ings, people, etc.), monetary value, surface area,and the importance of the elements-at-risk. Indi-cators are used to determine the specific physicalvulnerability of each hazardous phenomenon andsub-system (Scaini et al. 2014).

The value is the number of human lives atstake, together with the capital value (land,

buildings, etc.) and productive capacity (facto-ries, power plants, highways, etc.) exposed to thedestructive events (Blong 2000).

Risk management is a complex process(Fig. 1) since different steps are necessary forevaluating and minimising risk. It can be thoughtof as the sum of risk assessment—which includesrisk analysis and risk evaluation—and risk con-trol. Risk analysis aims to improve preventiontools through the collection and acquisition ofdata on hazards and risks, and then to dissemi-nate it in the form of maps and scenarios (Thierryet al. 2015). This phase is characterised by fivedifferent steps: hazard identification, hazardassessment, elements-at-risk/exposure analysis,vulnerability assessment, and risk estimation(Van Westen 2013). In particular, it is importantto distinguish between long- and short-termhazard assessments, which will vary accordingto the expected period of time over which theprocess will display significant variations.

Fig. 1 Risk management schematic: steps for evaluating and minimising risk

50 S. Bartolini et al.

Long-term assessment is based on historical andgeological data, as well as theoretical models,and refers to the time window available—duringwhich time the volcanic system shows no signsof unrest—before an unrest episode occurs. Onthe other hand, short-term assessment refers tothe unrest phase, and complementary informationderived from the combination of a long-termanalysis with real-time monitoring data is neededto update the status of the volcanic hazard (Blong2000). Short-term evaluation helps forecastwhere and when the eruption may take place andthe most likely eruptive scenarios.

Once the long-term risk analysis is computed, itis then possible to adopt mitigation measures suchas land-use planning and emergency preparationsto reduce the risk. In addition, this long-termanalysis will help manage the volcanic crisis as itwill constitute the basis for the short-term analysisand, combined with a cost-benefit analysis, willassist in correct decision-making (Sobradelo et al.2015). To evaluate the total risk related to a par-ticular volcanic eruption we have to repeat theevaluation of the vulnerability and the cost-benefitanalysis (risk evaluation) for each possible hazardscenario and then sum the results. This will allowus to estimate the impact and the economic lossesthat will affect society and the environment, and toidentify a range of risk management alternatives.

Finally, the second part of risk management isrisk control, which consists of the decision-making process involved in managing riskswhose aim is to improve crisis managementcapabilities and implement risk-mitigation mea-sures using the results of risk assessment as aninput (Western 2013). During this phase mea-sures must be adopted for reducing vulnerability(people and infrastructure) and developingrecovery and resilience capacities after an eventhas taken place.

E-tools for Volcanic Hazard and RiskManagement

In this section, we present different e-tools thathave been specifically designed to assess andmanage volcanic risk (Fig. 2). The objective is tocombine freely available models to produce anew approach for minimising and managingvolcanic risk. These e-tools are based on theassumption that the best way to show howprobabilities work is to use the possible scenariosand outcomes of volcanic unrest (an increase involcanic activity that may or may not precede avolcanic eruption) to design an integrated modelthat can act as a descriptor of scenarios. Theeffectiveness of these e-tools has been analysed

Fig. 2 E-tools for assessing and managing volcanic risk that allow to evaluate the possible hazards that could affect avolcanic area and develop appropriate hazard and risk maps

Probabilistic E-tools for Hazard Assessment … 51

in different volcanic areas showing pros andcontras of these approaches, such as in the caseof El Hierro Island in Canary Islands (Becerrilet al. 2014; Sobradelo et al. 2015), Tenerife inCanary Islands (Scaini et al. 2014), DeceptionIsland in Antarctica (Bartolini et al. 2014a), andLa Garrotxa Volcanic Field in Spain (Bartoliniet al. 2014b). User manuals are available athttps://volcanbox.wordpress.com/ and http://www.vetools.eu.

These e-tools are designed to be implementedbefore an emergency in order to identify (i) theoptimum mitigating actions and (ii) the mostappropriate probabilistic and statistical tech-niques for volcanological data analysis andtreatment in the context of quantitative hazardand risk assessment, and (iii) how appropriateresponses may change as new information isobtained. They constitute different steps, asshown in Fig. 3, in hazard assessment and riskmanagement.

Storing Data

The results of our analyses are as important asthe input parameters that are used. Given thenature, variety, and availability of the data weneed to handle, one of the most importantaspects of risk assessment is the managementand exchange of information (De la Cruz-Reyna1996). The quality of the data will determine theevaluation of the volcanic risk, which is anessential part of risk-based decision-making inland-use planning and emergency management.Some of the most relevant issues include howand where to store the data, in which formatshould they be made available, and how tofacilitate its use and exchange. Thus, it isessential to have an appropriate database that isspecifically adapted to the task of evaluating andmanaging volcanic risk. For that reason, wedesigned VERDI (Volcanic managEment RiskDatabase desIgn) (Bartolini et al. 2014c), a

Fig. 3 Steps in volcanic risk assessment applying e-tools to be implemented before an emergency

52 S. Bartolini et al.

geodatabase with an appropriate architecture forvolcanic risk assessment and management thatstores the data from which we will extract theinput parameters to run our e-tools. VERDI is aspatial database structure (Fig. 4) that allowsdifferent types of data, including geological,volcanological, meteorological, monitoring, andsocio-economic information, to be manipulated,organised, and managed. The data contained inthis database are the basis for applying the

probabilistic models that are the first step in ourrisk analysis.

Hazard E-tools

Volcanic hazard assessment consists of simulat-ing eruptive scenarios for use in risk-baseddecision-making, land-use planning, and emer-gency management. They must necessarily be

Fig. 4 The design of the VERDI database (after Bartolini et al. 2014c)

Probabilistic E-tools for Hazard Assessment … 53

based on a good knowledge of the past eruptivehistory of the volcano or volcanic area, and willreveal how volcanoes erupted in the past, therebyproviding clues to how they will erupt in thefuture. The first step in the quantitative assess-ment of volcanic hazards is the spatial probabilityof occurrence of a hazard, i.e. where the nexteruption may take place and its extent. Thisanalysis is based on the development of suscep-tibility maps (Martí and Felpeto 2010), that is,the spatial probability of a future vent openinggiven the past eruptive activity of a volcano andthe simulation of possible eruptive scenarios.Another important task is to investigate thetemporal probability, in other words, when thenext eruption will occur in the future and the typeof scenarios that are most likely to be involved.Thus, an evaluation of volcanic hazard enablesus to infer where and when the next eruption maytake place and its magnitude.

Spatial AnalysisSusceptibility analysis is the evaluation of thespatial distribution of future vent openings(Fig. 5a). This challenging issue is generallytackled using probabilistic methods that use thecalculation of a kernel function at each datalocation to estimate probability density functions(PDFs). Commonly, a Gaussian kernel, describ-ing a normal distribution, is used to estimatelocal event densities in volcanic fields, whichwill give the intensity of a new vent opening.This method is based on the distance from nearbyvolcanic structures and a smoothing parameter,also known as bandwidth. This factor is the mostimportant parameter in the kernel function andrepresents the degree of randomness in the dis-tribution of past events.

QVAST (QGIS for VolcAnic SuscepTibility),developed by Bartolini et al. (2013), is a new tooldesigned to generate user-friendly quantitativeassessments of volcanic susceptibility (e.g. theprobability of hosting a new eruptive vent).QVAST allows an appropriate method for eval-uating the bandwidth for the kernel function to beselected on the basis of input parameters and theshapefile geometry, and can also evaluate thePDF with the Gaussian kernel. When different

input data sets are available for the area, the totalsusceptibility map is obtained by assigning dif-ferent weights to each of the PDFs, which arethen combined via a weighted sum and modeledin a non-homogeneous Poisson process. Thise-tool has been used to evaluate susceptibility onthe island of El Hierro (Canary Islands) (Becerrilet al. 2014) and on Deception Island (Antarctica)(Bartolini et al. 2014a).

When monitoring data generated during anunrest phase are available, the QVAST e-tool canalso be used to update the susceptibility map. Infact, seismicity and surface deformation are goodindicators of magma movement and during vol-canic unrest variations in shallow volcano-tectonic and long-period seismicity, as well asground deformation, are observed as the magmamigrates within the volcanic system (Martí et al.2013). Thus, QVAST is a highly useful tool thatcan be applied to both long- and short-termevaluations.

Temporal AnalysisHASSET (Hazard Assessment Event Tree),developed by Sobradelo et al. (2014), is aprobability tool built on an event tree structurethat uses Bayesian inference to estimate theprobability of occurrence of a future volcanicscenario. It also evaluates the most relevantsources of uncertainty in the correspondingvolcanic system. Event tree structures (Newhalland Hoblitt 2002) constitute one of the mostuseful and necessary tools in modern volcanol-ogy for assessing the volcanic hazard of futureeruptive scenarios. An event tree is a graphicrepresentation of events in the form of nodesand branches. It evaluates the most relevantsources of uncertainty when estimating theprobability of occurrence of a future volcanicevent.

The objective of this e-tool is to outline allrelevant possible outcomes of volcanic unrest atprogressively greater detail and to assess thehazard of each scenario by estimating its proba-bility of occurrence within a future time interval.Each node of the event tree represents a step andcontains a set of possible branches (the outcomesfor that particular category). The nodes are

54 S. Bartolini et al.

alternative steps from a general prior event, state,or condition that move towards increasinglyspecific subsequent events and a final outcome.HASSET (Fig. 5b) uses this event tree structureto make estimations of the probabilities for eachpossibility (branches and nodes) using a statisti-cal methodology known as Bayesian Inference(Newhall and Hoblitt 2002; Marzocchi et al.2004; Sobradelo et al. 2014). In particular, andbased on comparisons with previous event treesfor volcanic eruptions, HASSET accounts for thepossibility of (i) flank eruptions (as opposed to

only central eruptions), (ii) geothermal or tec-tonic unrest (as opposed to only magmaticunrest), and (iii) felsic or mafic lava composition,as well as (iv) certain volcanic hazards as pos-sible outcomes of an eruption, and (v) the dis-tance reached by each hazard.

A user-friendly interface guides the userthrough all steps and helps

– enter all the data needed for the analysis;– compute the estimated probability for each

branch in the event tree;

Fig. 5 Methodological approach for obtaining a qualitative hazard map: its application to Deception Island (afterBartolini et al. 2014a)

Probabilistic E-tools for Hazard Assessment … 55

– compute the total estimated probability andcompare up to five different scenarios.

This tool has been used for determining theeruption probability on El Teide (Tenerife,Spain) (Sobradelo and Martí 2010), El Hierro(Canary Islands) (Becerril et al. 2014), andDeception Island (Antarctica) (Bartolini et al.2014a).

In both long-term and short-term evaluationsHASSET can be useful for determining theoccurrence probabilities of eruptive scenarios andcan assist decision-makers assess the requiredmitigation actions associated with each scenarioand estimate the corresponding potential risk.

Simulation ModelsSimulating eruptive scenarios caused directly(e.g. lava flows, fallout, surges) and indirectly(earthquakes, landslides) by an eruption requiresa detailed analysis of the past activity of thevolcano or volcanic area, and must take intoaccount all the possible hazards associated withthe eruptive activity. Volcanic hazard can beassessed via two different types of approaches:deterministic, which defines a maximum exten-sion area affected by an eruptive episode basedon deposits generated by past activity, or prob-abilistic, based on the probability that a certainarea will be affected by an eruptive process. Inorder to generate hazard maps (Fig. 5d), it isimportant to understand past eruptive behaviourand to employ physical simulation models thatwill permit the behaviour of future volcanicactivity to be foreseen. In this type of approach,accurate and detailed geographic and carto-graphic data are required for high-quality analy-sis with a GIS.

Here, we describe some of the e-tools freelyavailable for download that allow volcanic haz-ards to be evaluated (Fig. 5c):

– VORIS 2.0.1 is a GIS-based tool, developedby Felpeto et al. (2007) that allows users tosimulate lava flows, fallout, and pyroclasticdensity current scenarios. Lava-flow simula-tions are based on a probabilistic model thatassumes that topography is the most

important factor determining the path of alava flow. The determination of the proba-bility of each point being invaded by lava isperformed by computing several randompaths with a Monte Carlo algorithm. Falloutsimulation models are advection diffusionmodels that assume that away from the ventthe transport of the particles from a Pliniancolumn is controlled by the advective effectof the wind, by diffusion due to atmosphericturbulence, and by the settling velocity of theparticles. The model for simulating pyro-clastic density currents is the energy conemodel proposed by Sheridan and Malin(1983). The input parameters are the topog-raphy, the collapse equivalent height (H), andthe collapse equivalent angle (h). The inter-section of the energy cone, originating at theeruptive source, with the ground surfacedefines the distal limits of the flow.

– HAZMAP is a free program for simulatingthe sedimentation of volcanic particles atdiscrete point sources that predicts the corre-sponding ground deposits (deposit mode)(Macedonio et al. 2005). HAZMAP is alsoable to evaluate the probability of overcominga given loading threshold in ground depositsby using a set of different wind profilesrecorded on different days (probability mode).Using a statistical set of recorded wind pro-files (and/or other input parameters), it canalso be used to draw hazard maps for ashfalldeposits. In HAZMAP, settling velocities canbe calculated using several models as a directfunction of particle diameters, densities, andshapes. The advantage of HAZMAP is that itis a simple tool able to predict ashfall duringhypothetical or real eruptions of a givenmagnitude and wind profile.

– LAHARZ is a semi-empirical code for cre-ating hazard-zonation maps that depict esti-mates of the location and extent of areasinundated by lahars (Schilling 1998). Theinput parameters for this model are the DigitalElevation Model and the lahar volume, whichprovide an automated method for mappingareas of potential lahar inundation. Thesehazard zones can be displayed in a GIS with

56 S. Bartolini et al.

other types of volcano hazard informationsuch as the proximal hazard zone, infras-tructure, hydrology, and population, as wellas contours and shaded relief, to producevolcano hazard-zonation maps. Such mapsshow the proximity and intersection ofpotential hazard zones to people andinfrastructures.

– TITAN2D is a computer program modeldeveloped by the University at Buffalo (Patraet al. 2005) that simulates granular flows overdigital elevation models, based on a “thinlayer model”. The input for the computercode includes simulation time, minimumthickness of the final deposit, internal and bedfriction angles, starting coordinates, and theinitial speed and direction of the flow. Addi-tionally, this program allows users to definethe specific starting pile dimensions or adynamic flow source. The outputs from theprogram (represented dynamically) are flowdepth and momentum, which yield thedeposit limit, run-out path, average flowvelocity, inferred deposit thickness, and traveltime.

Vulnerability E-tool

Once we have obtained hazard maps, the nextstep consists of adding population, infrastruc-tures, and land-use data to evaluate the vulnera-bility associated with the impact of a determinedhazard. The data required for generating vulner-ability maps are very complex and varied, anddepend on the observation scale. Vulnerability isdirectly dependent on the type of phenomena inquestion and on the socio-economic characteris-tics of the environment. VOLCANDAM is a newe-tool based on the methodology developed byScaini et al. (2014) that generates maps

estimating the expected damage caused by vol-canic eruptions. VOLCANDAM (Fig. 6) con-sists of three main parts: exposure analysis,vulnerability assessment, and the estimation ofexpected damages. The exposure analysis iden-tifies the elements exposed to the potential haz-ard and focuses on the relevant assets of thestudy area (population distribution, social andeconomic conditions, and productive activitiesand their role in the regional economy). Thevulnerability analysis defines a physical vulner-ability indicator for all exposed elements, as wellas a corresponding qualitative vulnerabilityindex. Systemic vulnerability considers the pos-sible relevance of each element in the system andtheir interdependencies by taking into account allexposed and non-exposed elements (people,buildings, transportation network, urban services,and productive activities). Damage assessment isperformed by associating a qualitative damagerating to each combination of hazard and vul-nerability, bearing in mind their specific contextsand roles in the system. The way one element canbe damaged—and thus lose its functionality—depends in fact on the type of hazardous eventand the characteristics of the element. The resultis damage maps that can be displayed at differentlevels of detail, depending on user preferences.This tool aims to facilitate territorial planningand risk management in active volcanic areas.

Decision-Making

The evaluation of the “direct costs” and “factors”(indirect costs) that have an impact on the eco-nomic growth of an area affected by a volcanicevent needs to take into account a number ofelements. A cost-benefit analysis may assist thedecision-making process by evaluating the eco-nomic impact of the different scenarios. Theapproach used by Sobradelo et al. (2015) (Fig. 7)

Fig. 6 Steps in the vulnerability analysis in the VOLCANDAM approach (after Scaini et al. 2014). See text for details

Probabilistic E-tools for Hazard Assessment … 57

is a Bayesian decision model that applies a gen-eral, flexible, probabilistic approach to the man-agement of volcanic crises by combining thehazard and risk factors that decision-makers needfor a holistic analysis of a volcanic crisis. Thesefactors include eruption scenarios and theirprobabilities of occurrence, the vulnerability ofpopulations and their activities, and the costs offalse alarms and erroneous forecasts. This modelcan be implemented before an emergency to(i) pinpoint actions for reducing the vulnerabilityof a particular area, (ii) identify the optimummitigating actions during an emergency and howthese may change as new information is obtained,(iii) assess after an emergency the effectiveness ofa mitigating response and, in light of results,(iv) how to improve strategies before anothercrisis occurs. BADEMO (BAyesian DEcisionMOdel) is part of this integrated approach, andenables the previous analysis of the distributionof local susceptibility and vulnerability to erup-tions to be combined with specific costs andpotential losses. Indeed, BADEMO should beseen as a tool for improving communicationbetween the monitoring scientists who providevolcanological information and those responsiblefor deciding which action plans and mitigatingstrategies should be put into practice.

Discussion and Conclusions

Modern volcanology is a scientific disciplinewith social and practical applications that derivefrom its direct involvement in risk reduction.Today, a multidisciplinary approach is requiredto risk assessment since new methodologies areconstantly being developed and explored, and,for example, geo-spatial technologies arebecoming extremely helpful in the disaster-riskmanagement. A defined and solid methodologyfor assessing volcanic hazard and managing riskis fundamental in both long-term evaluations andin unrest phases when monitoring information isavailable. Furthermore, e-tools can be progres-sively modified and implemented in light of theoutcomes obtained in order to integrate as manymodels as possible into the assessment andmanagement of volcanic risk.

The possibility of foreseeing an event such asa volcanic eruption is more effective—andguarantees greater economic savings—than act-ing after a disaster. The potential of simplee-tools such as those presented in this chapter liesin the fact that they are freely available and havebeen developed on accessible and dynamicgraphical user interfaces. The main advantage of

Fig. 7 The volcanic crisis management cycle: stages and phases (after Sobradelo et al. 2015). See text for details

58 S. Bartolini et al.

using these e-tools is that new data or new modelresults can be easily incorporated into the pro-cedures for updating the hazard assessment. Bycontrast, the use of expensive commercial e-toolshampers the exchange of information and com-plicates their testing in situ in volcanic fields.Furthermore, non-free tools may hinder effectiverisk assessment since they are often beyond themeans of advisory and management groups withlimited financial, technological, and manpowerresources (Leidig and Teeuw 2015). However,this does not mean that anybody can use thee-tools discussed here for, on the contrary, theyrequire expertise in volcanic hazards and relatedissues and their use implies scientific knowledgethat will avoid incorrect outcomes.

In this chapter we have presented differentstatistical methodologies and e-tools for inter-preting volcanic data and assessing long- andshort-term volcanic hazards and vulnerability,and for carrying out cost-benefit analyses. Sta-tistical analysis enables us to extract informationabout the future behaviour of a volcano bylooking at the geological and historical activityof the volcanic system (VERDI database, Bar-tolini et al. 2014c). The QVAST tool (Bartoliniet al. 2013) can be used to analyse past activityand to calculate the possibility that new ventswill open (volcanic susceptibility), while theBayesian event tree statistical method HASSET(Sobradelo et al. 2014) can be applied to calcu-late eruption recurrence. Using these calcula-tions, we can identify a number of significantscenarios using GIS-based e-tools (i.e. VORIS2.0.1, HAZMAP, …) and evaluate the potentialextent of the main volcanic hazards expected tooccur in volcanic areas. The results obtainedallow us to generate volcanic hazard maps fordifferent levels of hazards, evaluate vulnerability(VOLCANDAM e-tools, Scaini et al. 2014),conduct cost-benefit analysis (BADEMO e-tools,Sobradelo et al. 2015), and, finally, managevolcanic risk.

During a volcanic crisis, emergency plansmust be put into practice and so different gov-ernment departments need to be prepared inadvance. Therefore, it is important to have

conducted previous long-term hazard assessment—among many other tasks—to properly managea volcanic crisis. They allow scientists andmanagers to understand the characteristics of thevolcano or volcanic area and its past eruptivehistory, and to infer the possible eruptive sce-narios that may occur in the future. With thisprevious information, short-term hazard assess-ment can be conducted when volcanic unreststarts and hazard maps can be drawn up andalert levels be defined; nevertheless, it isimportant to always bear in mind that the bestform of protection is the evacuation of thepopulation at risk. Volcanic monitoring is anessential part of short-term assessment and soshould be performed by experts and based on agood understanding of volcanic processes.Despite the possibility of conducting cost-benefitanalysis, which can help maintain the economicorder, the security and health of the populationshould always be the main concern.

One of the purposes in the near future is tocreate a new software platform (VolcanBox, seeVETOOLS European Project—www.vetools.eu)with a user-friendly interface. This platform willcontain different e-tools that, via a homogeneousand systematic methodology, will help minimiserisk. However, the feasibility and applicability ofeach tool will have to be analysed by differentgroups of experts with experience in regionspossessing different volcanological andsocio-economic scenarios. This evaluation mustalso bear in mind potential end users (i.e. CivilProtection agencies), not only to test the abilityof existing tools but also to understanddecision-makers’ needs and requirements, andtrain them in the use of these tools. If we are toreduce volcanic risk we must ensure that scien-tists, managers, and decision-makers are all fullyprepared to confront this phenomenon since thebest way to guarantee risk reduction is to possessgood knowledge of its causes.

Acknowledgements This work was supported by theEuropean Commission (FP7 Theme: ENV.2011.1.3.3-1;Grant 282759: VUELCO). The English text was correctedby Michael Lockwood.

Probabilistic E-tools for Hazard Assessment … 59

References

Aspinall WP (2006) Structured elicitation of expertjudgment for probabilistic hazard and risk assessmentin volcanic eruptions. In: Mader HM et al (eds) Statis-tics in volcanology. Special Publication of IAVCEI,Geological Society of London

Bartolini S, Cappello A, Martí J, Del Negro C (2013)QVAST: a new Quantum GIS plugin for estimatingvolcanic susceptibility. Nat Hazards Earth Syst Sci 13(11):3031–3042

Bartolini S, Geyer A, Martí J, Pedrazzi D, Aguirre-Díaz G(2014a) Volcanic hazard on deception Island (SouthShetland Islands. J Volcanol Geotherm Res, Antarc-tica). doi:10.1016/j.jvolgeores.2014.08.009

Bartolini S, Bolós X, Martí J, Riera E, Planagumà LL(2014b) Hazard assessment at the Quaternary LaGarrotxa Volcanic Field (NE Iberia). Nat Hazards.doi:10.1007/s11069-015-1774-y

Bartolini S, Becerril L, Martí J (2014c) VERDI: a newvolcanic management risk database design. J VolcanolGeotherm Res 288:132–143

Becerril L, Bartolini S, Sobradelo R, Martí J, Morales JM,Galindo I (2014) Long-term volcanic hazard assess-ment on El Hierro (Canary Islands). Nat Hazards EarthSyst Sci 14:1853–1870

Blong R (2000) Volcanic hazards and risk management.In: Sigurdsson H et al (eds) Encyclopedia of volca-noes. Academic, San Diego, pp 1215–1227

Cappello A, Neri M, Acocella V, Gallo G, Vicari A, DelNegro C (2012) Spatial vent opening probability mapof Etna volcano (Sicily, Italy). Bull Volcanol74:2083–2094. doi:10.1007/s00445-012-0647-4

Cappello A, Bilotta G, Neri M, Del Negro C (2013)Probabilistic modelling of future volcanic eruptions atMount Etna. J Geophys Res Solid Earth 118:1925–1935. doi:10.1002/jgrb.50190

Chester DK, Dibben CJL, Duncan AM (2002) Volcanichazard assessment in Western Europe. J VolcanolGeotherm Res 115:411–435

De La Cruz-Reyna S (1996) Long term probabilisticanalysis of future explosive eruptions. In: Scarpa R,Tilling RI (eds) Monitoring and mitigation of volcanichazards. Springer, Berlin, pp 599–629 (IAVCEI/UNESCO Volume)

Felpeto A, Martí J, Ortiz R (2007) Automatic GIS-basedsystem for volcanic hazard assessment. J VolcanolGeotherm Res 166:106116

Leidig M, Teeuw R (2015) Free software: a review, in thecontext of disaster management. Inter J Appl EarthObserv Geoinfo 42:49–56

Lirer L, Petrosino P, Alberico I (2001) Volcanic hazardassessment at volcanic fields: the Campi Flegrei casehistory. J Volcanol Geotherm Res 101(1–4):55–75

Macedonio G, Costa A, Longo A (2005) A computermodel for volcanic ash fallout and assessment ofsusbsequent hazard. Comput Geosci 31:837–845

Martí J, Felpeto A (2010) Methodology for the compu-tation of volcanic susceptibility. An example for mafic

and felsic eruptions on Tenerife (Canary Islands).J Volcanol Geotherm Res 195:69–77

Marti J, Spence R, Calogero E, Ordoñez A, Felpeto A,Baxter P (2008) Estimating building exposure andimpact to volcanic hazards in Icod de los Vinos,Tenerife (Canary Islands). J Volcanol Geotherm Res178:553–561

Martí J, Sobradelo R, Felpeto A, García O (2012)Eruptive scenarios of phonolitic volcanism atTeide-Pico Viejo volcanic complex (Tenerife, CanaryIslands). Bull Volcanol 74:767–782. doi:10.1007/s00445-011-0569-6

Martí J, Pinel V, López C, Geyer A, Abella R, Tárraga M,Blanco MJ, Castro A, Rodríguez C (2013) Causes andmechanisms of the 2011–2012 El Hierro (CanaryIslands) submarine eruption. J Geophys Res SolidEarth 118(3):823–839

Marzocchi W, Sandri L, Gasparini P, Newhall C,Boschi E (2004) Quantifying probabilities of volcanicevents: the example of volcanic hazard at MountVesuvius. J Geophys Res 109. doi:10.1029/2004JB003155

Neri A, Aspinall W, Cioni R, Bertagnini A, Baxter P,Zuccaro G, Andronico D, Barsotti S, Cole P,Ongaro TE, Hincks T, Macedonio G, Papale P,Rosi M, Santacroce R, Woo G (2008) Developingan event tree for probabilistic hazard and risk assess-ment at Vesuvius. J Volcanol Geotherm Res 178(3):397–415

Newhall CG, Hoblitt RP (2002) Constructing event treesfor volcanic crisis. Bull Volcanol 64:3–20

Sandri L, Jolly G, Lindsay J, Howe T, Marzocchi W(2012) Combining long-and short-term probabilisticvolcanic hazard assessment with cost-benefit analysisto support decision making in a volcanic crisis fromthe Auckland Volcanic Field, New Zealand. BullVolcanol 74(3):705–723

Sandri L, Thouret JC, Constantinescu R, Biass S,Tonini R (2014) Long-term multi-hazard assessmentfor El Misti volcano (Peru). Bull Volcanol 76(2):1–26

Scaini C, Felpeto A, Martí J, Carniel R (2014) AGIS-based methodology for the estimation of potentialvolcanic damage and its application to Tenerife Island,Spain. J Volcanol Geotherm Res 278–279:40–58

Schilling SP (1998) LAHARZ: GIS program for auto-mated mapping of lahar-inundation hazard zones. USGeological Survey Open-File Report, pp 98–638

Sheridan MF, Malin MC (1983) Application ofcomputer-assisted mapping to volcanic hazard evalu-ation of surge eruptions: Vulcano, Lipari. J VolcanolGeotherm Res 17:187–202

Sobradelo R, Martí J (2010) Bayesian event tree forlong-term volcanic hazard assessment: application toTeide-Pico Viejo stratovolcanoes, Tenerife, Canaryislands. J Geophys Res 115. doi:10.1029/2009JB006566

Sobradelo R, Bartolini S, Martí J (2014) HASSET: aprobability event tree tool to valuate future volcanicscenarios using Bayesian inference presented as aplugin for QGIS. Bull Volcanol 76:770

60 S. Bartolini et al.

Sobradelo R, Martí J, Kilburn C, López C (2015)Probabilistic approach to decision-making underuncertainty during volcanic crises: retrospective appli-cation to the El Hierro (Spain) 2011 volcanic crisis.Nat Hazards 76(2):979–998

Thierry P, Neri M, Le Cozannet G, Jousset P, Costa A(2015) Preface: approaches and methods to improverisk management in volcanic areas. Nat Hazards EarthSyst Sci 15:197–201

Toyos GP, Cole PD, Felpeto A, Martí J (2007) AGIS-based methodology for hazard mapping of smallpyroclastic density currents. Nat Hazards 41:99–112

Twigg J (2004) Disaster risk reduction: mitigation andpreparedness in development and emergency pro-gramming. Overseas Development Institute, Human-itarian Practice Network, London, Good PracticeReview no. 9

Van Westen CJ (2013) Remote sensing and GIS fornatural hazards assessment and disaster risk manage-ment. In: Shroder J (editor in Chief), Bishop MP(ed) Treatise on geomorphology. vol 3, RemoteSensing and GIScience in Geomorphology. AcademicPress, San Diego, pp 259–298

Open Access This chapter is licensed under the terms ofthe Creative Commons Attribution 4.0 InternationalLicense (http://creativecommons.org/licenses/by/4.0/),which permits use, sharing, adaptation, distribution andreproduction in any medium or format, as long as yougive appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons licenseand indicate if changes were made.

The images or other third party material in this chapterare included in the chapter’s Creative Commons license,unless indicated otherwise in a credit line to the material.If material is not included in the chapter’s CreativeCommons license and your intended use is not permittedby statutory regulation or exceeds the permitted use, youwill need to obtain permission directly from the copyrightholder.

Probabilistic E-tools for Hazard Assessment … 61