Volcanic Hazard Mapping Case Study
Armento MC, Genevois R, Tecca PR () Comparison of numerical models of two debris flows in the Cortina d’Ampezzo area, Dolomites, Italy. Landslides –CrossRefGoogle Scholar
Benvenuti M () Facies analysis and tectonic significance of lacustrine fan-deltaic successions in the Pliocene-Pleistocene Mugello Basin, Central Italy. Sediment Geol –CrossRefGoogle Scholar
Brooks RH, Corey AT () Hydraulic properties of porous media, Hydrol. Pap., 3. Colorado State University, Fort CollinsGoogle Scholar
Bruno P, de Alteriis G, Florio G () The western undersea section of the Ischia volcanic complex (Italy, Tyrrhenian Sea) inferred by marine geophysical data. Geophys Res Let 29(9):1–4, Art. no. CrossRefGoogle Scholar
Canuti P, Casagli N, Catani F, Falorni G () Modeling of the Guagua Pichincha volcano (Ecuador) lahars. Phys Chem Earth Parts A/B/C 27(36)CrossRefGoogle Scholar
Casagli N, Dapporto S, Ibsen ML, Tofani V, Vannocci P () Analysis of the landslide triggering mechanism during the storm of 20thth November , in Northern Tuscany. Landslides –21CrossRefGoogle Scholar
De Vita S, Sansivero F, Orsi G, Marotta E () Cyclical slope instability and volcano-tectonism in resurgent calderas: the Ischia island (Italy) case study. Eng Geol –CrossRefGoogle Scholar
De Vita P, Di Clemente E, Rolandi M, Celico P () Engineering geological model of the initial landslides occurred on the April 30th at the Mount Vezzi (Ischia Island, Italy). Ital J Eng Geol Environ –Google Scholar
Fiorillo F, Guadagno F, Aquino S, De Blasio A () The December Cervinara landslides: further debris flow in the pyroclastic deposits of Campania (Southern ltaly). Bull Eng Geol Environ 60()–Google Scholar
Fredlund DG, Morgenstern NR, Widger RA () The shear strength of unsatured soil. Can Geotech J 15(3)–Google Scholar
Fredlund DG, Rahardjo H () Soil mechanics for unsaturated solis. John Wiley And Sons Inc., New YorkGoogle Scholar
Geostudio (a) Geostudio , SEEP/W user manual, Geo Slope InternationalGoogle Scholar
Geostudio (b) Geostudio , SLOPE/W user manual, Geo Slope InternationalGoogle Scholar
Green RE, Corey GJ (). Calculation of hydraulic conductivity: further evaluation of same predictive methods. Soil Sci Soc AM Proc –8Google Scholar
Guzzetti F () Landslide fatalities and evaluation of landslide risk in Italy. Eng Geol –Google Scholar
Heim A () Bergstruz und Menschenleben. Fretz und Wasmuth, Zürich, p Google Scholar
Hungr O () A model for the runout analysis of rapid flow slides, debris flows, and avalanches. Can Geotech J –CrossRefGoogle Scholar
Hungr O. (). Analytical models for slides and flows. In Proceeding International Symposium on Landslide Risk Mitigation and protection of Cultural and Natural Heritage, Kyoto, 21–25 January (), pp. –Google Scholar
Hungr O, Evans SG, Bovis MJ, Hutchinson JN () A review of the classification of landslides of the flow type. Environ Eng Geosci VII(3)–CrossRefGoogle Scholar
Hungr O, Morgan GC, Kellerhals R () Quantitative analysis of debris torrent hazard for design of remedial measures. Can Geotech J –CrossRefGoogle Scholar
Jackob M. & Hungr O. (). Debris flow hazards and related phenomena. Springer, woaknb.wz.sk Scholar
Johnson AM, Rodine JR () Debris flow. In: Brunsden D, Prior DB (eds) Slope instability. John Wiley and Sons Ltd., London, Chap. 8, p –Google Scholar
O’Brien JS, Julien PY, Fullerton WT () Two-dimensional water flood and mudflow simulation. J Hydraul Div, ASCE (HY2)–CrossRefGoogle Scholar
O'Brien J.S. &. Julien P.Y (). Physical processes of hyperconcentrated sediment flows. In Utah Water Research Laboratory (eds) () Proceeding of the ASCE Specialty Conf. on the Delineation of Landslides, Floods, and Debris Flow Hazards in Utah, Series UWRL/g/03, pp. –Google Scholar
Orsi G, de Vita S, Di Vito M, Isaia R, Nave R, Heiken G () Facing volcanic and related hazards in the Neapolitan area. In: Heiken G, Fakundiny R, Sutter J (eds) Earth sciences in cities. American Geophysical Union (Special Publication), Washington, pp –Google Scholar
Migale LS, Milone A () Colate di Fango in terreni piroclastici della Campania. Primi dati della ricerca storica. Rassegna Storica Salernitana –Google Scholar
Morelli S, Monroi VHG, Gigli G, Falorni G, Rocha EA, Casagli N () The Tancitaro debris avalanche: characterization, propagation and modeling. J Volcanol Geotherm Res (12)–CrossRefGoogle Scholar
Morgenstern R, Price VE () The analysis of stability of general slip surfaces. Géotechnique –Google Scholar
Pareschi MT, Favalli M, Giannini F, Sulpizio R, Zanchetta G, Santacroce R () May 5, , debris flows in cirmum-Vesuviuan areas (southern Italy): insights for hazard assessment. Geology –CrossRefGoogle Scholar
Philip JR () Approximate analysis of the borehole permeameter in unsaturated soil. Water Res –Google Scholar
Piochi M, Civetta L, Orsi G () Mingling in the magmatic system of Ischia (Italy) in the past 5 ka. Mineral Petrol –CrossRefGoogle Scholar
Revellino P, Hungr O, Guadagno FM, Evans SG () Velocity and runout simulation of destructive debris flows and debris avalanches in pyroclastic deposits, Campania region, Italy. Environ Geol –CrossRefGoogle Scholar
Rosi M, Principe C, Vecci R () The eruption of Vesuvius reconstructed from the review of chronicles and study of deposits. J Volcanol Geotherm Res –CrossRefGoogle Scholar
Sansivero F () Assetto stratigrafico ed evoluzione vulcanologica del settore orientale dell'isola d'Ischia negli ultimi 10 Ka. PhD Thesis. University of Naples, Italy, ppGoogle Scholar
Sassa K, Wang G () Mechanism of landslide-triggered debris flows: liquefaction phenomena due to the undrained loading of torrent deposit. In: Jakob M, Hungr O (eds) Debris-flow hazard and related phenomena. Sprinter Verlag, Berlin, p 81–Google Scholar
Scotto di Santolo A, Evangelista A () Some observations on the prediction of the dynamic parameters of debris flows in pyroclastic deposits in the Campania region of Italy, Nat. Hazards –CrossRefGoogle Scholar
Smith GA, Lowe DL () Lahars: volcano-hydrologic events and deposition in the debris flow-hyperconcentrated flow continuum. In: Fisher RV, Smith GA (eds) Sedimentation in volcanic settings, vol 45, SEPM Special Publication., pp –CrossRefGoogle Scholar
Tofani V, Dapporto S, Vannocci P, Casagli N () Infiltration, seepage and slope instability mechanisms during the 20–21 November rainstorm in Tuscany, central Italy. Nat Hazards Earth Syst Sci –CrossRefGoogle Scholar
Vezzoli L., (). Island of Ischia. CNR Quaderni de “La ricerca scientifica”, vol. (10). woaknb.wz.sk Scholar
Voellmy A. (). Über die Zerstörungskraft von Lawinen Schweiz Bauzeitung 73, –Google Scholar
Wagner AA () The use of the unified soil classification system by the bureau of reclamation. Proceedings of Fourth International Conference on Soil Mechanics and Foundation Engineering, London, p Google Scholar
A sound management of volcanic unrest in high-risk areas requires a quantitative assessment of volcanic hazard in a short-time perspective, i.e., in time windows of few hours/few weeks [e.g., Marzocchi et al., a]. Yet short-term volcanic hazard models are still lacking, and during volcanic unrest, volcanologists usually assist decision makers producing maps of the deterministic outcome, or at best the probabilistic impact, of one or a few selected scenarios [Folch et al., ; Scollo et al., ]. These maps are undoubtedly important for decision makers, but only a fully probabilistic volcanic hazard analysis (PVHA) can provide a quantitative solid basis for a rationale decision making [Cornell and Krawinkle, ; Der Kiureghian, ; Marzocchi and Woo, , ; Sandri et al., ]. PVHA accounts properly for the intrinsic variability of the system and all known uncertainties [Senior Seismic Hazard Analysis Committee (SSHAC), ; Marzocchi and Jordan, ], and it offers a more complete view of the real volcanic hazard than any single scenario could do [Selva et al., ].
In this paper, we introduce a first short-term PVHA model, rooted on the Bayesian event tree philosophy [Marzocchi et al., , , ], that aims at (1) including within a formal procedure all the relevant information that can constrain short-term forecasts, from past and monitoring data to external forecast models (like weather forecast); (2) managing all the uncertainties, both natural (related to the intrinsic nonpredictability of the phenomena) and epistemic (related to the limited knowledge about the system); and (3) producing in output probabilistic assessments suitable for risk analysis and decision making.
Even the state-of-the-art operative procedures for dealing with short-term volcanic hazards are typically based only on scenario approaches [e.g., Scollo et al., ], which have been demonstrated to be strongly limited in their description of potential hazards and consequent impact [Selva et al., ] and inapplicable for rational decision making [Cornell and Krawinkle, ; Marzocchi and Woo, ].
First, the probability of occurrence of an imminent eruption and of the impact of potential consequences is not evaluated. Such probability is a necessary factor in quantitative studies for rational decision making, like cost-benefit analysis [e.g., Marzocchi and Woo, ; Sandri et al., ]. In recent literature, several methods have been developed to evaluate short-term eruption probability, based on quantitative analysis of monitoring information [Aspinall et al., ; Aspinall, ; Marzocchi et al., ; Lindsay et al., ; Selva et al., a]. However, the link between short-term eruption forecast and consequent eruptive hazard has been explored only for time-independent hazardous phenomena [e.g., Aspinall et al., ; Sparks, ; Sandri et al., ]. Much more challenging is the investigation of the time-dependent hazardous events, such as tephra fall or gas dispersion, whose hazard is strongly linked to the time-varying wind field.
Second, no assessment of completeness and likelihood of the selected scenario(s) is performed, although not all the scenarios are equally probable, neither in size nor in vent position. In addition, no attention is paid in covering the whole range of potential eruptions that may occur, strongly biasing the forecast. This lack in modeling aleatory uncertainty does largely influence the results. In terms of vent positions, the most likely ones can be constrained by structural information and location of past vents [e.g., Martin et al., ; Selva et al., b] and, in a short-term perspective, by the localization of several preeruptive phenomena (e.g., seismic events and deformations) [Lindsay et al., ]. In terms of explosive eruptive sizes, large sizes are usually less likely [e.g., Marzocchi et al., ; Neri et al., ; Orsi et al., ] than smaller ones, but still they may occur and are connected to higher levels of damage.
Third, other known and relevant uncertainties are hidden, for example, the ones related to specific modeling choices or to accessory forecasts setting either initial and/or boundary conditions for modeling procedures. Indeed, given the potential regulatory concern of short-term PVHA, attention should be spent in quantifying all the known unknowns, in order “to represent the center, the body, and the range of technical interpretations that the larger technical community would have if they were to conduct the study” [SSHAC, ].
The proposed model is then applied to estimate the tephra hazard during a realistic volcanic unrest at Mount Vesuvius. Tephra dispersal during explosive volcanic eruptions and its consequent deposition can produce multiple hazardous effects, such as damage to human settlements, agriculture, and transportation systems [Blong, ; Casadevall, ; Sparks et al., ; Miller and Casadevall, ; Horwell and Baxter, ]. Although the modelistic aspects of the tephra dispersion are well explored, as well as some application for the long-term assessment [e.g., Barberi et al., ; Cioni et al., ; Bonadonna et al., ; Magill et al., ; Macedonio et al., ; Bonasia et al., , ], the short-term tephra hazard analysis is still in its infancy. Besides the paramount importance in assisting decision makers during volcanic unrest in high-risk zone like in Neapolitan area, the recent disruption to civil aviation over Europe caused by the Eyjafjallajökull eruption [e.g., Bonadonna et al., ; Folch et al., ] and the repeated closures of the Catania airport caused by the Etna activity highlight the importance to build reliable and flexible models of short-term hazard assessment.
In the following sections, we first describe BET_VH_ST; then we show the potential of its application through a retrospective analysis. In particular, in section 2 we describe the main features of BET_VH_ST, showing (i) how the monitoring observations are used to update the short-term volcanic hazard, (ii) the calibration of each node of the event tree, and (iii) how the tephra modeling is incorporated into the final assessment of the volcanic hazard. Then, in section 4, we retrospectively apply BET_VH_ST to the Major Emergency Simulation Exercise (MESIMEX) exercise at Mount Vesuvius [Barberi and Zuccaro, ; Marzocchi et al., ] for the tephra fallout. This exercise was promoted by the European Commission together with Italian Civil Protection Department to focus on the preparatory phase of a possible reactivation of such volcano, located in a densely inhabited area like Naples and thus posing a high volcanic risk. MESIMEX exercise took place from 19 to 23 October and consisted in a blind simulation of Vesuvius reactivation, from the early warning phase up to the final eruption, and it included evacuation of a sample of about people from the area at risk, as established by the Emergency Plan.
2 Short-Term Volcanic Hazard Assessment: The Model BET_VH_ST
PVHA quantifies the exceedance probability of an intensity measure of a specific phenomenon (e.g., tephra load) in a target area and within a time window (exposure time) [e.g., SSHAC, ]. These exceedance probability curves are commonly referred to as hazard curves, and they represent the full information about the probabilistic hazard assessment [Rougier et al., ]. Indeed, they can be used to produce hazard maps and probability maps and for probabilistic risk assessments [e.g., Elefante et al., ].
The short-term PVHA method proposed here is a specific development of the Bayesian event tree (BET) model [Marzocchi et al., , ] for short-term hazard purposes. The BET model applies a Bayesian inference procedure to assess the probability of occurrence of the relevant phenomena that may occur at one target volcano. Based on a structured event tree, monitoring data, past eruptive history, and theoretical models are dynamically used at different levels, in order to deal with both short- and long-term analyses and explicitly evaluate aleatory and epistemic uncertainties. There is a rather extensive literature presenting different methodological aspects of BET model [e.g., Marzocchi et al., , ; Lindsay et al., ; Selva et al., ]. The new model we propose combines different aspects of methods previously developed and introduces some key improvements and changes.
The main goal of BET_VH_ST is to quantify the hazard curve associated with one (or more) volcanic phenomenon, in the target area and in a time window T, that is
where is a threshold of the intensity L in ; E indicates an eruption in the time window T; ESu the uth eruptive settings, defined as an eruption in a given vent location of a given size and the sum runs over a complete set of .
In our application the length of the time window is defined according to the needs of the decision makers.
In Figure 1a, we report the event tree used by BET_VH_ST. This event tree is composed of eight levels, or nodes, which describe with increasing details the possible evolution of an eruption. In particular, (1) nodes 1 to 3 deal with the problem of eruption forecasting, starting from the detection of unrest (node 1), the detection of movements of magma bodies (node 2), and the onset of the eruption (node 3); (2) nodes 4 and 5 deal with the problem of forecasting the eruptive scenario, treating the variability on the potential vent positions (node 4) and the potential eruption magnitude and intensity (i.e., size class) (node 5); and (3) nodes 6 to 8 deal with the problem of forecasting the impact of one specific eruptive phenomenon (like tephra fall), analyzing the possibility of the occurrence of the phenomenon (node 6) and the involvement, at various levels, of specific areas around the volcano (nodes 7 and 8).
These three macrolevels correspond to the three probabilistic factors in equation (1), and they deal with the uncertainty on the occurrence of one eruption (eruption forecasting), the uncertainty related to the natural variability of eruptive scenarios (eruptive scenario forecasting), and the uncertainty on the effects of each eruptive scenario (impact forecasting), respectively. At all levels, BET_VH_ST estimates both aleatory (due to natural variability) and epistemic (due to limitations on knowledge and/or modeling capability) uncertainty so that each probabilistic term (hereinafter indicated with θ) is assessed along with the uncertainty in its evaluation, described through a probability distribution (hereinafter indicated with [θ]).
In Figure 1b, we show as the uncertainties considered in these macrolevels of the event tree propagate into probabilistic hazard analyses, allowing the assessment of hazard curves either conditioned to the occurrence of one eruption with whatever scenario (conditional PVHA) or unconditioned (absolute PVHA). Hazard curves, and relative epistemic uncertainties, represent all the information required to characterize the hazard at each target [e.g., SSHAC, ]. To provide an idea of the aleatory uncertainty on the assessment, different maps, with different probability thresholds, may be produced, starting from the best guess (mean) hazard curve. An example of this is reported in Figure 2, where three different hazard maps are obtained from the mean hazard curve with three exceedance probability levels of , , and On the other hand, to provide a visualization of the epistemic uncertainty on the assessment, different maps, at different level of confidence, may be produced by cutting the hazard curves at different percentiles. An example of this is reported in Figure 3, where three different hazard maps are obtained for the same exceedance probability level of but considering the mean hazard curve, as well as the hazard curve at the 10th and 90th percentiles.
Even if probabilities at nodes 1–5 in BET_VH_ST are calculated similarly to BET_EF model [Marzocchi et al., , ], and at nodes 6–8 to BET_VH model [Marzocchi et al., ; Selva et al., ], BET_VH_ST requires several key methodological developments with respect to those models. First, we introduce refinements of the combination of short- and long-term assessments, in order to produce a consistent treatment of the epistemic uncertainties throughout the hazard quantification. Second, we introduce specific methods to deal with time-dependent forecasts for different temporal duration. Third, we implement the hazard analysis to produce complete hazard curves [Selva and Sandri, ; Rougier et al., ], which represent the exceedance probability of predefined threshold values for the parameter describing the intensity of the hazard at a given site. Finally, we include a postprocessing analysis of the hazard results, allowing a simple and direct access to the results, accounting for uncertainties. Such developments are discussed in details in the next sections.
Combination Scheme for Short- and Long-Term Assessments
Short-term PVHA relies on two different kinds of information: measurements from monitoring systems (data set ) and all the other kinds of data/information (data set , which include past data, statistical and physical models, and theoretical beliefs regarding the target volcano and/or analogous ones). For example, monitoring measures are pivotal for providing short-term eruption forecast during unrest periods [e.g., Aspinall et al., ; Aspinall, ], while they do not carry any relevant information during a quiet period, apart from telling that the volcano is at rest.
For this reason, in Marzocchi et al.  these two kinds of information were used to produce two separate assessments, at each level of the event tree, and then combined as