Advances in Volcanology

Joachim Gottsmann
Jürgen Neuberg
Bettina Scheu Editors

Volcanic
Unrest
From Science to Society
Advances in Volcanology
An Official Book Series of the International
Association of Volcanology and Chemistry
of the Earth’s Interior – IAVCEI, Barcelona,
Spain
Series editor
Karoly Nemeth, Palmerston North, New Zealand
More information about this series at http://www.springer.com/series/11157
Joachim Gottsmann Jürgen Neuberg
•

Bettina Scheu
Editors

Volcanic Unrest
From Science to Society
Editors
Joachim Gottsmann Bettina Scheu
School of Earth Sciences Ludwig Maximilian University of
University of Bristol Munich
Bristol, Bath and North East Somerset Munich, Bayern
UK Germany

Jürgen Neuberg
Faculty of Environment
University of Leeds
Leeds, UK

ISSN 2364-3277 ISSN 2364-3285 (electronic)

Advances in Volcanology
ISBN 978-3-319-58411-9 ISBN 978-3-319-58412-6 (eBook)
https://doi.org/10.1007/978-3-319-58412-6

Library of Congress Control Number: 2018961003

© The Editor(s) (if applicable) and The Author(s) 2019. This book is published open access.
Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give appro-
priate credit to the original author(s) and the source, provide a link to the Creative Commons
licence and indicate if changes were made.
The images or other third party material in this book are included in the book’s Creative
Commons licence, unless indicated otherwise in a credit line to the material. If material is not
included in the book’s Creative Commons licence and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in
this book are believed to be true and accurate at the date of publication. Neither the publisher nor
the authors or the editors give a warranty, express or implied, with respect to the material
contained herein or for any errors or omissions that may have been made. The publisher remains
neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

Volcanic unrest is a manifestation of complex subsurface processes which

might or might not herald an imminent volcanic eruption. Short-term (hours
to few weeks) unrest activity before an eruption may in hindsight be inter-
preted as “pre-eruptive”. Protracted periods or waning of unrest without a
clear relationship between unrest and future eruptive activity may be deemed
“non-eruptive”.
Periods of unrest challenge both scientists and decision-makers to interact
effectively and efficiently to minimise societal risk during the unrest and in
anticipation of a possible imminent eruption.
The scientific challenge is to identify whether unrest phenomena are
pre-eruptive or non-eruptive early on in a developing unrest crisis. The
challenge for decision-makers is to respond with the “right” decision during
an emerging crisis.
Expectations from many different stakeholders including the media and
the general public need to be managed amid epistemic (the unknown sub-
surface processes hidden from the observer) and aleatoric uncertainty (their
stochastic variability).
Numerous examples in the recent past have shown that periods of unrest
are hazardous and costly even without leading to imminent eruption (e.g. at
La Soufriere, Guadeloupe, in 1976; at Campi Flegrei, Italy, in 1982–1984; at
Long Valley Caldera/Mammoth Mountain, USA, in the 1980s and 1990s).
Tensions between different stakeholders can arise particularly during
protracted periods (years to decades) of unrest at volcanoes with significant
and growing populations, when trust in scientific knowledge is challenged
amid the need for economic and societal development and growth.
Volcanic Unrest: From Science to Society—highlights the complexities of
volcanic unrest from both scientific and societal perspectives and provides a
summary of findings from the project entitled “Volcanic Unrest in Europe
and Latin America: Phenomenology, eruption precursors, hazard forecast,
and risk mitigation (VUELCO)”. The multi-disciplinary and cross-boundary
research project was funded by the European Commission’s 7th framework
programme for research, technological development and demonstration
under grant agreement no. 282759 and brought together an international team
of academics from the social and natural sciences, as well as personnel from
volcano observatory and civil protection agencies.

v
vi Preface

The project was designed in response to a call by the EC for a “collab-

orative project (small- or medium-scale focused research project) for specific
cooperation actions (SICA) dedicated for international cooperation partner
countries with focus on Latin America”.
The work was conducted at ten partner institutions across Europe and
Latin America during 2011–2015. Most of the findings have been published
in dedicated scientific journals over the past few years, and others will be
published in the years to come.
The purpose of this open-access book is to summarise and publicise the
key findings of the project for a broad audience. Seventeen and one appendix
chapters focus on four broad topics relevant to the understanding of volcanic
unrest in the context of scientific and societal challenges:

1. The significance of volcanic unrest at the natural hazard and risk interface,
2. Geophysical and geochemical fingerprints of unrest and precursory
activity,
3. Subsurface dynamics leading to unrest phenomena,
4. Stakeholder interaction and volcanic risk governance.

This book aims to make our research accessible to both scientific and
non-scientific audiences with interest in the different aspects of volcanic
unrest, its impact and consequences. The chapters have been written with the
intention to make the findings accessible to a broad audience. This entails a
balancing act between keeping the scientific jargon at bay, whilst also sat-
isfying the curiosity of scientific readers from different disciplines. In addi-
tion, most chapters are accompanied by Spanish-language abstracts.
As a consequence, the style of the chapters is different in several ways
from the general peer-reviewed scientific literature. Most chapters have a
summary/review character of findings from the project, which were originally
published in dedicated journals. The chapters hence provide syntheses and
articulations of concepts rather than comprehensive compilation of data and
the available literature. However, all chapters provide the reader with ref-
erences to original publications that will permit a wider reading and study
of the findings behind the chapters.
The book is a joint effort between editors, authors, reviewers and
publishers.
All chapters have undergone peer review, and we are indebted to all
reviewers from the VUELCO community as well as the following external
reviewers who provided their expert opinions and comments:
S. de Angelis, F. Arzilli, O. Bachmann, F. Costa, N. Deligne, J. Gardner,
H. Gonnermann, A. Hicks, S. Hurwitz, S. Jenkins, P. Lesage, C. Newhall,
C. Pritchard, G. Woo.
We thank J. Schwarz at Springer for her patience and expert handling of
all things related to publishing this book.

Munich, Germany Bettina Scheu

Bristol, UK Joachim Gottsmann
Leeds, UK Jürgen Neuberg
Contents

Volcanic Unrest and Pre-eruptive Processes: A Hazard

and Risk Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
J. Gottsmann, J.-C. Komorowski and J. Barclay
The Role of Laws Within the Governance of Volcanic Risks. . . . . 23
R. J. Bretton, J. Gottsmann and R. Christie
Deterministic Versus Probabilistic Volcano Monitoring:
Not “or” But “and”. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
D. Rouwet, R. Constantinescu and L. Sandri
Probabilistic E-tools for Hazard Assessment
and Risk Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Stefania Bartolini, Joan Martí, Rosa Sobradelo and Laura Becerril
The Need to Quantify Hazard Related to Non-magmatic Unrest:
From BET_EF to BET_UNREST . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Laura Sandri, Roberto Tonini, Dmitri Rouwet,
Robert Constantinescu, Ana Teresa Mendoza-Rosas,
Daniel Andrade and Benjamin Bernard
Groundwater flow and volcanic unrest . . . . . . . . . . . . . . . . . . . . . . 83
Alia Jasim, Brioch Hemmings, Klaus Mayer and Bettina Scheu
Experimental Simulations of Magma Storage and Ascent . . . . . . . 101
C. Martel, R. A. Brooker, J. Andújar, M. Pichavant, B. Scaillet
and J. D. Blundy
Magma Chamber Rejuvenation: Insights from
Numerical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
C. P. Montagna, P. Papale, A. Longo and M. Bagagli
Magma Mixing: History and Dynamics of an Eruption
Trigger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Daniele Morgavi, Ilenia Arienzo, Chiara Montagna, Diego Perugini
and Donald B. Dingwell
Gases as Precursory Signals: Experimental Simulations,
New Concepts and Models of Magma Degassing . . . . . . . . . . . . . . 139
M. Pichavant, N. Le Gall and B. Scaillet

vii
viii Contents

Crystals, Bubbles and Melt: Critical Conduit Processes

Revealed by Numerical Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
M. E. Thomas, J. W. Neuberg and A. S. D. Collinson
When Does Magma Break? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Fabian B. Wadsworth, Taylor Witcher, Jérémie Vasseur,
Donald B. Dingwell and Bettina Scheu
Volcano Seismology: Detecting Unrest in Wiggly Lines . . . . . . . . . 185
R.O. Salvage, S. Karl and J.W. Neuberg
The Ups and Downs of Volcanic Unrest: Insights from
Integrated Geodesy and Numerical Modelling . . . . . . . . . . . . . . . . 203
J. Hickey, J. Gottsmann, P. Mothes, H. Odbert, I. Prutkin
and P. Vajda
Fluid Geochemistry and Volcanic Unrest: Dissolving
the Haze in Time and Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Dmitri Rouwet, Silvana Hidalgo, Erouscilla P. Joseph
and Gino González-Ilama
Geophysical Footprints of Cotopaxi’s Unrest and Minor
Eruptions in 2015: An Opportunity to Test Scientific and
Community Preparedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Patricia A. Mothes, Mario C. Ruiz, Edwin G. Viracucha,
Patricio A. Ramón, Stephen Hernández, Silvana Hidalgo,
Benjamin Bernard, Elizabeth H. Gaunt, Paul Jarrín, Marco A. Yépez
and Pedro A. Espín
Volcanic Unrest Simulation Exercises: Checklists
and Guidance Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
R. J. Bretton, S. Ciolli, C. Cristiani, J. Gottsmann, R. Christie
and W. Aspinall
Appendix: Volcanic Unrest: Terminology and Definitions . . . . . . . 299
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
 Volcanic Unrest and Pre-eruptive
Processes: A Hazard and Risk
Perspective

J. Gottsmann, J.-C. Komorowski and J. Barclay

Abstract
Volcanic unrest is complex and capable of producing multiple hazards that
can be triggered by a number of different subsurface processes. Scientific
interpretations of unrest data aim to better understand (i) the processes
behind unrest and their associated surface signals, (ii) their future
spatio-temporal evolution and (iii) their significance as precursors for
future eruptive phenomena. In a societal context, additional preparatory or
contingency actions might be needed because relationships between and
among individuals and social groups will be perturbed and even changed
in the presence of significant uncertainty. Here we analyse some key
examples from three international and multidisciplinary projects
(VUELCO, CASAVA and STREVA) where issues around the limits of
volcanic knowledge impact on volcanic risk governance. We provide an
overview of the regional and global context of volcanic unrest and
highlight scientific and societal challenges with a geographical emphasis
on the Caribbean and Latin America. We investigate why the forecasting
of volcanic unrest evolution and the exploitability of unrest signals to
forecast future eruptive behaviour and framing of response protocols is

J. Gottsmann (&)
School of Earth Sciences, University of Bristol,
Bristol, UK
e-mail: [email protected]
J. Gottsmann
The Cabot Institute, University of Bristol, Bristol,
UK
J.-C. Komorowski
Institut de Physique du Globe de Paris et Université
Paris Diderot, Université Sorbonne Paris Cité, CNRS
UMR 7154, Paris, France
J. Barclay
School of Environmental Sciences, University of
East Anglia, Norwich, UK

Advs in Volcanology (2019) 1–21

DOI 10.1007/11157_2017_19
© The Author(s) 2017
Published Online: 17 August 2017
2 J. Gottsmann et al.

challenging, especially during protracted unrest. We explore limitations of

current approaches to decision-making and provide suggestions for how
future improvements can be made in the framework of holistic volcanic
unrest risk governance. We investigate potential benefits arising from
improved communication, and framing of warnings around
decision-making timescales and hazard levels.

Resumen
La agitación volcánica es compleja y capaz de generar múltiples peligros
que pueden ser desencadenados por un número diferente de procesos
subsuperficiales. Las interpretaciones científicas sobre datos de agitación
volcánica tienen como objetivo el mejor entendimiento de (i) los procesos
detrás de la agitación volcánica y sus señales superficiales asociadas, (ii) su
evolución espacial-temporal y (iii) su significado como precursores de
fenómenos eruptivos a futuro. Dentro de un contexto social, acciones
adicionales preparatorias o de contingencia podrían ser requeridas debido a
que las relaciones entre individuos y dentro de grupos sociales serán
perturbadas e inclusive modificadas ante la presencia de incertidumbre
significativa. Aquí nosotros analizamos algunos ejemplos clave a partir de
tres proyectos internacionales y multidisciplinarios (VUELCO, CASAVA
y STREVA) en los cuales las cuestiones alrededor de los límites del
conocimiento volcánico tienen impacto en la gestión pública del riesgo
volcánico. Proveemos una perspectiva general del contexto regional y
global de la agitación volcánica y sobresaltamos retos científicos y sociales
con énfasis geográfico en el Caribe y América Latina. Investigamos porqué
el pronóstico de la evolución en la agitación volcánica y el aprovecha-
miento de señales de agitación volcánica para el pronóstico de compor-
tamiento eruptivo a futuro y el enmarque de protocolos de respuesta es un
reto, especialmente durante periodos de agitación prolongada (años a
décadas) en los que algunos retos surgen desde la utilización de señales de
agitación para pronosticar la evolución de agitación a largo plazo y sus
eventuales consecuencias. Exploramos las limitantes de actuales enfoques
para la toma de decisiones y proveemos sugerencias acerca de cómo
pueden hacerse reformas a futuro dentro del marco holístico de gobern-
abilidad ante el riesgo de agitación volcánica. Investigamos los potenciales
beneficios que surgen por comunicación mejorada, y delimitando alertas
alrededor de escalas de tiempo para la toma de decisiones y los niveles de
alerta. Proponemos la necesidad de la cooperación a través de las fronteras
científicas tradicionales, una valoración más amplia del riesgo natural y una
mayor interacción de los sectores interesados.
Volcanic Unrest and Pre-eruptive Processes … 3

1 Introduction eruption were to ensue it may involve the erup-

tion of magma or may be non-magmatic and
Volcanic unrest is a complex multi-hazard phe- mainly driven by expanding steam and hot water
nomenon of volcanism. Although it is fair to (hydrothermal fluids) (Table 1). These conun-
assume that probably all volcanic eruptions are drums contribute significant uncertainty to
preceded by some form of unrest, the cause and short-term hazard assessment and forecasting of
effect relationship between subsurface processes volcanic activity and have profound impact on
and resulting unrest signals (geophysical or the management of unrest crises (e.g., Marzocchi
geochemical data recorded at the ground surface, and Woo 2007).
phenomenological observations) is unclear and While institutional and individual
surrounded by uncertainty (e.g., Wright and decision-making in response to this unrest should
Pierson 1992). Unrest may, or may not lead to promote the efficient and effective mitigation or
eruption in the short-term (days to months). If an management of risk, informed decision-making

Table 1 Summary of processes contributing to unrest signals in space and time, possible outcomes and hazards/impact
of unrest
Nature of processes Processes Signals Hazards/Impact Unrest
Outcome
Magmatic Magma and/or melt Seismicity, ground Ground Waning and
and/or volatile deformation, deformation, return to
migration (input, changes in potential shaking and rupture background
loss or ascent from fields, changes in and associated activity;
reservoir), chemical gas and/or ground infrastructure eruptive
differentiation, water chemistry, damage; water table activity
thermal convection, changes in heat flux, level changes; toxic (magmatic
thermal perturbation changes in volatile gas emissions, and/or
(heating or cooling), flux contamination of phreatic)
pore fluid migration ground water,
reservoir atmosphere and
rejuvenation, crops; edifice
crystallization and destabilization; toxic
other phase changes gas emissions
Tectonic/gravitational Faulting, changes in Waning and
local/regional stress return to
fields, edifice background
gravitational activity;
spreading, crustal eruptive
loading, pore fluid activity
migration (magmatic
and/or
phreatic)
Hydrothermal Fluid migration, Waning and
phase changes, return to
changes in background
temperature and/or activity;
pressure, chemical phreatic
changes, pore eruptive
pressure variations, activity
porosity and
permeability
changes (sealing),
host-rock alteration
Processes can act individually, in unison or in any combination
4 J. Gottsmann et al.

is fundamentally dependent on the early and volcanic unrest and pre-eruptive processes from
reliable identification of changes in the subsur- the scientific contributions generated by collab-
face dynamics of a volcano and their “correct” oration of ten partners in Europe and Latin
assessment as precursors to an impending erup- America. Dissecting the science of monitoring
tion. However, uncertainties in identifying the data from unrest periods at six target volcanoes
causative processes of unrest impact significantly in Italy (Campi Flegrei caldera), Spain (Tener-
on the ability to “correctly” forecast the ife), the West Indies (Montserrat), Mexico
short-term evolution of unrest. (Popocatepetl) and Ecuador (Cotopaxi) the con-
When a volcano evolves from dormancy sortium created strategies for (1) enhanced
through a phase of unrest, scientific interpreta- monitoring capacity and value, (2) mechanistic
tions of data generated by this unrest relate to data interpretation and (3) identification of
(i) the processes behind unrest and their associ- eruption precursors and (4) crises stakeholder
ated surface signals, (ii) their potential future interaction during unrest.
spatio-temporal evolution (i.e., hydrothermal vs. The CASAVA project (2010–2014; Agence
phreatic vs. magmatic processes and their inten- nationale de la recherche, France; Understanding
sity) and (iii) their significance as precursors for and assessing volcanic hazards, scenarios, and
future eruptive phenomena. Scientific interpreta- risks in the Lesser Antilles—implications for
tions framed towards the governance of and decision-making, crisis management, and prag-
social responses to the risk implicit in the matic development; https://sites.google.com/site/
potential onset of an eruption focus on: (i) un- casavaanr/, last accessed 11-10-2016) imple-
derstanding the epistemic (relating to the limits mented an original strategy of multi-disciplinary
of existing knowledge) and aleatoric (relating to fundamental research on the quantitative assess-
the intrinsic variability of natural processes) ment of volcanic risk for the Lesser Antilles
uncertainties surrounding these data and their region with emphasis on Guadeloupe and Mar-
impact on decision making and emergency tinique. The aim of the project was to improve
management, (ii) the communication of these the capacity to anticipate and manage volcanic
uncertainties to emergency managers and the risks in order to reduce reactive ‘repairing’
citizens at risk, and (iii) understanding how best post-crisis solutions and promote the emergence
to manage evolving crises through the use of of a society of proactive volcanic risk prevention
forecasted scenarios. in case of a future eruption. Part of this was
achieved via a forensic analysis of past crises,
described here.
2 Motivation The STREVA Project (2012–2018 funded by
the UK Natural Environment and Economic and
The analysis presented in this chapter synthesises Social Research Councils; www.streva.ac.uk)
wider results and experiences gained in three was designed as a large interdisciplinary project
major research consortia with focus on volcanic to develop new means to understand how vol-
hazards and risks: (1) The VUELCO project, canic risk should be assessed and framed. It uses
(2) the CASAVA project, and (3) the STREVA the ‘forensic’ interdisciplinary analysis of past
project. volcanic eruptions in four settings to understand
The European Commission funded VUELCO the key drivers of volcanic risk. The aim is to use
project (2011–2015; “Volcanic unrest in Europe this analysis to generate future plans that will
and Latin America: Phenomenology, eruption reduce the negative consequences of future
precursors, hazard forecast, and risk mitigation; eruptions on populations and their assets.
www.vuelco.net) focused on multi-disciplinary STREVA works closely with partners in the
research on the origin, nature and significance of Caribbean, Ecuador and Colombia, focussing the
Volcanic Unrest and Pre-eruptive Processes … 5

forensic analysis on long-lived eruptions of whether (in hindsight) “correct” or “false” fore-
Soufrière Hills Volcano (Montserrat) and Tun- casts are issued to suggest there could be an
gurahua (Ecuador) and shorter duration eruptions imminent eruption are among the central ques-
of La Soufrière (St. Vincent) and Nevado del tions that need answering as soon as unrest is
Ruiz (Colombia). The focus of the ‘forensic detected.
analysis’ process in the STREVA project has The cost of scientific uncertainty regarding the
been to understand the key drivers of risk and causes and outcome of volcanic unrest may be
resilience during long-lived volcanic crises. substantial not only in terms of direct or indirect
Nonetheless the analysis of the initial phases of financial implications such as explored in
activity from these eruptions provide some Sect. 5, but also regarding knock-on (secondary)
insights into the acute uncertainties of unrest and effects such as public trust in the accuracy or
the social, political and scientific consequences inaccuracy of scientific knowledge, public per-
of that uncertainty. ception of the relationships between signals of
unrest and volcanic risk and future public com-
pliance with orders to evacuate or improve pre-
3 Volcanic Unrest: Scientific paredness in the medium to long term.
and Social Context A multitude of subsurface processes may
contribute to unrest signals and some are sum-
Volcanic unrest can be defined in a scientific marised in Table 1. Not all processes are
context: “The deviation from the background or pre-eruptive and the challenge lies in deciphering
baseline behaviour of a volcano towards a the causes of unrest with a view to establish early
behaviour or state which is a cause for concern in on in a developing crises whether a volcanic
the short-term (hours to few months) because it system develops towards a state where an erup-
might prelude an eruption” (Phillipson et al. tion may ensue. Whether or not unrest leads to
2013). The term “eruption” in the context of a eruption depends on many parameters. In general
possible unrest outcome could either relate to a the main concern during volcanic unrest lies with
magmatic or non-magmatic (phreatic or the potential for a magmatic eruption. For this to
hydrothermal) origin including the possible occur magma must rise from depth and break
evolution from phreatic to magmatic activity or through the surface. The dilemma for scientists is
an alternation or mix between the two (e.g., that magma movement does not create uniquely
Rouwet et al. 2014). In a social context, these attributable unrest signals and does not neces-
concerns might necessitate additional preparatory sarily lead to eruption (Table 1). For example,
or contingency actions in response to the unrest seismicity and ground uplift, both common
phenomena or the preparation for an eruption indicator of unrest, may be induced by the
given that the organisation and preparedness of replenishment of a magma reservoir, the ascent
communities and those who manage them will be of magma towards the surface or the redistribu-
perturbed and even changed in the context of tion of aqueous fluids and fluid phase changes
significant uncertainty (Barclay et al. 2008 and (see Salvage et al. 2017; Hickey et al. 2017;
next section). Mothes et al. 2017 for examples from VUELCO
volcanoes). Similarly, an increase in the gas and
heat flux (Christopher et al. 2015) at the surface
4 Challenges and Key Questions may be induced by magmatic or hydrothermal
Relating to Volcanic Unrest processes and even tectonic stress changes have
also been shown to trigger such behaviour (e.g.,
4.1 Wider Perspective Hill et al. 1995). In fact, non-magmatic eruptions
are associated with significant hazards and have
Whether or not unrest results in eruption, either or could have caused fatalities in the past such as
of magmatic or non-magmatic origin, and for example Bandai in 1888 (Sekiya and Kikuchi
6 J. Gottsmann et al.

1890), Te Maari Tongariro in 2012 (e.g., Jolly The fundamental limitation for volcanologists
et al. 2014) and recently at Ontake in 2014 (e.g., is that it is not possible to directly observe cau-
Maeno et al. 2016). Many unrest processes sative processes at depth. Thus interpretations of
contribute to non-eruptive secondary hazards these drivers rely on the secondary interpretation
such as flank instability and collapse (e.g. Reid of observable signals associated with those pro-
2004). cesses (Salvage et al. 2017) or the reproduction
of interpreted processes via laboratory experi-
ments (Wadsworth et al. 2016). In addition,
4.2 Uncertain Causes and Uncertain many volcanic processes are intrinsically
Effects non-linear and characterized by a chain-link
reaction such that minor variations of some
Substantial uncertainties surround both the uncertain parameters might have ultimately sig-
interpretation of the drivers of unrest and the nificant consequences on the eruptive outcome.
assessment of the potential evolution and out- Such non-linear processes coupled with epis-
come of unrest. Critical questions include: Will temic and aleatoric uncertainties are complex to
an eruption ensue? If so, will it occur in the understand and model. This chapter analyses
short-term (days to months) or long-term (years some key examples across the three aforemen-
to decades)? What will be the nature and inten- tioned projects where issues around the limits of
sity of the eruption (magmatic vs. phreatic)? volcanic knowledge exacerbated risk and makes
In the case of magmatic unrest, magma ascent suggestion for how future improvements can be
towards the surface can lead to a magmatic made.
eruption with potential for the formation of lava
flows, pyroclastic flows, lahars, ash-fall and
ballistics. These processes impact the proximal 4.3 The Hazard and Risk Interface
(few tens to hundreds of meters), medial (kilo-
meters) and distal (tens of kilometres or more) Scientific Challenges
areas around the volcano. Conversely unrest In the light of the above, from a scientific point
driven by sub-surface hydrothermal activity may of view the early identification of the cause of
peak in a phreatic eruption and while impacted unrest and its likely outcome and evolution is
areas are rather proximal to the volcano, associ- pivotal for effective and efficient risk assessment,
ated ballistics and dilute pyroclastic density risk management and the design of mitigation
currents triggered by laterally-directed explo- efforts. In order to address the key scientific
sions and emplacement of a debris avalanche question of whether unrest is a prelude to
from a partial edifice collapse can lead to an imminent eruption or whether it will wane after
anomalously high loss of lives as recently some time without eruption several questions
shown by the September 27, 2014 Mount require answering first (note, that the list is not
Ontake eruption, the deadliest eruption in exhaustive):
more than 100 years in Japan (e.g. Maeno et al.
2016). • Is the anomalous behaviour unambiguously
The challenge, however, is to identify and indicative for a change in the volcano’s
discriminate signals that are indicative of reacti- behaviour and for a deviation from its back-
vation leading towards a major expulsion of ground state?
magmatic material from those associated with a • How reliable is the assessment of unrest as a
slight deviation from background levels and prelude to eruption, particularly in the
potential waning of unrest phenomena (Table 1). absence of data on past events?
Volcanic Unrest and Pre-eruptive Processes … 7

• What are the mechanistic processes at depth what is the likelihood that the eruption
leading to observed unrest signals? (a) will have its paroxysmal phase in the first
• Are monitoring signals indicative of mag- 24 h of eruption (42% of eruptions do,
matic, hydrothermal or tectonic unrest? according to Siebert et al. (2015)); or (b) will
• Can the unrest be caused by perturbations and have a more progressive escalation over
changes in the host-rock properties (e.g. several months that will culminate in a
porosity, permeability, mechanical properties) paroxysm; or (c) will be characterised by
rather than by distinct endogenic processes of peaks in activity separated by more or less
hydrothermal or magmatic origin? long-lasting pauses or strong decline of
• What are the uncertainties surrounding mon- activity preceding another rapid increase and
itoring signals and inferred sub-surface pro- peak of activity?
cesses (see Hickey et al. 2017 and Salvage
et al. 2017)? Societal Challenges
• Do secondary processes (e.g. hydrothermal At the same time, the political, sociological,
system perturbation, meteorological forcing) cultural and economic (grouped here under the
modify primary signals from deeper-seated term ‘societal’) implications from unrest need
magmatic processes? addressing in order to respond appropriately to
• What are the consequences of signal modifi- the emerging natural hazard (Wynne 1992) Here
cation for the assessment of the we provide a (non-exhaustive) list of questions
process-to-signal-to-outcome causal link? for risk managers and/or politicians in the context
• Does one follow a deterministic or proba- of risk governance during volcanic unrest:
bilistic approach for observations and fore-
casting (e.g., Hincks et al. 2014; Aspinall and • What is the best-practice to provide maxi-
Woo 2014; Rouwet et al. 2017)? mum response time, while minimizing vul-
• What is the likelihood of a specific eruptive or nerability and optimizing the cost/benefit
non-eruptive scenario to manifest (e.g., Bar- ratio (see Fig. 1) of mitigation actions in a
tolini et al. 2017)? developing unrest crises?
• Which types of eruptions did the volcano • What is the best practice to issue or raise an
produce in the past? alert?
• If an eruption is to occur, what is its likely • When and how to decide to raise an alert and
nature: magmatic, or phreatic or a mix? to take action?
• How much lead-time before eruption is there • What are the potential (legal) consequences
based on previous experience; how much of a false positive or false negative (see
lead-time is there in the absence of previous Table 2 and Bretton et al. 2015)?
experience? • What are the consequences of a true positive
• Which eruptive or non-eruptive unrest epi- (Table 2)?
sodes at analogue volcanoes can provide • What is the basis for raising an alarm: the
clues for the interpretation of signals and outcomes of unrest (e.g., instability of build-
forecasting of unrest evolution and outcome ings due to ground deformation or seismicity;
(e.g., Sheldrake et al. 2016)? toxic degassing and environmental contami-
• What is the likely size of the eruption and the nation) or the potential for eruption?
associated hazards and risks and impacted • How to best disseminate what information on
area? unrest and its potential consequences, when,
• What is the temporal evolution of eruptive and via which communication vehicle(s) to
intensity once the eruption has started? i.e., the public?
8 J. Gottsmann et al.

Fig. 1 Cost-benefit relationship as a tool for decision- capital management theory presented in Brealey et al.
making. a In the context of volcanic unrest risk manage- (2011). b Cost (C) versus loss (L) model for volcanic risk
ment, actions of given quantity Q (for example, number of management (after Marzocchi and Woo 2007). If, in this
shelters or evacuees) are associated with costs in relation to decision-making framework, the expected expense (cost)
their expected benefits (expressed by a financial value). An for mitigation action is to be minimised, then action is
optimal relationship between costs of mitigation efforts required if the probability (p) of an adverse event to occur
and resultant benefits can be achieved when the difference exceeds the ratio between the cost of the action and the
between investment and benefit is greatest (shown by expected loss (L/C). See discussion for a wider appraisal of
stippled red line). The example is based on concepts of the challenges arsing from such an analysis

Table 2 Concept of Event forecast Event not forecast

successful and unsuccessful
forecasting Event occurs True positive False negative
(Type II error)
Event does not occur False positive True negative
(Type I error)

• How to account for uncertainty and the • In what context does this occur such as
diversity of expert opinions in deciding the political pressures, concurrent natural or other
alert level? hazards (pandemic, famine, cyclone, etc.)
Volcanic Unrest and Pre-eruptive Processes … 9

4.4 Cost-Benefit Analysis (CBA) 5 Global and Regional Context

of Volcanic Unrest
In the previous paragraphs, several questions
related to how to get both the scientific analysis 5.1 Unrest Durations
(‘what is going on?’) and the societal response and Characteristics
(‘how to respond?’) ‘right’. One measure
employed to quantify the economic consequences Phillipson et al. (2013) reviewed global unrest
of action or no-action under imminent threat and a reports of the Smithsonian Institution Global
tool for informed institutional decision-making is Volcanism Program (GVP) between January
the cost-benefit analysis (Marzocchi and Woo 2000 and July 2011 to establish the nature and
2007) whereby one aims to find a good answer to length of unrest activity, to test whether there are
the question: “Given an assessment of costs and common temporal patterns in unrest indicators
benefits related to risk mitigation efforts, which and to test whether there is a link between the
actions should be recommended?” Figure 1 shows length of inter-eruptive periods and unrest dura-
the concept of evaluating the optimum ratio tion across different volcano types.
between the cost and benefit of mitigation efforts Using available formation on unrest at 228
and provides a cost-benefit matrix for the design of volcanoes they defined unrest timelines to
action plans in response to a future [short-term in demonstrate how unrest evolved over time and
context of this chapter) adverse event of given highlight different classes of unrest including
probability (p) (Brealey et al. 2011; Marzocchi and reawakening, pulsatory, prolonged, sporadic and
Woo 2007)]. A critical issue in CBA is the ‘min- intra-eruptive unrest (see Fig. 2 for an example
imum value of a human life’, which we will not from Cotopaxi volcano). Statistical analyses of
discuss further here. The interested reader is the data indicate that pre-eruptive unrest (where
referred to, for example, Woo (2015) for further there is a causal link between unrest and an
details on this quantification. Another interesting eruption within the observation period) duration
point relates to what might be regarded as a ‘cost’ was different across different volcano types with
and a ‘benefit’ in a response to an unfolding unrest 50% of stratovolcanoes erupting within one
crisis with an uncertain outcome (see also Sect. 6). month of reported unrest. The median average

Fig. 2 Timeline of reported anomalous activity at months), but Cotopaxi entered an eruptive phase in
Cotopaxi volcano (Ecuador) in 2001/2002. This period August 2015 after a short-period of renewed unrest
of pulsatory unrest lasted for more than 3 years with a activity starting in April 2015 (see Mothes et al. 2017 for
heightened level of activity in 2001 and 2002. The unrest details). The data shown in the graph are from Phillipson
did not lead to an eruption in the short-term (weeks to et al. (2013)
10 J. Gottsmann et al.

duration of pre-eruptive unrest at large calderas satellite remote sensing data in the available
was about two months, while at shield volcanoes reports. Recently Biggs et al. (2014), addressed
a median average five months of unrest was the latter and systematically analysed 198 vol-
reported before eruptive activity. The shortest canoes with more than 18 years of satellite
median average duration is reported for complex remote sensing deformation data for their defor-
volcanoes where eruptive unrest was short at mation behaviour. 54 volcanoes that showed
only two days. Overall there appears to be only a deformation also erupted during the observation
very weak correlation between the length of the period. Their analysis does not imply any causal
inter-eruptive period and pre-eruptive unrest link, or even a temporal relationship between any
duration. This may indicate that volcanoes with specific eruptions and episodes of deformation
long periods of quiescence between eruptions and is hence not directly comparable to the
will not necessarily undergo prolonged periods causal and predictive analysis by Phillipson et al.
of unrest before their next eruption (Fig. 3). (2013). However, given that 46% of deforming
Phillipson et al. (2013) found statistically rele- volcanoes erupted while 94% of non-deforming
vant information only from reports of anomalous volcanoes did not erupt provides “strong evi-
seismic behaviour, most other monitoring signals dential worth of using deformation data as a trait
are either not recorded or not reported as unrest association with eruption” (Biggs et al. 2014).
criteria. The authors reported a noteworthy lack It is important to note that exploitable records
of geodetic data/information and in particular on volcanic unrest are limited and the available

Fig. 3 Comparison between the inter-eruptive period (n = 60). The null hypothesis (“the UD is independent of
(IEP) and unrest duration (UD) from the data set the IEP”) is hence statistically acceptable when consid-
presented in Phillipson et al. (2013). a shows entire data ering the entire data set. Considering the subset of
set (n = 118) b shows a subset of the data for clarity of pre-eruptive unrest, however, the statistical tests do not
inter-eruptive periods <150 years. The p-values of the provide enough evidence to fully accept the null hypoth-
Pearson’s correlation test are p = 0.93 for the entire data esis since the associated p-value of 0.20 might
set, p = 0.60 for the subset of non-eruptive unrest hint towards some weak correlation between the two
(n = 58) and p = 0.20 for the subset of eruptive unrest variables
Volcanic Unrest and Pre-eruptive Processes … 11

data sets are far from complete. Key issues are example disease or starvation as a result of vol-
the lack of or poor instrumentation at most vol- canic activity. Beyond increased human vulner-
canoes, the lack of reporting by observers par- ability, the direct and indirect financial impacts
ticularly if an unrest turns out to be minor and from volcanic activity can be immense as
without immediate consequences, and the lack of demonstrated by the relatively small-scale erup-
integrating unrest data from satellite remote tion of Iceland’s Eyjafjallajökull volcano in April
sensing. The GVP generally lacks the post-facto 2010 and the associated air travel disruption.
integration of unrest indicators from satellite- This eruption demonstrated the vulnerability of
remote sensing data (e.g., Fournier et al. (2010) modern infrastructure to volcanic hazards on an
and Biggs et al. (2014) for deformation and Carn unprecedented scale with losses to the aviation
et al. (2011) for degassing). In this respect, it is industry alone at a minimum of US$2.5 Billion
vitally important to recognise and support ini- (European Commission 2010).
tiatives to collate and exploit worldwide volcano Equally there are social, political and financial
monitoring data such as for example the implications for “false positives” related to vol-
WOVOdat project (Venezky and Newhall 2007). canic unrest. In these instances actions are taken in
Only by significantly increasing the knowledge- response to an imminent threat, which then did not
base on the spatial and temporal evolution of the manifest. In the case of volcanic unrest the
unrest-eruption relationship can we embark on imminent threat is generally defined as a volcanic
statistically sound exploitations of the data with a eruption, although the multi-hazard nature of vol-
potential to improve forecasting capabilities early canic unrest (e.g., ground shaking, ground uplift or
on in developing unrest crises. subsidence, ground rupture, ground instability,
toxic gas emissions, contaminated water supplies)
and possibly ensuing eruptive activity (magmatic
5.2 Socio-Economic Contexts vs. phreatomagmatic vs. phreatic) makes the defi-
nition of ‘imminent threat’ rather complex.
The Wider Perspective Although there is little systematic gathering
Nowadays, about 800 million people live on or in and synthesis of data relating to financial or
direct vicinity of active volcanoes (Brown et al. social losses associated with these episodes there
2015). The overwhelming majority of this pop- are some well-documented analyses. Examples
ulation lives in low and middle income countries include:
(countries with an annual gross national income
per capita of less than US$12,700) including the (1) On Guadeloupe in the French West Indies a
focus area of the VUELCO, STREVA and major evacuation over a period of 4 months
CASAVA projects: the wider Latin American in excess of 70,000 individuals was initiated
(LA) region extending from Mexico, through in 1976, as a result of abnormal levels of
Central America and the Caribbean to South volcanic seismicity and degassing (see also
America. This region hosts about 330 Holocene next section). The estimated cost of the
volcanic centres compared to 84 in Europe and unrest was about US$340 Million at the 1976
one quarter of the reported global fatalities exchange rate (data compiled using Lepointe
attributed to volcanic events occurred there 1999; Tazieff 1980; Blérald 1986; Baunay
(Global Volcanism Program 2013). 1998; Kokelaar 2002; Annen and Wagner
Volcanic disasters are among the least audited 2003), which translates to more than US$ 1.2
of all natural disasters and therefore our knowl- Billion at the time of this writing (July 2016).
edge on the impact of volcanic activity beyond At the time the cost equaled to ca. 60% of the
claiming lives is largely incomplete (Benson Gross National Product of the Guadeloupe
2006; Auker et al. 2013). Huge uncertainty sur- economy (Blérald 1986). 90% of these costs
round estimates for indirect losses from for were incurred by the costs of the evacuation,
12 J. Gottsmann et al.

and the costs associated with the rehabilita- “proportion of evacuees who would have owed
tion and salvage of the economy in Guade- their lives to the evacuation, had there been a
loupe after the evacuation. major eruption, was substantial” (Woo 2008).
(2) Unrest at Rabaul volcano in Papua New The CASAVA project undertook an exhaustive
Guinea (an LDC) between 1983 and 1985, hindsight analysis of the process of scientific
had significant adverse implications for both decision-making for the unrest and eruptive
the private and public sectors. Considerable crisis of 1976–1977 at La Soufrière de Guade-
economic costs were incurred, estimated at loupe. The crisis caused significant hardships
over US$22.2 Million at the 1984 rate of and loss of livelihood for the evacuated popu-
exchange although an eruption did not occur lation and the whole society in Guadeloupe as a
until 10 years later (Benson 2006). result of controversial crisis management asso-
(3) Evacuation and rehousing of 40,000 inhabi- ciated with a forecast of a major magmatic
tants of the Pozzuoli area in the Campi Fle- eruption that did not occur (false positive)
grei volcanic area of Italy resulted as a (Feuillard et al. 1983; Fiske 1984; Komorowski
response to intense seismicity and ground et al. 2005; Hincks et al. 2014). Given the evi-
uplift in the early 1980s. Although decision- dence of continued escalating pressurisation and
makers did not release notice that this was in the uncertain transition to a devastating mag-
part due to the threat from an imminent matic eruption, authorities declared a 4-month
eruption (see also Sect. 4.3.2), it is true that evacuation of ca. 70,000 people on August 15,
the re-location of these inhabitants moved 1976 that provoked severe socio-economical
them from the area of highest threat in the consequences for months to years thereafter.
event of an eruption. At the time there was This evacuation is still perceived as unnecessary
no agreement amongst scientists as to the and reflecting an exaggerated use of the “prin-
cause of the unrest (Barberi et al. 1984) and ciple of precaution” on behalf of the
the scientific discussion as to the cause of government.
these events is still ongoing more than However, some level of risk governance (i.e.
30 years after the crisis. evacuation of the most exposed area) was justi-
fied in hindsight given the persistent ashfalls and
The following paragraphs focus on two examples environmental contamination from acid degas-
of short-term and long-term volcanic unrest cri- sing as well as the hazards from a series of
ses response and provide more detailed insights non-magmatic eruptions (e.g., pyroclastic flows
into the volcanic risk governance in two different from laterally directed explosions, partial edifice
jurisdictions. collapse, mudflows) (Komorowski et al. 2005;
Hincks et al. 2014).
Short-Term Crisis Example: The 1976– The (in hindsight) erroneous identification of
1977 La Soufrière of Guadeloupe Unrest the presence of ‘fresh glass’ in the ejecta and its
The unrest on Guadeloupe culminated in a series interpretation as evidence of the magmatic origin
of explosive eruptions of hot gas, mud and rock of the unrest and thus of its possible outcome, led
(termed phreatic eruption) without the direct to a major controversy amongst scientists that
eruption of magma before waning in 1977 was widely echoed in the media. Lack of a
(Feuillard et al. 1983; Komorowski et al. 2005; comprehensive monitoring network prior to the
Hincks et al. 2014). Fortunately no fatalities crisis, limited knowledge of the eruptive history,
were caused by the activity. Had the unrest on and living memory of past devastating eruptions
Guadeloupe led to a magmatic eruption, then the in the Lesser Antilles contributed to a high
cost of the unrest would have likely been neg- degree of scientific uncertainty and a publically-
ligible. Although the precautionary evacuation expressed lack of consensus and trust in available
caused a substantial economic loss with severe expertise. Consequently analysis, forecast, and
social consequences, it is acknowledged that the crisis response were highly challenging for
Volcanic Unrest and Pre-eruptive Processes … 13

scientists and authorities in the context of esca- The long-lived volcanic crisis of the Soufrière
lating and fluctuating activity and societal pres- Hills Volcano is probably one of the most written
sure. The high uncertainty about a so-called about volcanic eruptions, encompassing a wide
“unequivocal” impending disaster fostered a variety of perspectives, scientific, social-
binary zero-sum strongly opinionated approach scientific and personal, in that writing. As a
in the scientific discourse. The public debate thus consequence of the activity on the island of
became polarized on issues of opposing “truths” Montserrat a population of over 10,500 was
served by contrasted scientific expertise rather reduced to just 2850 (the population has since
than on how science could help constrain epis- risen to 4922 [2011 census], Hicks and Few
temic and aleatoric uncertainty and foster 2015). At the onset of eruption (1995) an
improved decision-making in the context of assessment of risk existed (Wadge and Isaacs
uncertainty (Komorowski et al. 2017). This sit- 1988) but was not acted on or acknowledged by
uation acted as an ideal crucible to fuel a the authorities, and so preparedness was low,
media-hyped controversy on the crisis and its exacerbated by the recent passage of Hurricane
management. A recent retrospective Bayesian Hugo (1989) which had caused 11 fatalities and
Belief Network analysis of this crisis (Hincks rendered 3000 homeless. Governance on
et al. 2014) demonstrates that a formal evidential Montserrat was reforming in the wake of the
case could have been made to support the economic and social crisis induced by the hurri-
authorities’ concerns about public safety and cane (Wilkinson 2015). The protracted uncer-
decision to evacuate in 1976. tainty in the early stages of the eruption coupled
As part of the CASAVA project we conducted with a lack of coherence in governance between
focus group interviews, issued questionnaires, the UK and local governments lead to the pro-
and ran role playing games with the population tracted evacuation of 1300 people in temporary
currently living in areas potentially threatened by public shelters, which suffered from overcrowd-
renewed unrest and eruptive activity from La ing, lack of privacy, poor sanitation and lack of
Soufrière, (be it magmatic or non-magmatic). We access to good nutrition. Ultimately, this led to a
found that the current population’s risk percep- partial disregard for evacuation advice and a
tion increases to a level of preparing to evacuate strong pulse of outwards migration. In the longer
chiefly on the basis of the timing and nature of term, the long-lived volcanic eruption has acted
scientific information issued publically by the to exaggerate pre-existing vulnerabilities in the
volcano observatory. This implies that the pop- local population (Hicks and Few 2015).
ulation is prone to self-evacuate ahead of any The early stages of the current Tungurahua
official evacuation order given by the authorities (Ecuador) eruptive episode that started in 1999
in charge of civil protection and crisis response. typify a further challenge for the management of
unrest prior to or between surface activity at the
Long-Term Crises Examples: Soufrière early stages of a volcanic crisis. Initially the local
Hills (Montserrat) and Tungurahua population were evacuated by a compulsory
(Ecuador) evacuation order but when the initial phases
The forensic analyses of the STREVA project proceeded more slowly than had been expected
have focussed on the integration of new by local authorities and communities, civil unrest
social-science based understandings of popula- and disturbance happened with the re-occupation
tion response and recovery with the scientific by force and ultimately abandonment of the
insights prompted by these long-lived eruptions. evacuation order. These arose from the acute
This has similarities with the ‘FORIN’ approach economic and social pressures visited on the
advocated by the International Program on Inte- population by the evacuation (Mothes et al.
grated Risk for Disaster Reduction (Burton 2015). Subsequently, the response of the moni-
2010). In this description we focus particularly toring organisation to these pressures represents a
on the initial stages of the eruptions. new archetype for collaborative monitoring and
14 J. Gottsmann et al.

management of restive volcanoes (Mothes et al. precursor, its exploitability regarding forecasting
2015; Stone et al. 2014). The growth of trust, and of potential eruptive behaviours and framing of
attempts to maximise resilience in the face of response protocols (e.g., CBA) remains weak for
repeated unrest episodes provides strong evi- a number of reasons:
dence for collaborative approaches to risk man-
agement (Few et al. 2017). Nonetheless tensions (1) The scientific interpretation of volcanic
still exist, largely arising from our current inca- unrest is surrounded by substantial uncer-
pacity to predict the intensity or magnitude of tainty, ambiguity and ignorance (Stirling
eruptions from signals relating to new unrest. 2010) regarding causes and eventual out-
There can be problems in this risk system come. Since the contributing subsurface
implicit in anticipating the ‘maximum expected’ processes cannot be directly observed, vol-
outcome from unrest. canic unrest is likely among the least
understood phenomena in volcanology for a
variety of reasons:
6 Discussion (i) Incomplete knowledge of the mechanistic
processes and their dynamic behaviour over
6.1 The Caveats of Volcanic Unrest time within a magma reservoir and its sur-
Response roundings (host-rock, hydrothermal system,
meteoric recharge, local and regional
Managing volcanic unrest episodes is extremely structural context) that trigger the geo-
complex and challenging due to the multi-hazard physical, geochemical and geodetic signals
nature of unrest. The risks to be assessed and recorded at the surface during unrest peri-
mitigated include both those associated with the ods (Table 1).
unrest itself as well as those from the potential (ii) Consequently, the interpretation, of the
future eruptive activity. Whilst ground deforma- departure of monitoring signals from a
tion, seismicity, thermal flux or anomalous long-term baseline level or in the absence of
degassing are indicators of possible future baseline data a crescendo or decrescendo of
activity these phenomena also pose significant signals collected during periods of unrest are
immediate threats to population, infrastructure often ambiguous or non-unique. While this
and other assets in affected areas during the can in practice be addressed in models
unrest. through epistemic and aleatoric uncertainties,
From a scientific point of view, hazard ambiguities in the interpretation will remain.
assessment relating to eruptive activity has made (2) Ambiguity, uncertainty and ignorance (Stir-
considerable progress in recent years partly ling 2010) have impact on probabilistic
through the deployment of increasingly powerful forecasting of duration, spatio-temporal
computational models and simulation capabilities evolution, causal relationship between
(e.g., Esposti Ongaro et al. 2007; Manville et al. sequential events and outcomes of unrest
2013) as well as through advances in the devel- episodes (see Sandri et al. 2017) and on
opment of probabilistic eruption forecasting tools remedial actions to mitigate current and
(e.g., Marzocchi et al. 2008; Aspinall 2006; future adverse effects. Uncertainties in the
Aspinall and Woo 2014) and improvements to decision-making process may give rise to
fundamental understandings of the root drivers of “false alerts” (i.e., false positives; see
changing activity (e.g., Cashman and Sparks Table 2) and actions by civil protection with
2013). adverse impacts on the compliance of
Despite these crucial advances for short-term affected communities in future unrest events.
eruption forecasting, the knowledge-base on (3) Lack of globally accepted and standardised
volcanic unrest, its significance as an eruption approach for the terminology, methodology,
Volcanic Unrest and Pre-eruptive Processes … 15

criteria, protocols and best practice 6.2 Some Ways Forward

employed to evaluate and respond to vol-
canic unrest by different stakeholders such The issues identified above can contribute less
as academia, volcano observatories and the optimal unrest response and risk mitigation
Civil Protection Agencies. This absence of actions. Although there are other strong con-
commonly recognised standards can result tributors to societal vulnerability, we have shown
in the critical issue of managerial risk vul- that scientific uncertainty combined with a lack
nerability (i.e., standard equivocality’ after of social awareness and preparedness does act to
Bretton et al. 2015). It also often impacts increase the vulnerability of a society to haz-
negatively on the effectiveness of commu- ardous unrest phenomena with possibly adverse
nication between stakeholders, hinders or outcomes. Here, we propose future avenues
delays effective and efficient decision- which can form part of a Risk Governance
making processes and hampers the dia- Framework (IGRC 2017; Fig. 4) including
logue among members of scientific, gov- research that could gather critical evidence for
ernmental and civil communities (De la some of the key drivers of decisions that result in
Cruz-Reyna and Tilling 2008). However, it adverse outcomes for affected populations in the
is important to note that internationally- face of an unrest crisis. Such research could also
defined standards should not be rigidly contribute to the analysis of and identification of
imposed irrespective of local, cultural, key targets for future research in volcanology
political and social practices (e.g., Bretton and the social sciences.
et al. 2015; IAVCEI 2016).
(4) Globally, there is no commonly accepted (a) Cost-Benefit Analysis
and standardised denominator between
those that provide and those that receive CBA, where the economic impacts of different
scientific advice regarding the level of decisions are quantified, can be difficult at the
appropriate scientific complexity to be unrest hazard and risk interface. The case studies
considered. This may hamper a wider dis- presented here demonstrate that intangible assets
course on scientific and technological such as social and cultural cohesion and capital
advances in the quantification of unrest as well as trust (in the context of CBA analysis
phenomena and resultant uncertainties with this would be intentional trust in the sense of
other stakeholders. From the scientist’s Dasgupta (1988); i.e., the subjective probability
perspective this may generate the notion assigned to compassionate action by an individ-
that the public, administrators, mass media ual or a group of individuals) between different
and governmental entities do not appreciate stakeholders can have a strong impact on indi-
the “excellence of the science” behind vidual and institutional vulnerabilities during
unrest characterisation and use the inherent crises.
uncertainty as a rationale to go into denial Analyses that include a wider range of defi-
over the hazards posed during unrest. From nitions and types of ‘costs’ and ‘benefits’ of
a sociological point of view, however, mitigation efforts (e.g. loss of empowerment, a
decision-making apparently prompted loss of cultural identity or cultural references),
solely by the present or likely volcanic informed by past experiences would facilitate a
hazards, does not account for local context discourse between different stakeholders. This
and can result in a lack of trust in either would entail the need to attribute a financial
scientific expertise or government repre- value to, for example, mental well-being, social
sentatives, (Johnson 1987; Haynes et al. networks and cohesions and would necessarily
2008; Christie et al. 2015; Komorowski trigger a wider discourse of the impacts of
et al. 2017). decision-making beyond the avoidance of ‘cost
16 J. Gottsmann et al.

protracted (several years or decades) unrest

requiring an above back-ground level of
long-term vigilance (e.g. yellow/vigilance level
for La Soufrière of Guadeloupe since 1999,
Komorowski et al. 2005; OVSG-IPGP 1999–
2016 or at the Campi Flegrei caldera since 1969;
Ricci et al. 2013). In such cases there are obvious
long-term strategies that could be developed to
improve social well-being and economic devel-
opment (e.g. developing resilient critical infras-
tructures such as roads, bridges, public electrical
water and sewer systems and communications
networks) that would significantly enhance the
quality of life for years of “peace time” from the
volcano while ensuring a more efficient crisis
response and recovery should the volcano erupt
Fig. 4 The International Risk Governance Council and impact the society.
(IRGC 2017) Risk Governance Framework adapted for
the specific case of volcanic unrest. Hazard and Risk
Pre-assessment—“peacetime framing” the hazard and risk (b) Improved communication
in order to provide a structured definition of the baseline
behaviour of the volcano and its consequences, of how the Open and multi-directional communication pro-
hazard and risk are framed by different stakeholders, of
cesses are of paramount importance in fostering
how the risk may best be handled, and of the thresholds to
be met or exceeded to declare a state of unrest. Hazard the development of a shared representation and
and Risk Appraisal—combining a scientific risk assess- understanding among all stakeholders of the
ment of the current unrest hazards (using for example a nature, magnitude, dynamics, and societal and
rating scheme of unrest intensity, e.g., Potter et al. 2015)
environmental consequences of unrest and its
and its probability with a systematic concern assessment
(of public concerns and perceptions) to provide the potential eruptive outcome on multiple spatio-
knowledge base for subsequent decisions in an emerging temporal scales (e.g. Barclay et al. 2008, 2015;
unrest crises. Risk Characterisation and Evaluation—in Komorowski et al. 2017). While communicating
which the scientific data and a thorough understanding of
this information in a timely and comprehensible
societal values affected by the risk are used to evaluate the
risk as acceptable, tolerable (requiring mitigation), or format is challenging, the evidence presented
intolerable (unacceptable). Risk Management—the here suggests that a continuous discourse is
actions and remedies needed to avoid, reduce transfer or needed between different stakeholders ideally
retain the unrest risk and risks from probable unrest
both before, during and after an unrest situation.
outcomes. Risk Communication—how stakeholders and
civil society understand the unrest risk and participate in The case studies presented here demonstrate that
the risk governance process. Risk Categorisation and part of this discourse should involve a discussion
Evaluation—categorising the knowledge about the about the appropriate scientific complexity in the
cause-effect relationships as either simple, complex,
communication between scientific and
uncertain or ambiguous. In the context of volcanic unrest
this may include the categorisation of the outcome of non-scientific stakeholders is essential, so that
unrest and probable future eruptive activity the information exchange is ‘useful, usable and
used’ (Aitsi-Selmi et al. 2016) and fit for the
to lives’. Ideally this discourse would be framed decision-making purpose to which it is intended
during ‘peace-time’ (i.e. not in response to an (Fischhoff 2013). Wider discourse could for
unfolding unrest crisis) and involve participation example include regular information bulletins
from a wide spectrum of scientific and societal from monitoring agents to the civil society and
stakeholders. The necessity to move beyond authorities, the development of scenario-based
circumscribed appraisal methods such as the approaches in simulation exercises involving the
CBA is also evident from the response to civil society and the wider appraisal of less
Volcanic Unrest and Pre-eruptive Processes … 17

tangible ‘assets’ (i.e., live stock or cultural cap- society to take better decisions when times of
ital) in risk governance efforts. impeding adversity arise.
Dialogues between those responsible for
monitoring hazards and those responsible for (d) Framing of warnings around
managing risk, as well as the communities at risk decision-making timescales and hazard level
cannot only help to understand the most impor-
tant aspects of scientific information to convey Typically changes in alert levels are strongly tied
but could also lead to an improved understanding to pre-determined changes in geophysical and
of the context into which emergency response geochemical signals or phenomenological
actions must be made (e.g. Christie et al. 2015), observations and have carefully worked out
and encourage citizens at risk to act on advice. In associated actions. In our case studies, difficulties
particular more systematic studies that analyse have arisen when the time-scale over which
the effectiveness of different techniques and mitigating actions can be taken is much shorter
strategies in achieving these goals would be than needed to implement mitigating actions
very useful (see Fearnley et al. 2017 for a recent such as evacuation or much longer than the
compilation). These efforts should help address timescale over which unrest or new eruptive
reluctance by the public to follow emergency- activity impacts on the population at risk. In the
response advice in an emerging unrest case of the former, lives or assets may be put at
crises. risk and in the case of the latter, possessions and
livelihoods can be negatively impacted with
(c) Wider natural risk appraisal repercussions on trust and political stability.
Managing decade or longer periods of protracted
In a similar vein, the implementation of advice moderate-level unrest amid significant epistemic
on volcanic risk could be more effective if it is and aleatoric uncertainty on its outcome consti-
considered in the context of other natural risks tutes major challenges for scientists, authorities,
and social challenges (e.g., Wilkinson et al. the population, and the media.
2016). By definition the onset of a volcanic The development of novel probabilistic for-
eruption involves the anticipation of impacts malism for decision-making could help reduce
from multiple hazards but the risk associated scientific uncertainty and better assist public
with volcanic hazards are often considered in officials in making urgent evacuation decisions
isolation, and as a low probability, high conse- and policy choices should the current and
quence hazard, ignored in advance of an unrest ongoing unrest lead to renewed eruptive activity.
crisis. This lack of dialogue and preparation has To improve decision-making around changing
been identified above as a strong contributor to alert or hazard levels, improved modelling efforts
tensions during unrest crises. Volcanic regions of the time-scales and pathways of population
only very rarely suffer solely from the impacts of mobilisation or actions (both as forward mod-
a single natural hazard (e.g. volcanic small-island elling and as analysis of past events) and better
developing states discussed in Wilkinson et al. understanding of the consequences of protracted
2016; Komorowski et al. 2017). Therefore unrest or eruptive activity on the vulnerabilities
methods that consider the multi-hazard context of affected populations (e.g. Few et al. 2017)
more clearly may ultimately help communities at could improve choices to be made in responding
risk cope with uncertainty in face of volcanic to changing or escalating activity as well as
hazards. This may be particularly the case, if they chain-link scenarios.
are able to identify ‘co-benefits’ during volcano Further, focussing on the time-scales associ-
“peace time” where preparedness or mitigation ated with the responses to unrest (from the time
measures yield benefits for more than one hazard taken to mobilise populations in an acute emer-
scenario (Wilkinson et al. 2016). This improve- gency, to the time-limits of tolerability of evac-
ment of social well-being is likely to allow the uation processes and finally the time-scales over
18 J. Gottsmann et al.

which services and livelihoods deteriorate in References

response to protracted unrest) could provide
important indicators for the time-scales over Aitsi-Selmi A, Blanchard K, Murray V (2016) Ensuring
which alert levels (and attendant actions) need science is useful, usable and used in global disaster
attention. In turn this perspective could inform risk reduction and sustainable development: a view
through the Sendai Framework lens. Palgrave Com-
scientific targets for improved forecasting, with
mun (2) doi:10.1057/palcomms.2016.16
strong effort expended to reduce uncertainty over Annen C, Wagner J-J (2003) The impact of volcanic
time intervals that match those most critical to eruptions during the 1990s. Nat Hazards Rev
effective societal action. 4:169–175
Aspinall WP (2006) Structured elicitation of expert
judgment for probabilistic hazard and risk assessment
in volcanic eruptions. In: Mader HM, Coles SG,
7 Conclusions Connor CB, Connor LJ (eds) Statistics in volcanology.
The geological society for IAVCEI, 15–30
Aspinall WP, Woo G (2014) Santorini unrest 2011–2012:
We have identified a number of scientific and
an immediate Bayesian belief network analysis of
sociological problems surrounding volcanic eruption scenario probabilities for urgent decision
unrest and have highlighted key aspects of risk support under uncertainty. J Appl Volcanol 3:12
governance at the interface between scientists, Auker MR, Sparks RSJ, Siebert L, Crossweller S, Ewert J
(2013) A statistical analysis of the global historical
emergency managers and wider societal stake- volcanic fatalities record. J Appl Volcanol 2:2
holders. We have in particular focussed on the Barclay J, Haynes K, Mitchell T, Solana C, Teeuw R,
issue of scientific uncertainty and its impact on Darnell A, Crosweller HS, Cole P, Pyle DM, Lowe C,
preparatory or contingency actions that might be Fearnley C, Kelman I (2008) Framing volcanic risk
within disaster risk reduction: finding ways for the
needed because relationships between and social and physical sciences to work together. In:
among individuals and social groups will be Liverman DGE, Pereria CPG, Marker B (eds) Com-
perturbed or even changed. Especially, during municating environmental geoscience. Geological
periods of protracted unrest (years to decades) Society Special Publication, 305
Barclay J, Haynes K, Houghton BF, Johnston DM (2015)
challenges arise from the exploitability of unrest Social processes in volcanic risk reduction. In:
signals to forecast long-term unrest evolution and Sigurdsson H et al (eds) Encyclopaedia of volcanol-
its eventual outcome. This impacts directly on ogy, 2nd edn. Elsevier, Amsterdam.
establishing the probability for and the timing Barberi F, Corrado G, Innocenti F, Luongo G (1984)
Phlegraean Fields 1982–1984: brief chronicle of a
and type of future eruptive behaviour as well as volcano emergency in a densely populated area. Bull
the definition of appropriate response protocols. Volcanol 47:175–185
To improve communication and trust between Bartolini S, Marti J, Sobradelo R, Becerril L (2017)
stakeholders as well as the framing of warnings Probabilistic e-tools for hazard assessment and risk
management. Adv Volcanol 1–15
around decision-making timescales and hazard Baunay Y (1998) Gestion des risques et des crises:
levels of unrest, we propose that bridging across l’exemple de deux crises d’origines volcaniques dans
traditional scientific boundaries, wider natural les Caraïbes: Montserrat 1995–1998 et la Guadeloupe
risk appraisal and broader stakeholder interaction 1976–1977. Mémoire de Maîtrise de Géographie,
Université de Paris 1, Panthéon-Sorbonne, Paris,
is needed. unpublished, 164 pp.
Biggs J, Ebmeier SK, Aspinall WP, Lu Z, Pritchard ME,
Acknowledgements The VUELCO project received Sparks RSJ, Mather TA (2014) Global link between
funding by the EC-FP7 program under grant agreement deformation and volcanic eruption quantified by
number 282759, the CASAVA project (ANR-09- satellite imagery: Nat Commun 5
RISK-002) was funded by the Agence Nationale de la Blérald APh (1986) Histoire éruptive de la Guadeloupe et
Recherche (France) and the STREVA Project de la Martinique du XVIIème siècle à nos jours,
(NE/J020052/1) was funded by the UK NERC and ESRC. Editions Karthala, Paris, 336 p
We acknowledge the useful insights and comments from Benson C (2006) Volcanoes and the economy. In: Marti J,
two anonymous reviewers. Ernst GGJ (eds) Volcanoes and the environment.
Cambridge University Press, pp 440–467.
Volcanic Unrest and Pre-eruptive Processes … 19

Brealey R, Myers S, Allen F (2011) Principles of Few R, Armijos T, Barclay J (2017) Living with Volcan
corporate finance. McGraw, 872 p Tungurahua: the dynamics of vulnerability during
Bretton R, Gottsmann J, Aspinall WP, Christie R (2015) prolonged volcanic activity. Geoforum 80:72–81
Implications of legal scrutiny processes (including the Fischhoff B (2013) The sciences of science communica-
L’Aquila trial and other recent court cases) for future tion. In: Proceedings of the National Academy of
volcanic risk governance. J Appl Volcanol 4:18 Sciences of the United States of America, 110
Brown SK, Auker MR, Sparks RSJ (2015) Populations (Supplement 3), 14033–14039. doi:10.1073/pnas.
around Holocene volcanoes and development of a 1213273110
population exposure index. In: Loughlin SC et al Fiske R (1984) Volcanologists, journalists, and the
(eds) Global volcanic hazards and risk ‘populations concerned local public: a tale of two crises in the
around volcanoes and the development of a population eastern Caribbean. In: Boyd F (ed) Explosive volcan-
exposure index. Cambridge ism: inception, evolution and hazards. Studies in
Burton I (2010) Forensic disaster investigations in depth: geophysics. National Academy Press, Washington DC,
a new case study model. Environ Mag 52(5):36–41 pp 170–176
Cashman KV, Sparks RSJ (2013) How volcanoes work: a Fournier TJ, Pritchard ME, Riddick SN (2010) Duration,
25 year perspective. Geol Soc Am Bull. doi:10.1130/ magnitude, and frequency of subaerial volcano defor-
b30720.1 mation events: new results from Latin America using
Carn SA, Froyd KD, Anderson BE, Wennberg P, InSAR and a global synthesis. Geochem Geophys
Crounse J, Spencer K, Dibb JE, Krotkov NA, Brow- Geosyst 11:Q01003. doi:10.01029/02009GC002558
ell EV, Hair JW, Diskin G, Sachse G, Vay SA (2011) Global Volcanism Program (2013) Volcanoes of the
In situ measurements of tropospheric volcanic plumes World, v. 4.5.5. In: Venzke E (ed) Smithsonian
in Ecuador and Colombia during TC4. J Geophys Res Institution. Downloaded 04 May 2017. http://dx.doi.
Atmos, 116 org/10.5479/si.GVP.VOTW4-2013
Christie R, Cooke O, Gottsmann J (2015) Fearing the Haynes K, Barclay J, Pidgeon NF (2008) The issue of
knock on the door: critical security studies insights trust and its influence on risk communication during a
into limited cooperation with disaster management volcanic crisis. Bull Volcanol 70:605–621
regimes. J Appl Volcanol 4:19 Hicks A, Few R (2015) Trajectories of social vulnerabiltiy
Christopher TE, Blundy J, Cashman K, Cole P, during the Soufriere Hills volcanic crisis. J Appl
Edmonds M, Smith PJ, Sparks RSJ, Stinton A Volcanol 4:10
(2015) Crustal-scale degassing due to magma system Hill DP, Johnston MJS, Langbein JO, Bilham R (1995)
destabilization and magma-gas decoupling at Sou- Response of Long Valley caldera to the Mw = 7.3
frière Hills Volcano, Montserrat: Geochemistry. Geo- Landers, California, Earthquake. J Geophys Res 100
phys Geosyst 16(9):2797–2811 (B7):12985–13005
Dasgupta P (1988) Trust as a commodity. In: Gambetta D Hickey J, Gottsmann J, Mothes P, Odbert HM, Prutkin I,
(ed) Trust: making and breaking of cooperative Vajda P (2017) The ups and downs of volcanic unrest:
relations. Blackwell, Oxford, 49–72 insights from integrated geodesy and numerical mod-
De la Cruz-Reyna S, Tilling RI (2008) Scientific and elling. Adv Volcanol 1–17
public responses to the ongoing volcanic crisis at Hincks T, Komorowski J-C, Sparks RSJ, Aspinall WP
Popocatépetl Volcano, Mexico: importance of an (2014) Retrospective analysis of uncertain eruption
effective hazards-warning system. J Volcanol precursors at La Soufrière volcano, Guadeloupe,
Geotherm Res 170:121–134 1975–77: volcanic hazard assessment using a Baye-
Esposti Ongaro T, Cavazzoni C, Erbacci G, Neri A, sian Belief Network approach. J Appl Volcanol 3:3
Salvetti MV (2007) A parallel multiphase flow code IAVCEI task force on crises protocols (2016)
for the 3D simulation of volcanic explosive eruptions. Toward IAVCEI guidelines on the roles and respon-
Parallel Comput 33:541–560 sibilities of scientists involved in volcanic hazard
European Commission (2010) Press Release 27. April, evaluation, risk mitigation, and crisis response. Bull
2010). http://ec.europa.eu/commission_2010-2014/ Volcanol 8(31). doi:10.1007/s00445-00016-01021-
kallas/headlines/news/2010/04/doc/information_note_ 00448
volcano_crisis.pdf IRGC (2017) https://www.irgc.org/risk-governance/irgc-
Fearnley C, Bird D, Haynes K, Jolly G, McGuire B risk-governance-framework/. Last accessed 05/02/2017.
(eds) (2017) Observing the volcano world: volcano Johnson BB (1987) Accounting for the social context of
crisis communication. Advances in volcanology, risk communication. Sci Technol Stud, 103–111
IAVCEI. Springer, Berlin, 350 p Jolly AD, Jousset P, Lyons JJ, Carniel R, Fournier N,
Feuillard M, Allegre CJ, Brandeis G, Gaulon R, Le Fry B, Miller C (2014) Seismo-acoustic evidence for
Mouel JL, Mercier JC, Pozzi JP, Semet MP (1983) an avalanche driven phreatic eruption through a
The 1975–1977 crisis of La Soufrière de Guadeloupe beheaded hydrothermal system: an example from the
(FWI): a still-born magmatic eruption. J Volcanol 2012 Tongariro eruption. J Volcanol Geotherm Res
Geotherm Res 16:317–334 286:331–347
20 J. Gottsmann et al.

Kokelaar BP (2002) Setting, chronology and conse- Phillipson G, Sobradelo R, Gottsmann J (2013) Global
quences of the eruption of Soufrière Hills Volcano, volcanic unrest in the 21st century: an analysis of the
Montserrat (1995–1999). In: Druitt TH, Kokelaar BP first decade. J Volcanol Geotherm Res 264:183–196
(eds) The eruption of Soufrière Hills Volcano, Potter S, Scott B, Jolly G, Neall V, Johnston D (2015)
Montserrat, from 1995 to 1999, vol 21. Geological Introducing the volcanic unrest index (VUI): a tool to
Society, London, Memoirs, pp 1–44 quantify and communicate the intensity of volcanic
Komorowski J-C, Boudon G, Semet M, Beauducel F, unrest. Bull Volcanol 77:1–15
Anténor-Habazac C, Bazin S, Hammouya G (2005) Reid ME (2004) Massive collapse of volcano edifices
Guadeloupe. In: Lindsay JM, Robertson REA, Shep- triggered by hydrothermal pressurization. Geology
herd JB, Ali S (eds) Volcanic Atlas of the Lesser 32:373–376
Antilles, Seismic Research Unit, The University of the Ricci T, Barberi F, Davis MS, Isaia R, Nave R (2013)
West Indies, Trinidad and Tobago, WI, 65-102 Volcanic risk perception in the Campi Flegrei area.
Komorowski J-C, Morin J, Jenkins S, Kelman I (2017) J Volcanol Geoth Res 254:118–130
Challenges of volcanic crises on small islands states. In: Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J,
Fearnley C, Bird D, Haynes K, Jolly G, McGuire B Tonini R, Papale P (2014) Recognizing and tracking
(eds) Observing the volcano world: volcano crisis volcanic hazards related to non-magmatic unrest: a
communication’, Advances in Volcanology, IAVCEI, review. J Appl Volcanol 3:17
Springer, Berlin, pp 1–18. doi:10.1007/11157_2015_15 Rouwet D, Constantinescu R, Sandri L (2017) Determin-
Lepointe E (1999) Le réveil du volcan de la Soufrière en istic versus probabilistic volcano monitoring: not “or”
1976: la population guadeloupéenne à l’épreuve du but “and”. Adv Volcanol. doi:10.1007/11157_2017_8
danger. In: Yacou A (ed) Les catastrophes naturelles Sandri L, Tonini R, Rouwet D, Constantinescu R,
aux Antilles – D’une Soufrière à l’autre. CERC Mendoza-Rosas AT, Andrade D, Bernard B (2017)
Université Antilles et de la Guyane, Editions Karthala, The need to quantify hazard related to non-magmatic
Paris, pp 15–71 unrest: from BET_EF to BET_UNREST. Adv Vol-
Maeno F, Nakada S, Oikawa T, Yoshimoto M, Komori J, canol. doi:10.1007/11157_12017_11159
Ishizuka Y (2016) Reconstruction of a phreatic Salvage RO, Karl S, Neuberg J (2017) Volcano seismology:
eruption on 27 September 2014 at Ontake volcano, detecting unrest in wiggly lines. Adv Volcanol 1–17
central Japan, based on proximal pyroclastic density Sekiya S, Kikuchi Y (1890) The eruption of Bandai-san.
current and fallout deposits. Earth, Planets and Space Tokyo Imp Univ Coll Sci J 3:91–172
68:82 Sheldrake TE, Sparks RSJ, Cashman KV, Wadge G,
Manville V, Major JJ, Fagents SA (2013) Modeling lahar Aspinall WP (2016) Similarities and differences in the
behavior and hazards. In: Fagents S, Gregg T, Lopes R historical records of lava dome-building volcanoes:
(eds) Modeling volcanic processes: the physics and Implications for understanding magmatic processes
mathematics of volcanism. Cambridge University and eruption forecasting. Earth Sci Rev 160:240–263
Press, pp 300–330 Siebert L, Cottrell E, Venzke E, Andrews B (2015)
Marzocchi W, Woo G (2007) Probabilistic eruption Earth’s volcanoes and their eruptions: an overview. In:
forecasting and the call for an evacuation. Geophys Sigurdsson H et al (eds) The encyclopedia of volca-
Res. Lett 34:L22310. doi:10.21029/22007GL031922 noes, 2nd edn. Elsevier, Amsterdam. http://dx.doi.org/
Marzocchi W, Sandri L, Selva J (2008) BET_EF: a 10.1016/B978-0-12-385938-9.00012-2
probabilistic tool for long- and short-term eruption Stone J, Barclay J, Simmons P, Cole PD, Loughlin SC,
forecasting. Bull Volc 70:623–632 Ramon P, Mothes P (2014) Risk reduction through
Mothes P, Yepes HA, Hall ML, Ramon PA, Steele AL, community-based monitoring: the vigías of Tungu-
Ruiz MC (2015) The scientific-community interface rahua. Ecuador. J. Appl. Volcanol. 3:11
over the 15 year eruptive episode of the Tungurahua Stirling A (2010) Keep it complex. Nature 468:1029–1031
Volcano, Ecuador. J Appl Volcanol 4:9 Tazieff H (1980) About the Soufrière of Guadeloupe.
Mothes P, Ruiz MC, Viracucha EG, Ramón PA, J Volcanol Geotherm Res 8:3–6
Hernández S, Hidalgo S, Bernard B, Gaunt EH, Jarrín Venezky D, Newhall C (2007) “WOVOdat Design
P, Yépez MA, Espín PA (2017) Geophysical foot- Document; The Schema, Table Descriptions, and
prints of Cotopaxi´s unrest and minor eruptions in Create Table Statements for the Database of World-
2015: an opportunity to test scientific and community wide Volcanic Unrest (WOVOdat Version 1.0).” U.S.
preparedness. Adv Volcanol 1–30 Geological Survey Open-File Report, 184.
OVSG-IPGP (1999–2013) Bulletin mensuel de l’activité Woo G (2008) Probabilistic criteria for volcano evacua-
volcanique et sismique de Guadeloupe. Monthly tion decisions. Nat Hazards 45(1):87–97
public report, Observatoire Volcanologique et Sis- Woo G (2015) Cost-Benefit analysis is volcanic risk. In:
mologique de Guadeloupe - Institut de Physique du Papale P (ed) Volcanic Hazards, Risks and Disasters.
Globe de Paris, Gourbeyre (ISSN 1622-4523). http:// Springer, pp 289–300. https://www.elsevier.com/
www.ipgp.fr/fr/ovsg/bulletins-mensuels-de-lovsg. books/volcanic-hazards-risks-and-disasters/papale/
Last accessed 11-10-2016 978-0-12-396453-3
Volcanic Unrest and Pre-eruptive Processes … 21

Wadge G, Isaacs MC (1988) Mapping the volcanic hazards Wilkinson E, Lovell E, Carby B, Barclay J, Robert-
from Soufriere Hills Volcano, Montserrat, West Indies son REA (2016) The dilemmas of risk sensitive
using an image processor. J Geol Soc 145(4):541–551 development on a small volcanic island. Resources 5
Wadsworth FB, Vasseur J, Scheu B, Kendrick JE, (2):21. doi:10.3390/resources5020021
Lavallée Y, Dingwell DB (2016) Universal scal- Wright TL, Pierson T (1992) Living with volcanoes.
ing of fluid permeability during volcanic welding and No. 1073. US Geologic Survey
sediment diagenesis. Geology 44(3):219–222 Wynne B (1992) Uncertainty and environmental learning:
Wilkinson E (2015) Beyond the volcanic crisis: reconceiving science and policy in the preventive
co-governance of risk in Montserrat. J Appl Volcanol 4:3 paradigm. Glob Environ Change 2:111–127

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 The Role of Laws Within
the Governance of Volcanic Risks

R. J. Bretton, J. Gottsmann and R. Christie

Abstract
The governance of volcanic risks does not take place in a vacuum. In
many cultures, volcanic risks are perceived to be susceptible to
governance with the objective of achieving their effective mitigation,
and have become the responsibility of the institutions and stakeholders of
relevant social communities. An array of international, national and local
laws dictate governance infrastructures, the roles of duty holders and
beneficiaries and the relationships between them (the stakeholders), duties
and rights (the stakes) and acceptable standards of safety and wellbeing
(the ultimate rewards). Many regional, national and local stakeholders
(individuals and entities) have a range of different, yet complementary,
roles, duties, rights and powers. Much of this chapter, which has two main
sections, represents a summary of a longer paper (Bretton et al. 2015) that
addresses legal aspects of the future governance of volcanic risks. After a
general introduction to relevant terminology in the first section, the second
section describes the significant threat posed by periods of volcanic unrest.

R. J. Bretton (&)
J. Gottsmann

School of Earth Sciences, University of Bristol,
Wills Memorial Building, Queens Road, BS8 1RJ
Bristol, UK
R. J. Bretton
J. Gottsmann
R. Christie
The Cabot Institute, University of Bristol, Wills
Memorial Building, Queens Road, BS8 1RJ Bristol,
UK
R. Christie
School of Sociology, Politics and International
Studies, University of Bristol, 4 Priory Road, BS8
1TU Bristol, UK

Advs in Volcanology (2019) 23–34

DOI 10.1007/11157_2017_29
© The Author(s) 2017
Published Online: 28 November 2017
24 R. J. Bretton et al.

The third section contains a general introduction to the critical concept of

risk which lies at the heart of governance and provides a more detailed
description of the many roles that national laws play. Reference is also
made to international law which has an increasingly important role in the
absence of relevant national laws, or when national laws are inadequate,
ineffective or unenforced.

Keywords
Hazard
Risk
Risk governance
Legal duties

example, non-resident workers may be less

1 Introduction exposed than full-time residents and ‘vulnera-
bility’ may be related to length of exposure).1
This chapter describes the ways in which laws
‘Governance’ is a complex concept attracting
create the administrative and functional infras-
a multitude of definitions (Walker et al. 2010).
tructures that facilitate the effective mitigation of
Although it “encompasses an array of organisa-
volcanic risks.
tions, practices and ideas” (Rothstein et al. 2012)
For the sake of brevity and clarity, we draw
that change over time, for sake of clarity, we
upon the existing rich discourse on relevant ter-
embrace following definition.
minology (e.g. Fournier d’Albe 1979; Luhmann
1998, 1992; Power 2007, 2009; UN/ISDR 2009; Governance is the sum of the many ways indi-
viduals and institutions, public and private, man-
Smith and Petley 2009; MIAVITA 2012) and age their common affairs. It is a continuing process
adopt a number of brief working definitions. through which conflicting or diverse interests may
A ‘hazard’ is an event (defined by risk-related be accommodated and co-operative action may be
temporal, spatial and other parameters) that may taken. It includes formal institutions and regimes
empowered to enforce compliance, as well as
cause adverse effects. It is a complex function informal arrangements that people and institutions
being the “probability of any particular area either have agreed to or perceive to be in their
being affected by a destructive volcanic mani- interest (Commission on Global Governance 1995,
festation within a given period of time” (Fournier 4).
d’Albe 1979, 321). A ‘volcanic hazard’ is a ‘Risk governance’ includes all attempts to man-
volcanic scenario (defined by risk-related tem- age the three constituent variables of risk
poral, spatial and physical parameters) that may including steps to mitigate volcanic hazards
cause adverse consequences to people and/or (there are very few successful examples of this),
valued assets. reduce the exposure of people, assets etc. and
In the early 1970s, definitions of risk identi- reduce their vulnerability when exposed. This
fied the product of three separate and distinct expression is adopted not only as an analytic
elements—‘vulnerability’ on ‘exposure’ to a term to describe stakeholders undertaking miti-
defined ‘hazard’ (UNESCO 1972). Bankoff et al. gation activities but also for ‘normative’ (i.e.
(2004) refers to volcanic hazards being one of evaluative standard) purposes. Risk governance
three variables (hazard, exposure and vulnera-
bility) that are convolved to produce volcanic 1
The 18th report of the Scientific Advisory Committee on
risks. For the purposes of governance, it may be Montserrat Volcanic Activity contains a good example of
helpful to quantify each identified exposure in a risk assessment which adopts this approach. It differ-
units of both number (i.e. the number of people entiates between the risks faced by residents in Zone A
and those of workers involved in the shipment of sand
exposed) and time (given that, by way of from Plymouth Jetty.
The Role of Laws Within the Governance of Volcanic Risks 25

has a set of definable ‘good’ qualities that pro- response may take the form of the release of
vide for the effective integration of the key inappropriate advice, media speculation, unwar-
components of how risks are handled by risk ranted emergency declarations and premature
stakeholders (Walker et al. 2010; IRGC 2009). cessation of economic activity and community
services” (Johnston et al. 2002, 228).

2 Geological Background
3 Risk Governance and Roles
Active volcanism can involve complex of Law
multi-hazard phenomena. Precursory unrest pro-
vides, by means of its signals, the monitoring The concept of risk as something that can be
data upon which evidence-based short-term managed through human intervention is a rela-
hazard analysis is grounded. However, periods tively new one and important because it has
of mild unrest, even if they may not lead to an become an increasingly pervasive concept in
eruption, can themselves present a range of many societies. Risk is also associated with
hazards including earthquakes, ground deforma- notions of choice, responsibility and blame
tion, hydrothermal changes/eruptions and gas/ (OECD 2015).
water chemistry changes. These precursory haz- Risk evolved from its modest origins in the
ards can create societal risks that can escalate seventeenth century and became in the nine-
unnecessarily and therefore require very careful teenth century a principle for the objectification
management. Unrest periods create, not only of possible experience—not only of the hazards
uncertainty about what is happening and of personal life and private venture, but also of
resulting public alarm, anxiety and speculation, the common venture of society (Gordon 1991).
but also demands for information and advice By the late nineteenth century, risk had “be-
(Johnston et al. 2002). come central to the rhetoric of regulation”. State
The evolution of an unrest period will depend regulation of risk emerged as the means by which
upon its underlying causative processes, which the state controlled economic activities in Wes-
can lead to different outcomes in different loca- tern societies. The traditional objects of state
tions and with different spatial and physical regulation were manufactured risks, most par-
properties (Rouwet et al. 2014; Sobradelo and ticularly those resulting from scientific and
Marti 2015). technological innovation within manufacturing
Volcanic hazard communications and risk processes. The usual style of state regulation was
mitigation decisions rely upon the suitable and “command and control” by imposing formal,
sufficient collection, and the correct analysis and structured and active risk management duties.
interpretation of monitoring data, and the geo- The state exercised control through the promul-
logical record (Newhall and Hoblitt 2002; Sparks gation of primary (i.e. enabling) and secondary
et al. 2012; Rouwet et al. 2014). The analysis of (i.e. detailed implementing) laws and policing
monitoring data, which will often be limited in through specialist inspectorates.
both quantity and quality, is challenging and In the twenty first century, regulation is no
there are many uncertainties in identifying longer confined to non-natural, human-made
causes and thereafter anticipating the evolution risks. Many risks are, in whole or in part,
of unrest and imminent eruption (Sparks et al. recurring social manifestations (i.e. human-made
2012; Phillipson et al. 2013; Sobradelo and phenomena) with negative consequences (Lauta
Marti 2015). 2014). In many cultures, particularly western
Hazard analysis is difficult and the risk gov- cultures, they are no longer perceived as the
ernance stakes are high. Poorly handled unrest consequences of external forces occurring inde-
periods cause social, economic and political pendently of society and insusceptible to miti-
problems, even without an eruption. “Adverse gation by society. Accordingly, they are now
26 R. J. Bretton et al.

positioned within, and have become the respon- Notwithstanding these challenges, many
sibility of, the institutions and stakeholders of jurisdictions have national laws that attempt to
relevant social communities (Lauta 2014). These regulate the management of risks arising from
human-made risks are perceived to be susceptible natural hazards. Many reflect the shift in para-
to regulation with the objective of achieving their digm, at both international and national levels,
effective mitigation. By way of illustration, the from focussing on ex-post, reactive response (the
population of Naples has greatly increased since phases of emergency response and post-disaster
1944 and many would argue that the resulting longer term recovery) to ex-ante, pro-active risk
increase in volcanic risk exposure is human- management and mitigation (the phase of plan-
made and capable of regulation. ning and preparedness) (UN SC-DRR 2009).
Low probability-high impact risks pose a As illustrated in Fig. 1, national laws create
particular challenge for legislators. In fact there governance infrastructures, duties of care and
are three related challenges, namely scientific duty holders, rights and rights holders, enabling
uncertainty, a low likelihood of occurrence, and powers, regulators, enforcement powers, and
significant societal consequences. Whilst the lastly scrutiny venues. Each will now to consid-
elevated consequences of these risks call for ered in turn.
some level of regulation, the intrinsic uncertainty
and low probability of their occurrence make it
difficult to review the evidentiary scientific jus- 3.1 The Creation of National Risk
tification, to assess costs and benefits, and to Governance Infrastructures
identify means by which chosen regulatory goals
can be pursued (Simoncini 2013). National laws tend to identify, authorise and fund
In the absence of a tragedy, it is difficult to risk governance bodies (e.g. government
measure the performance of law-backed societal departments and agencies, and public corpora-
risk governance by the usual measures of: tions) and public officials (e.g. individuals such
(1) economy (e.g. value for money) for input and governors, mayors, prefects and village heads)
process; (2) efficiency (e.g. quality delivered on within a coherent legal and administrative
time) for process and output; and (3) effective- framework, in other words, a risk governance
ness for output and outcome. The indicators of infrastructure. These laws often use and build
outcome (the intended and unintended results) of upon existing entities within existing adminis-
the integrated governance system will be related trative frameworks that have multi-level national,
to the impacts on, and the consequences for, regional, district, municipal etc. political divi-
public good, safety, security, health and welfare sions and subdivisions.
but it will be a challenge for any related targets In some jurisdictions, formal legal infrastruc-
(e.g. benchmarks and performance standards) to tures anticipate and rely upon less formal struc-
be SMART—Specific, Measurable, Achievable, tures and relationships at local levels nearer
Relevant and Timed (OECD 2002). at-risk communities. For example, in Ecuador,
By contrast, in a fact-finding process of the risk governance infrastructure relies upon the
scrutiny after a tragedy, the use of SMART tar- engagement and commitment of local represen-
gets may become more practicable. It may be tatives (e.g. chiefs and elders) and volunteers,
possible to measure hazard characterisation out- such as hazard wardens/monitors, for both hazard
puts against planned targets for timely delivery, data gathering and risk mitigation.
user-friendliness, outcome-focussed, and In some jurisdictions, such as Italy, the laws
temporal/spatial/intensity forecast accuracy. favour the imposition of duties upon individuals,
Based upon findings of fact, it may be feasible to rather than impersonal legal entities such as
quantify the resulting risk-mitigation impact government departments/agencies and public
measured in lives and assets saved. companies. Legal duties may be founded upon an
The Role of Laws Within the Governance of Volcanic Risks 27

Duty Holders Managerial risks Scrutiny venues/procedures

Criminal
Civil claims Fatal accident
prosecutions
& enquiries &
&
risk of compensation risk of factual
risk of penal
payments findings
Duties of care owed in respect of societal risks sanctions

Actual practice to fulfil duties of care

as influenced by Regulators
acceptable current practice but
subject to ‘standard equivocality’
Note
The 'constants' of the generalised legal
infrastructure are shown in black-lined
square boxes whereas the only 'variable'
Rights Holders is in a green-lined box with rounded
corners.

Fig. 1 The many roles of law in the governance of natural hazards set out in a generalised legal framework

individual having effective decision-making 3.2 The Creation of Duty and Rights
powers and control over financial resources
rather than upon an individual holding a partic- National laws allocate to bodies and individuals
ular title or occupying a particular post (Bergman (duty holders such as volcano observatories and
et al. 2007). civil protection authorities) high level manage-
These infrastructures can be complex, con- ment functions with responsibilities (duties of
fusing, fragmented and multi-level. They are care), which are owed to the particular classes of
often the creations of multiple sets of national people for whose benefit the duties were created
primary (enabling) and secondary (detailed (rights holders).
implementing) legislation supplemented as nec- Since modest beginnings in the 1840’s near
essary by further provisions at ministerial, Vesuvius Italy, the role of over 100 volcano
inter-ministerial, regional, provincial and local observatories around the world has evolved.
levels of government. Observatories have at least two overlapping roles
Occasionally additional specialised bodies are which involve a synergy of observation and
established (e.g. emergency management agen- theory. They have been described as ‘critical in
cies, research/monitoring institutes and volcano the volcanic risk reduction cycle’2 (Jolly 2015,
observatories) with the creation of statutory roles 302), and employ and/or engage scientists who
to be filled by appointed individuals. practice at the hazard-risk interface. The World
In a few known jurisdictions (e.g. USA, Organisation of Volcano Observatories
Canada and the Philippines) laws also regulate to (WOVO), a Commission of the International
varying degrees the qualification, licensing and
2
registration of geologists and the practice of This cycle includes periods before, during and after
periods of volcanic unrest that may or may not lead to an
geology per se.
eruption (Jolly 2015, 302).
28 R. J. Bretton et al.

Table 1.1 Volcano observatories—three contrasting regimes

Montserrat—One individual with no powers of delegation
Section 8 of the Montserrat Volcano Observatory Act 2002 states that the Director of the Observatory shall be
responsible for “reporting on the status of the volcanic activity in a regular and timely manner to the appropriate
authorities” and “assisting in the dissemination to the public of information concerning the status of volcanic activity”
Alaska, USA—A group of institutions
“The Alaska Volcano observatory is a joint programme of the United States Geological Survey (USGS), the
Geophysical Institute of the University of Alaska Fairbanks (UAFGI) and the State of Alaska Division of Geological
and Geophysical Surveys (ADGGS)” (Jolly 2015, 299)
New Zealand—One institution under powers of delegation given to one individual
The Institute of Geological and Nuclear Sciences Limited (or GNS Science) is a Crown research institute created in
1992. The GNS has “sole responsibility for providing volcanic activity warnings and hence provides the function of a
volcano observatory” (Jolly 2015, 299) under wide powers of delegation given to the Director of Civil Defence
Emergency Management in the Civil Defence Emergency Management Act 2002

Association of Volcanology and Chemistry of safety, not specifying any risk creator (a partic-
the Earth’s Interior (IAVCEI), is an organization ular hazard, natural or otherwise) or they may be
of and for volcano observatories of the world. more specific, identifying a particular hazard
WOVO’s website notes that its “members are (ground uplift, earthquakes etc.).
institutions that are engaged in volcano surveil- Rights holders may be given the right to a safe
lance and, in most cases, are responsible for and healthy environment, and to be represented,
warning authorities and the public about haz- consulted or engaged in risk decisions and/or
ardous volcanic unrest”. given information. Additional rights may be
Examples of three contrasting regimes are given to certain categories of persons due to
given in Table 1.1. special vulnerabilities and/or the influence of
In some jurisdictions, general disaster man- social structures and practices. These categories
agement obligations are also imposed on ‘the may include women, the very poor, older per-
community’ (i.e. members of the public) and sons, children and people with disabilities (IFRC
non-government business entities. At-risk indi- 2015).
viduals and communities, businesses (such as There are two main types of duty of care,
aeroplane makers and operators within the avia- which are called here, respectively, ‘functional’
tion industry) and insurers have active and critical and ‘goal-setting’. Functional duties dictate the
roles to play in the governance of volcanic risks, fulfilment of a particular role (e.g. a duty to
however, their roles are not the principal focus of undertake monitoring, to prepare plans and pro-
either this chapter or Bretton et al. (2015). The grammes for emergency preparedness or to pro-
needs and sentiments of all duty and rights vide emergency preparedness communications
holders, who depend upon and use geoscientific and warnings). Goal-setting duties require the
knowledge of volcanic hazards, must be identi- achievement of an outcome (e.g. a duty to ensure
fied and reflected carefully and clearly in the roles the safety and wellbeing of identified rights
and interface practices of volcanologists. holders). Not even within the highly regulated
During an emerging period of volcanic unrest, field of occupational health are these safety goals
the relevant duty holders may change as the absolute (i.e. unqualified). The imposition of an
defined duties are transferred from one duty unrealistic absolute duty would give rights
holder to another—sometimes as a result of holders a theoretical guarantee of health and
changing hazard or risk characterisations. These safety within a risk-free environment.
duties of care can be framed in a wide variety of As a general rule, ‘qualified’ duties of care are
ways. They may relate to general health and therefore laid down. These duties represent
The Role of Laws Within the Governance of Volcanic Risks 29

democratic statements or mandates of ‘accept- (1) law-based performance standards offering

able’ or ‘optimal’ risk after mitigation, and guidance to societal risk managers; or (2) law
express a rational trade-off between safety and endorsed self-regulatory regimes such as those
risk (Hood and Rothstein 2001; Rothstein 2014). that frequented the early stages of food
Rothstein (2014) notes “After all, what is an regulation.
acceptable risk other than a euphemistic bound- Within the ‘goal-setting’ legislative approach,
ary between an acceptable adverse outcome and referred to above, it is implicit that there is an
an unacceptable failure”. obligation on duties holders to establish the
Compliance with qualified duties inevitably nature and suitability (i.e. the legal adequacy) of
requires duties holders to perform a risk-focussed their societal risk management arrangements.
cost/benefit analysis and a test of proportionality. This difficult justification will usually be done
A duty holder wishing to establish that societal post-facto, in other words, only after the risk
risks have been reduced ‘as low as reasonably outcome (perhaps a disaster properly so-called) is
practicable’ has to show that the costs (the sac- known and legal consequences are already being
rifice) of further feasible safety measures would considered (Simoncini 2011). The justification
be grossly disproportionate to the additional will cover, but not be limited to, the arrange-
safety benefits of those measures (based upon ments that were necessary for the planning,
UK Office for Nuclear Regulation 2013). organisation, control, monitoring and review of
Laws rarely, if ever, attempt to dictate, in societal risk mitigation measures. To complete
either general or more detailed terms, the societal the authors’ footballing analogy, post-facto legal
risk management arrangements that will be processes are analogous to slow motion TV
required to either fulfil a functional role or replays, in full view of partisan onlookers and
achieve a stated safety outcome. In practice, an experts with hindsight. They determine what has
assessment of societal vulnerability has two main happened, whether or not a goal has been scored
stages. Firstly, the nature and scope of duties of and, if not, why not and what effect any missed
care must be identified and delineated. This is goal had on the final score (i.e. whether legal
essentially a matter of law and involves the legal compliance has actually been achieved and, if
interpretation of primary and secondary legisla- not, why not and what the consequences should
tion and, if relevant, case law. Secondly, it is have been if compliance had been achieved).
necessary to identify the actions that the duty ‘Standard equivocality’, which is the absence
holder should take to fulfil those duties. This is of commonly recognised standards (norms), is
far more difficult. Competent lawyers can likely to exist in the absence of clear ‘legal’
describe the safety function or outcome required requirements, approved codes of practice or
in law—in football terms the dimensions and guidance. The resulting challenges faced by duty
position of the goal. However, they can offer holders are: (1) to find or design authoritative
very limited guidance regarding the practical standards or benchmarks to steer their societal
measures that the competent societal risk man- risk management arrangements; and thereby
ager will need to take to achieve legal compli- (2) to increase their chances of fulfilling their
ance (i.e. how to actually get the ball over the societal risk duties of care and achieving legal
goal line). compliance; and thereby (3) to minimise their
In the case of food standards and occupational vulnerability to managerial risks.
health and safety standards, it is common for Rothstein (2002) and Hood (1986) have noted
national laws to set up government agencies to that, in the absence of commonly agreed and
carry out research, to set performance standards practical principles or methodologies by which
and offer approved codes of practice or authori- compliance can be measured (‘standard-
tative guidance. By contrast, in respect of vol- unequivocality’), process compliance is difficult
canic hazards, at both the international and to monitor and enforce. Other obstacles to moni-
national levels, there appear to be neither: toring, surveillance and enforcement include
30 R. J. Bretton et al.

inherent scientific uncertainty, a dynamic state of of care by regulators, generally health and safety
scientific knowledge, a lack of expertise within agencies; and (2) the ex-post facto reactive
regulatory agencies, and often complex and frag- scrutiny of events, the identification of duty
mented multi-level infrastructures. Donovan and holders, the assessment of what happened and
Oppenheimer (2014) note complexities in gov- what should have happened and, if appropriate,
ernmental structures presented major challenges the imposition of sanctions and/or the granting of
to managing volcanic eruptions in Montserrat. remedies. The latter procedures are required at a
Recent crises including the 2010 Icelandic national level to comply with the international
Eyjafjallajökull eruption have highlighted the law ex-post facto obligations which are now
difficulty of co-ordinating and synthesising sci- considered.
entific input from many different disciplines and
institutions and translating these into useful policy
advice at very short notice (OECD 2015). 3.5 The Role of International Law

In the absence of relevant national laws, or when

3.3 The Creation of Powers national laws are inadequate, ineffective or
unenforced, there is room for the intervention of
National laws have traditionally granted defined international law. The European Court of Human
‘authorities’ to government duty holders backed Rights (ECHR) has taken the lead and it is sug-
by administration, protection and intervention gested here that in time the Inter-American Court
powers (ordinary, extraordinary and emergency). of Human Rights will follow. The European
Governance, with an emphasis more on control Convention of Human Rights (EConHR) lays
than protection, has often been achieved by the down a positive obligation on States to take
exercise of authority using linear “coercion and appropriate steps to safeguard the lives of citi-
enforcement” (Walker et al. 2010). zens within their jurisdiction. Article 2(1)
EConHR provides that “Everyone’s right to life
shall be protected by law. No one shall be
3.4 The Creation of Regulators, deprived of his life intentionally save in the
Enforcement Powers execution of a sentence of a court following his
and Scrutiny Venues conviction of a crime for which this penalty is
provided by law.”
Laws often establish, resource and empower In the context of the management of natural
regulators to monitor the performance of duty hazards, the most important case involving
holders and to take enforcement actions, includ- Article 2 arose in 2008. Budayeva and others v
ing prosecutions, against them if necessary. Russia (2008) ECHR 15339/02 concerned a
Examples of regulators include the Labour mudslide. In this case, the EHCR considered
Standards Agency in Japan, the Department of principles that had been applied in Oneryildiz v
Labour Health and Safety Service in New Zeal- Turkey (2004) to a human-made hazard—an
and and the Occupational Safety and Health industrial risk or dangerous activity such as the
Agency in the United States of America. These operation of a waste site. They were subse-
regulators often have very wide powers similar quently adopted in Kolyadenko and others v
to, and sometimes exceeding, those of police Russia (2012) in respect of natural flash floods.
forces. They include the power to enter premises, Budayeva and others v Russia
to investigate and inspect, to acquire and pre- The town of T was situated in a mountain district.
serve evidence and to serve notices. Two tributaries passed through it and were known
Laws provide the formal scrutiny venues to have associated mudslides. A mud collector and
a dam were constructed in order to protect T. The
(courts, tribunals etc.) and related procedures for: dam was seriously damaged by a mud and debris
(1) the ex-ante pro-active enforcement of duties
The Role of Laws Within the Governance of Volcanic Risks 31

flow in August 1999, so funds were requested to investigation. In relation to the latter, procedures
construct observation posts to warn of mudslides must exist for identifying not only shortcomings
until it could be repaired, and to carry out certain
emergency works to the dam. in the ex-ante regulatory measures but also any
Those measures were never implemented. A num- errors committed by those responsible (i.e. duty
ber of mudslides occurred in July 2000, killing holders). If there are any shortcomings and the
eight residents, including the first applicant’s infringement of the right to life was not inten-
husband, and destroying the applicants’ homes.
It was decided to dispense with a criminal inves- tional, it is not necessary for criminal proceed-
tigation into the circumstances of the death of the ings to be brought in every case. It may
first applicant’s husband, and claims of compen- satisfactory to make available to the victims civil
sation by the first applicant and others were law remedies (either alone or in conjunction with
refused on the basis that a mudslide of such
exceptional force could neither have been pre- a criminal law remedy), enabling any responsi-
dicted nor stopped. However, the applicants were bility of the parties concerned to be established
granted substitute housing and a lump-sum emer- and any appropriate civil redress, such as an
gency allowance. order for the payment of damages, to be
The applicants complained to the ECHR, inter alia,
that the authorities had violated the substantive obtained.
limb of Article 2 of the EConHR. The first appli- The positive obligations of EConHR State
cant asserted that the authorities were responsible duty holders under the ECHR are summarised in
for the death of her husband and she and the other Fig. 2.
applicants asserted that the authorities had failed to
take appropriate measures to mitigate the risks to
their lives posed by natural hazards.
The Court concluded that the relevant authorities 3.6 The Role of International
were aware of the mudslides (the hazards) and Institutions and Agencies
their capacity to cause devastating consequences
(the risks). There was no ambiguity about the
scope and timing of the work that needed to be In March 2015, the International Federation of
performed (the risk mitigation actions). After Red Cross and Red Crescent Societies (IFRC)
1999, risk mitigation was not given proper con- and the United Nations (UN) Development Pro-
sideration by the decision makers and budgetary
bodies (the duty holders) and there was no func-
gramme issued the pilot version of “The check-
tioning early warning system. State responsibility list on law and disaster risk reduction”. It
for the deaths had never been investigated. Each encourages accountability mechanisms within
applicant was awarded compensation. legislation to address failures to fulfil risk gov-
The EHCR determined that the obligation in ernance responsibilities. In particular it advocates
Article 2 entails, above all, a primary duty on the laws: (1) to establish public reporting or parlia-
State to put in place a clear legislative and mentary oversight mechanisms and transparency
administrative framework designed to provide requirements for government entities tasked with
effective deterrence against threats to the right to risk governance responsibilities; (2) to give a
life. This applies in the context of any activity, mandated role to the judiciary in enhancing
whether public or not, in which the right to life accountability; (3) to provide enforceable incen-
may be at stake and extends not only to industrial tives for compliance and disincentives for
risks and dangerous activities but also to actions non-compliance; and (4) to establish legal and/or
and omissions to control natural hazards. administrative sanctions (as appropriate) for
In the cases of Oneryildiz v Turkey (2004), public officials individuals and businesses for a
Budayeva v Russia (2008), and Kolyadenko v gross (“particularly egregious”) failure to fulfil
Russia (2012), the EHCR determined that there is their duties (IFRC 2015, 16).
a positive obligation: (1) ex-ante to take sub- The prioritisation of mitigation before
stantive regulatory measures to manage risks; response and recovery was recognised within the
and (2) ex-post facto to ensure that any risk Sendai Framework for Disaster Risk Reduction
eventuated fatalities are followed by a public 2015–2020 (the Sendai Framework) which
32 R. J. Bretton et al.

EConRH States as Duty Holders

Duty of care to fulfil obligation under EConHR Article 2 “Right to Life”

Legislative & administrative framework designed to provide effective deterrence against threats to the right to life

Ex-ante: Ex-Post facto:

Regulatory system to protect lives Procedures for identifying shortcomings

Substantive regulatory system with measures to: A prompt, diligent, effective, impartial, independent
(1) assess the risks inherent in the natural hazard; and public investigation (judicial or otherwise):
(1) to audit the ex-ante regulatory system; and
(2) implement, with all possible diligence, essential practical
control measures needed to ensure effective protection of (2) to ascertain;
at-risk citizens; - what happened;
(3) put in place a coherent supervisory system to encourage - the causes of the deaths;
those responsible to take steps to ensure adequate
- any shortcomings in the operation of the
protection of the population living in the area;
regulatory system; and
(4) set in place an emergency warning system; - the State authorities and officials (at all levels)
involved and their shortcomings.
(5) inform the local population of the potential uncontrolled
risks linked to the hazard and other essential information;
and

(6) establish sufficient coordination and cooperation

between the various administrative authorities to ensure that
the risks brought to their attention did not become so
serious as to endanger human lives

Fig. 2 The obligations of EConHR ‘States’ to manage natural hazards

emerged from the United Nations 3rd World national laws, supplemented by international
Conference on Disaster Risk Reduction laws and initiatives, few countries would have
(UN/ISDR 2015). It is suggested here that one the complex administrative infrastructures nec-
inevitable effect of the Sendai Framework will be essary for the mitigation of volcanic risks.
to enhance the importance of not only the col- Although emergency response may still dom-
lection and interpretation of monitoring data but inate thinking and funding in some jurisdictions,
also the better characterisation of unrest periods. national laws are unlikely to diminish in number
For the reasons stated in Sect. 2, periods of vol- and/or reach in the light of the emerging inter-
canic unrest, even if they do not lead to an erup- national law governance norms, the IFRC/UN
tion, present multiple hazards and risks which law checklist and the Sendai Framework.
require very careful assessment and mitigation.

References
4 Conclusions
Bankoff G, Frerks G, Hilhorst D (2004) Mapping
In many cultures, volcanic risks are perceived to vulnerability – disasters, development & people.
be susceptible to governance and have become Earthscan, London and New York
the responsibility of the institutions and stake- Bergman D, Davis C, Rigby B (2007) International
comparison of health and safety responsibilities of
holders of relevant social communities.
company directors, HSE research report RR 53, 207.
Laws create the stakeholders, the stakes and http://www.hse.gov.uk/leadership/international.pdf
the standards of risk governance. Without
The Role of Laws Within the Governance of Volcanic Risks 33

Bretton RJ, Gottsmann J, Aspinall WP, Christie R (2015) policy papers, no 21, OECD Publishing, Paris. http://
Implications of legal scrutiny processes (including the dx.doi.org/10.1787/5js33l1jcpwb-en
L’Aquila trial and other recent court cases) for future Phillipson G, Sobradelo R, Gottsmann J (2013) Global
volcanic risk governance. J Appl Volcanol 4:18. volcanic unrest in the 21st century: an analysis of the
https://doi.org/10.1186/s13617-015-0034-x first decade. J Volcanol Geoth Res 264(2013):183–
Commission on Global Governance (1995) Our global 196
neighbourhood. Oxford University Press, Oxford Power M (2007) Organised uncertainty: designing a
Donovan A, Oppenheimer C (2014) Science, policy and world of risk management. Oxford University Press,
place in volcanic disasters: insights from Montserrat. Oxford
Environ Sci Policy 38:150–161 (May) Power M (2009) The risk management of nothing. Acc,
Fournier d’lbe EM (1979) Objectives of volcanic mon- Organ Soc 34(2009):849–855
itoring and prediction. J Geol Soc136:321–326. doi: Rothstein H (2002) Neglected risk regulation: the insti-
https://doi.org/10:1144/gsjgs.136.3.0321 tutional attenuation problem, centre for analysis of risk
Gordon C (1991) Governmental rationality: an introduc- and regulation. London School of Economics and
tion. In Burchell G, Gordon C, Miller P (eds) The Political Science, London
foucault effect: studies in Governmentality. Harvester Rothstein H (2014) Exploring national cultures of risk
Wheatsheaf, Hemel Hempstead, pp 1–52 governance accessed via http://www.lse.ac.uk/
Hood C (1986) Administrative analysis: an introduction researchAndExpertise/units/CARR/publications/
to rules, enforcement and organisations. Wheatsheaf CARRmagR&R25-Rothstein.pdf
Books, Sussex Rothstein H, Borraz O, Huber M (2012) Risk and the
Hood C, Rothstein H (2001) Risk regulation under limits of governance: exploring varied patterns of
pressure: problem solving or blame shifting? Adm risk-based governance across. Europe Regul Gov 1–
Soc 33(1):21–53 21 http://doi.org/10.111/j.1748-5991.2012.01153.x
International Federation of Red Cross, Red Crescent Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J,
Societies (IFRC) and United Nations Development Tonini R, Papale P (2014) Recognizing and tracking
Programme (2015) The checklist on law and disaster volcanic hazards related to non-magmatic unrest: a
risk reduction, Pilot version, March 2015. IFRC, review. J Appl Volcanol 3(1):17. https://doi.org/10.
Geneva, Switzerland. http://www.ifrc.org/PageFiles/ 1186/s13617-014-0017-3
115542/The-checklist-on-law-and-drr.pdf Simoncini M (2011) Regulating catastrophic risks by
International Risk Governance Council (IRGC) (2009) standards. EJRR (1):37–50
What is risk governance? www.irgc.org Simoncini M (2013) Governing air traffic management in
Jolly G (2015) The role of volcano observatories in risk the single European sky: the search for possible
reduction. In: Loughlin SC, Sparks RSJ, Brown SK, solutions to safety issues. Eur Law Rev Issue 2
Jenkins SF, Vye-Brown C (eds) Global volcanic (2013):209–228
hazards and risk. Cambridge University Press, Smith K, Petley DN (2009) Environmental hazards:
Cambridge assessing risk and reducing disaster. Routledge, Lon-
Johnston D, Scott B, Houghton B, Paton D, Dowrick D, don and New York
Villamor P, Savage J (2002) Social and economic Sobradelo R, Martí J (2015) Short-term volcanic hazard
consequences of historic caldera unrest at the Taupo assessment through Bayesian inference: retrospective
volcano, New Zealand and the management of future application to the Pinatubo 1991 volcanic unrest.
episodes of unrest. Bull New Zealand Soc Earthquake J Volcanol Geoth Res 290(2015):1–11
Eng 35(4):215–229 Sparks RSJ, Aspinall WP, Auker M, Crosweller S,
Lauta KC (2014) Disaster law. Routledge, Oxford and Hincks T, Mahony S, Nadim F, Polley J, Syre E
New York (2012) Mapping and characterising volcanic risk,
Luhmann N (1998) Observations on modernity. Stanford magmatic rifting and active volcanism conference,
University Press, Stanford, CA Addis Ababa, 12 Jan 2012
Luhmann N (1992) Risk: a sociology theory. de Gruyter, United Kingdom Office for Nuclear Regulation (2013)
Berlin Guidance for the demonstration of ALARP (As Low As
MIAVITA (2012) Handbook for volcanic risk manage- Reasonably Practicable) NS-TAST-GD-005 Revision 6
ment: prevention, crisis management and resilience. United Nations Educational, Scientific and Cultural
MIAVITA team, Orleans Organization (UNESCO) (1972) Report of consulta-
Newhall CG, Hoblitt RP (2002) Constructing event trees tive meeting of experts on the statistical study of
for volcanic crises. Bull Volcanol 64:3–20 natural hazards and their consequences Document
OECD (Organisation for Economic Co-operation and SC/WS/500 11pp, Paris, France
Development) (2002) Outcome focussed management United Nations Development Programme Safer Commu-
in the United Kingdom by Ellis K, Mitchell S. OECD nities through Disaster Risk Reduction in Develop-
Publications, Paris, France ment Programme SC-DRR (2009) Lessons learned:
OECD (2015) “Scientific advice for policy making: the disaster management legal reform—The Indonesian
role and responsibility of expert bodies and individual experience UND-, Jakarta, Indonesia
scientists”, OECD science, technology and industry
34 R. J. Bretton et al.

UN/ISDR (2009) Terminology on disaster risk reduction. Walker G, Whittle R, Medd W, Watson N (2010) Risk
Switzerland, Geneva governance and natural hazards, Cap Haz-Net WP2
UN/ISDR (2015) Sendai framework for disaster risk Report, Lancaster Environment Centre, Lancaster
reduction 2015-30 University, Lancaster. http://caphaz-net.org/outcomess-
results
Kolyadenko, others v Russia (2012) Applications
17423/05, 20534/05, 20678/05, 23263/05, 24283/05,
35673/05, ECHR 17423/05
Oneryildiz v Turkey (2004) Application 48939/99 18
Legal Authorities and Case Law BHRC 145, 41 EHRR 20

Budayeva, others v Russia (2008) Applications 15339/02,

21166/02, 20058/02, 11673 & 15343/02, judgment of
20 March 2008, ECHR 15339/02

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Deterministic Versus Probabilistic
Volcano Monitoring: Not “or”
But “and”

D. Rouwet, R. Constantinescu and L. Sandri

Abstract
Volcanic eruption forecasting and hazard assessment are
multi-disciplinary processes with scientific and social implications. Our
limited knowledge and the randomness of the processes behind a volcanic
eruption yield the need to quantify uncertainties on volcano dynamics.
With deterministic and probabilistic methods for volcanic hazard assess-
ment not always being in agreement, we propose a combined approach
that bridges the two schools of thoughts in order to improve future
volcano monitoring. Expert elicitation has proven to be an effective way to
bind deterministic research within a probabilistic framework aiming to
reduce the uncertainties related to any hazard forecast; yet, numerous
exercises based on expert elicitation have revealed that the attempt to
reduce uncertainties led to the creation of new ones, often unquantifiable,
created by human nature and reasoning during stressful situations. Such
reasoning ignores the complexity of volcanic processes and the fact that
every scenario has a probability to occur. The recent probabilistic methods
and tools marry probabilistic and deterministic approaches and lead to
unprecedented models. Nevertheless, probabilistic hazard assessment is
often misunderstood as not all of the researchers involved have
backgrounds in such matters. A probabilistic method cannot stand-alone

D. Rouwet (&)
L. Sandri

Istituto Nazionale di Geofisica e Vulcanologia,
Sezioni di Bologna, Bologna, Italy
e-mail: [email protected]; dmitri.
[email protected]
R. Constantinescu
Seismic Research Centre, University of West Indies,
Saint Augustine, Trinidad and Tobago
Present Address:
R. Constantinescu
School of Geosciences, University of South Florida,
4202 E. Fowler Avenue, NES 107, Tampa, FL
33620-5550, USA

Advs in Volcanology (2019) 35–46

DOI 10.1007/11157_2017_8
© The Author(s) 2017
Published Online: 21 June 2017
36 D. Rouwet et al.

as it depends on data input obtained by deterministic approaches. We

propose that, given the symbiotic relationship between the two methods, a
probabilistic framework can play a role of moderator between various
deterministic disciplines, thus creating a coherent environment for
discussion and debate among seismologists, geodesists, geochemists.
This can be achieved by training all scientists involved in hazard
assessment, probability theory and data interpretation, while at least one
group member objectively uses the information provided to produce the
probabilities. Hence, numerical outcomes can be interpreted transparently
as they represent the quantification of experts’ knowledge and related
uncertainties. A probabilistic method that incorporates the joint opinions
of a group of multi-disciplinary researchers facilitates a more straightfor-
ward way of communicating scientific information to decision-makers.

Resumen extendid
La previsión de erupciones volcánicas y la evaluación del peligro son
procesos multidisciplinarios, con implicaciones tanto científicas como
sociales. Nuestro conocimiento limitado de los procesos detrás de una
erupción volcánica y su aleatoriedad genera la necesidad de cuantificar las
incertidumbres sobre las dinámicas del volcán y de mejorar la política de
la toma de decisiones durante una crisis volcánica. Sabiendo que existe un
desacuerdo sobre el uso de métodos determinísticos o probabilísticos
durante la evaluación de la peligrosidad volcánica, revisamos ambos
métodos y proponemos un enfoque que sirve como puente entre las dos
escuelas de pensamiento y que pueda mejorar las capacidades de
monitoreo volcánico en el futuro, hacia el reconocimiento en tiempo de
manifestaciones volcánicas y amenazas relacionadas. La elicitación de
expertos resulta ser una manera efectiva para relacionar la investigación
determinística con el marco probabilístico para poder reducir la
incertidumbre relacionada a cualquier intento de previsión de erupción;
sin embargo, numerosos ejercicios basados en elicitaciones de expertos
revelaron el hecho que este intento de reducir la incertidumbre creó nuevas
incertidumbres, a menudo imposible de cuantificar, siendo generada por la
naturaleza del pensamiento humano durante situaciones de estrés. El
proceso general es sujeto a la personalidad de un/a investigador/a o un
grupo de investigadores y sus ideas basadas en su experiencia. Esta
manera de pensar interfiere con la complejidad intrínsica de los procesos
volcánicos y con el hecho que cada escenario tiene una probabilidad de
ocurrencia. Los métodos e instrumentos probabilísticos recientes juntaron
los investigadores probabilísticos y determinísticos lo que resultó en
modelos e interpretaciones de información sobre volcanes sin precedentes.
Sin embargo, la novedad de la evaluación probabilística de peligrosidad
es, a menudo, incomprendida debido al hecho que no todos los
investigadores involucrados tienen una formación en estas materias
teóricas. El método probabilístico no puede existir autónomamente ya que
depende de datos de entrada obtenido a través de los estudios
determinísticos. Proponemos que, dada la relación simbiótica entre ambos
métodos, un marco probabilístico puede jugar el papel como moderador
Deterministic Versus Probabilistic Volcano Monitoring … 37

entre las varias disciplinas deterministas, creando un ambiente coherente

para discusiones y debates entre científicos (e.g., sismólogos, geodetas,
geoquímicos). Este se puede obtener por medio de entrenamiento de todos
los científicos involucrados en el monitoreo volcánico en la teoría
probabilística y la interpretación de datos, mientras que al menos un
miembro del grupo utiliza de manera objetiva la información disponible
para producir las probabilidades numéricas. Así, los resultados numéricos
pueden ser interpretados sin duda alguna, ya que representan la cuantifi-
cación del conocimiento de los expertos y las incertidumbres relacionadas.
Un método probabilístico que incorpora las opiniones conjuntas de un
grupo de investigadores multidisciplinarios facilitará una manera más
transparente de comunicación de la información científica hacia las
autoridades civiles, así mejorando (1) el proceso de toma de decisiones
durante una crisis y la mitigación a largo plazo, y (2) el estado de medidas
de preparación que incorpora los varios aspectos sociales. En el futuro, los
reportes de previsiones emitidos por los científicos deberían incluir los
resultados numéricos de los modelos probabilísticos; la arquitectura de
monitoreo se debería expander más allá del arreglo clásico “sismo-
deformación-gas” hacia un arreglo “sismo-deformación-gas-probabilidad”.
Este capítulo de opinión pretende proponer una ideología posible, con el
máximo respeto para el volcán, la sociedad, los científicos individuales o
los grupos de científicos, y para las autoridades que toman las decisiones,
con un objetivo común: mejorar la previsión de amenazas relacionadas a
los volcanes para proteger la sociedad.

Keywords



Volcano monitoring Probabilistic hazard assessment Deterministic



research Bridging and symbiosis Best practice scheme

Palabras clave
Monitoreo volcánico
Evaluación probabilística de amenaza
Investi-
gación determinística
Puente y simbiosis
Esquema de mejor práctica

translate these to decision makers and, occasion-

1 Introduction
ally, to the population. The combination of social
implications and forecasting future behaviour of
Volcanoes are intrinsically complex and unpre-
complex natural systems makes volcanology a
dictable systems manifesting non-linear behaviour
rather unique but “tricky business”. Besides the
in space and time, on the long- and short-term.
need to quantify uncertainties and better under-
Understanding how volcanoes evolve with time
stand our limited knowledge on volcanoes it is
through the various stages of activity—from qui-
necessary to legally protect volcanologists when
escence through unrest to eruption—is highly
exporting their knowledge outside their protected
challenging. Awareness of these facts is a basic
professional community (Bretton et al. 2015).
requisite when working with/on volcanoes. A ma-
During the past 40–50 years, volcano moni-
jor goal in volcanology is quantifying uncertainties
toring and eruption forecasting during volcanic
on volcano dynamics, and learning how to
38 D. Rouwet et al.

crises has been largely dominated by the deter- view, the probability of an eruption occurring, or
ministic approach (Sparks 2003, i.e. most likely of a given hazardous event hitting a target area, is
scenarios). Experts with different research back- described both by a best-evaluation value (for
grounds track changes in “their” parameters example the mean or the median of the proba-
related to volcanic activity, afterwards discussed bility distribution), and by a dispersion around
in “protected” round tables, generally behind such value (represented by standard deviations or
locked doors, to eventually come up with a single by a confidence interval). These two quantities
voice. This single voice does not and should not can be directly related to two different sources of
reflect possible internal conflicts or disagree- uncertainty: the aleatory one and the epistemic
ments behind the closed door. Such disagreement one. In this way, Bayesian approaches allows
is an often-unstated expression of the uncertainty quantifying also to what extent our probabilistic
due to our lack of knowledge on volcanic pro- assessments are constrained by data and knowl-
cesses, and the possible unavailability of data edge. In other words, now we know, as a group
(i.e. epistemic uncertainty) and due to the of volcano-experts, to which degree we can be
intrinsic randomness of the volcanic process wrong in our forecasts behind the closed doors.
studied (i.e. aleatory uncertainty). The power of This black-on-white awareness brought to light
the single voice from the group of experts often by probability density functions has led to some
misleads the receivers of the message (decision key questions, from the in- and outside worlds:
makers, authorities or lay public), believing the (1) are we, as volcano-experts, replaceable by a
experts are “sure” on the evolution of volcanic numerical approach?, and (2) we thought you,
activity. This is one of the reasons why volca- volcanologists, knew what was happening, but it
nologists are often, correctly or incorrectly, seems you don’t know.
highly trusted professionals by the public (Hay- This chapter critically reviews both “philoso-
nes et al. 2008; Donovan et al. 2011). phies” of eruption forecasting and tracking of
During the last decade, this “untouchable volcanic unrest and related hazards, in search of a
aura” around volcano monitoring based on combined approach that could become a guide-
deterministic research has vanished with the line for future volcanic surveillance architectures.
introduction of probabilistic hazard and eruption But we still need bridges between two schools
forecasting (e.g., Newhall and Hoblitt 2002; (deterministic and probabilistic) apparently
Sparks 2003; Marzocchi et al. 2004, 2008; speaking a different language. Remember that
Sparks and Aspinall 2004; Marzocchi and Beb- both methodologies aim for the same goal: the
bington 2012; Sobradelo et al. 2014; Sobradelo timely recognition when volcanoes become
and Marti 2015). A recent chapter by Newhall hazardous in their various ways of expressions.
and Pallister (2015) starts from the same false This is our common professional and social
dichotomy, aiming to spouse the deterministic responsibility as volcanologists.
and probabilistic points of view in the highly
applicable method of “Multiple Data Sets”.
Marzocchi and Woo (2007) propose a rational 2 Forecasts based on Deterministic
on decision-making based on the hazard/risk Research
separation principal, using a cost-benefit analyses
as the guiding tool. The goal of volcano monitoring based on deter-
Among the methods for probabilistic hazard ministic research is to link temporal variations of
assessment and eruption forecasting, many are physical-chemical parameters with variations in
based on a Bayesian approach (e.g., Marzocchi the state of unrest of the physical object volcano
et al. 2004, 2008; Sobradelo et al. 2014; Sobra- (i.e. unrest, magmatic unrest, non-magmatic
delo and Marti 2015), that allows describing the unrest, eruption, hazard; Rouwet et al. 2014).
probability of interest not as a single numerical Every volcanic eruption is intrinsically preceded
value, but as a probability distribution. In this by magma rise towards the surface. The major
Deterministic Versus Probabilistic Volcano Monitoring … 39

aim in eruption forecasting is the quick recog- obvious conclusion, yet, there are numerous
nition of such magma rise by changes in the cases in which activity stopped or pulled back,
physical parameters (deformation, seismicity) sometimes for years before an actual eruption.
and chemical parameters. The most direct way to So, the deterministic approach, which is based
do so is to determine how, where, when and why mostly on a recurrence interval and ‘experience’
the physical object “volcano” responds to magma of the volcanologist, is limited in providing a
rise. sound time scale for the evolution of the ‘activ-
Despite the straight-to-goal approach, large ity’, and hence, for a forecast. In deterministic
uncertainties exist: (1) a volcano remains a monitoring the “time” concept is not unambigu-
complex system (aleatory uncertainty), and ously defined, can be case dependent or even
(2) our knowledge on the volcano remains lim- change within the evolution of an unrest phase.
ited (epistemic uncertainty). How do we know if Fortunately, with converging signals through
we have detected all the signals the volcano time, the monitoring time window often becomes
eventually releases? Which of these signals are narrower when building up towards increased
we considering in our forecast framework, and unrest or eruption, although the exact time win-
why? Sometimes we may dismiss some signals dow cannot be rigorously chosen.
as being not pertinent, or simply because we are Besides the instrumental accuracies (detection
unable to correlate them either with our ‘under- limits, analytical errors, data quality), the uncer-
standing’, or with the rest of the signals. Some tainty of the forecast cannot be quantified before
‘signals’ are considered as stand-alone, at the an eruption. The only way to decrease this “un-
moment they occur, others are considered within quantifiable” uncertainty is by increasing our
a time evolution. knowledge on volcanoes, be it the specific vol-
The quality of the forecast largely depends on cano in unrest or any volcano that has shown
the interpretation of the signals and the similar behaviour in the past. The current
hypothesis/model of future activity developed as development of methods to increase the quality
a result of this assumed scientific stringency. The (e.g., novel approaches, numerical modelling)
latter is related to the experience and expertise of and quantity of data (increase frequency, e.g., by
the deterministic researcher, or better, on how the remote sensing and real-time transmission) will
experience and expertise is perceived by indi- undoubtedly help to achieve better insights into
vidual researchers or groups, decision makers volcanic systems.
and the researcher her/himself. It is known that
the most informative and valid opinion may not
always be that of the most respected or distin- 3 Probabilistic Forecasts
guished professional (Selva et al. 2012).
A big advantage in volcano monitoring based Probabilistic methods and tools for both short-
on deterministic research is the fact that, if and long-term time windows are more and more
independent monitoring approaches (e.g., geo- in the spotlight (Marzocchi et al. 2008; Sandri
chemistry vs. geophysics) point toward a similar et al. 2009, 2012, 2014; Lindsay et al. 2010;
hypothesis on future hazard in time and space, Selva et al. 2010, 2011, 2012; Sobradelo et al.
the future scenario will become more likely. 2014; Sobradelo and Marti 2015; Bartolini et al.
Finding a larger number of arguments in favour 2013; Becerril et al. 2014). A key review on
of certain scenarios is surely an efficient way to probabilistic volcano monitoring can be found in
decipher volcanic unrest. Marzocchi and Bebbington (2012).
The timescale of the forecast is highly Within this opinion chapter, we highlight
ambiguous and based on the limited knowledge some critical aspects of the probabilistic fore-
on how the volcano (or analogue volcanoes) casting method, without entering in the technical
behaved in the past within the desired time-scale. and operational details (see Tonini et al. 2016
Sometimes all the ‘signals’ converge towards an and Sandri et al. 2017 for further reading).
40 D. Rouwet et al.

A probabilistic forecast can provide a global but intrinsic uncertainty, that can somehow be
clear, numerical view of the opinion of, gener- overcome by quite robust methods for estimating
ally, a group of people, based on the volcanic completeness of sections of catalogues (Moran
history and knowledge of the volcano. Lately, et al. 2011). The most intuitive solution, simply
probabilistic forecasts are more and more applied choosing a smaller dataset, usually representing
in real crisis situations. Thus far, the efficiency or the most recent years/centuries of a volcano’s
accuracy have hardly been evaluated, probably activity reduces this uncertainty. But this choice
due to the fact that only recently we are reaching alone will be reflected in the quality of the
statistically relevant numbers of cases to test this probabilistic assessment. The unknowns of the
critical issue (Newhall and Pallister 2015). Once data catalogue represent the uncertainty in a
high numbers of applications are reached, the probabilistic framework, thus forcing volcanol-
numerical outcomes of probabilistic methods can ogists to ‘select’ how far to track back in time,
even be considered to support long-term hazard and which information to use. In other words, we
analyses, by becoming input information itself. select e.g., only the last 300 years of activity of a
In practice, probabilistic hazard assessment volcano just simply because we believe to be
and eruption forecasting frameworks constructed more certain on what happened, instead of
on the Event-Tree methodology (Newhall and choosing the last 2000 years. What is the real
Hoblitt 2002; Newhall and Pallister 2015) rely on scientific control of these choices? We cannot
a dataset of information about the past activity of say with an acceptable certainty that a volcano
a volcano (i.e. past data), theoretical/ will behave like it did in the last 300 or
mathematical models (i.e. prior data) and a ser- 2000 years. Indeed, with the recent probabilistic
ies of monitoring signals (i.e. parameters) that methods we are able to quantify the uncertainty
allow us to track the changes in the system with of such choices but we are yet to find a sound
time (Marzocchi et al. 2008; Sobradelo et al. scientific mechanism that allows us to make
2014). This information allows us to compute the objective decisions regarding the data set. After
probabilities of a specific hazardous outcome. As all, the end goal of eruption forecasting is to give
any such application reveals, the quality of the a prediction by analysing signals from an extre-
numerical output depends on the quality and mely complex system governed by a large degree
quantity of the input. The risk exists that using a of freedom.
dataset for long-term probabilistic assessment Another aspect to tackle is the use of moni-
will introduce an uncertainty, since the operators toring information, especially for short-term
are often biased by the hypothesis or model forecasts. Asking for numerical thresholds for
coming forth of the dataset. Data should hence be monitoring parameters at the various nodes of
considered “just” facts. In both deterministic and event tree structures is intrinsically wrong, as a
probabilistic hazard assessment, volcanologists numerical threshold is an expression of certainty
rely on information about past eruptions: erup- on something we cannot be certain about. For
tive behaviour, eruption frequency, and eruption this, volcanologists use monitoring parameters in
style. Such catalogues of information are inevi- order to detect anomalies with respect to the
tably incomplete. For instance, traces of smaller volcano’s background activity to be able to track
scale events could have been literally eroded their evolution with time. Moreover, from the
away from the geological record, buried or beginning, we rely on a subjective choice when
masked by larger events and, hence, relics of we define the unrest, unrest being commonly
precursory activity cannot be deduced. Conse- agreed upon as a state of elevated activity above
quently, one should limit the part of the cata- background that causes concern (Phillipson et al.
logue used, for the period and specific kind of 2013). This cause of concern, expressed in
event you desire to forecast, for which it is rea- numerical thresholds is a subjective choice:
sonably complete. As such, the foundation of a experts involved in volcano monitoring usually
probabilistic framework represents a source of decide thresholds above/below which the
Deterministic Versus Probabilistic Volcano Monitoring … 41

volcano is considered in unrest. But is a volcano the experience they have, on every parameter
really in unrest simply because we observe one within a monitoring setup, with the goal to reach
day an anomaly in one of the parameters? And if consensus about the most likely scenario/
so, how is the choice of a threshold scientifically threshold. All data and interpretations should be
sound, since most of the time it is based on the heard and evaluated. However, this process is
“expert’s experience”? Of course, expert’s still unduly influenced by the “stronger voice” of
experience should not be dismissed and never the group. We may be certain on something until
replaced by computer codes, but indeed, such a someone else makes us doubt it. At the moment
choice is associated with a large uncertainty that we start doubting our opinions we will be easily
is extremely difficult to quantify. Even the act of influenced by other, stronger opinions. On the
reducing the uncertainty of such a choice relies other hand, volcanologists are often forced to
on corroboration with information from other make such decisions and be liable for their
sources (e.g., analogue volcanoes) that is again choices (in court of law, Bretton et al. 2015). The
subjective itself. pressure of a volcanic ‘crisis’ and the feeling of
The “fuzzy-threshold approach” (upper and liability increases scientists’ reservations when
lower thresholds) somehow resolves this problem, faced with such choices.
as it tracks the degree of anomaly, emphasising A suggestion to improve the expert elicitation
from a state in which the volcano is ‘not causing process is to introduce a person to act as a
concern’ to one ‘causing a degree of concern’. “Devil’s advocate”. This will mean that one of
Thresholds can be case-dependent and are there- the experts is supposed to do exactly the opposite
fore in most cases themselves biased. Especially to the group’s decision. If most of the experts
in the case of poorly monitored volcanoes, or of agree on a scenario, it is the duty of the latter to
volcanoes without a monitored stage of unrest completely disagree and evaluate the opposite
(e.g., towards higher nodes in the event tree), scenario. This can be a good way to “account”
Boolean (Y/N) parameters are highly preferable. for surprise scenarios. Other ways are to weight
During an unrest crisis it might be tempting to final results according to anonymous calibration
adapt the numerical values of the thresholds, tests (Cooke method; Cooke 1991; Aspinall et al.
when e.g., the previous threshold is exceeded 2003; Aspinall 2006, 2010) and/or anonymous
while the volcano does not “react on this estimation of the most reliable members of the
parameter” as we thought it would have. Nev- group, not necessarily the loudest.
ertheless, once thresholds for parameters are The introduction of probabilistic hazard
fixed, they should not be modified, in order to assessment methods in the multi-disciplinary
track the time evolution of probabilities (and volcanological community has first led to a dis-
related uncertainties). credit of the purely deterministic approach. After
the usefulness of the probabilistic approach has
been demonstrated, and confusion on the differ-
4 Recommendations: Not “or” ent philosophies has disappeared, or at least
But “and” decreased, the awareness on framing the various
“niche” research branches in a bigger picture
4.1 Expert Elicitation: A Solution? resulted into constructive discussions among the
various research groups and individuals
In general, any choice made by an expert panel involved. This results in coherent group thinking
regarding when a volcano enters a phase of and a more collaborative atmosphere among
unrest, what information is pertinent for hazard volcanologists. Expert elicitation on its turn has
analyses and how to interpret the precursory obliged researchers with various backgrounds to
signals is done by a discussion-based elicitation absorb new data and ideas from one another.
process. Each expert in a specific volcanological This is definitely a positive side-effect of the
sub-domain will exercise their opinions, based on probabilistic approaches and expert elicitations.
42 D. Rouwet et al.

4.2 How to Interpret Uncertainties? 4.3 Trust in Scientists?

Probabilistic volcanic hazard assessment and Even the best scientists can make mistakes. If
eruption forecasting is a relatively new concept in volcanic unrest or activity is badly forecasted,
modern volcanology, and often reserved for those initial trust in scientists may dissolve in legal
with a background in statistics. But, as reality proceeding. Ideally, scientists who act to the best
showed (Constantinescu et al. 2016), most of of their knowledge should be protected rather
the volcanologists are not fully aware of the than being blamed if they make a bad forecast
probability theory and how its results should be (Bretton et al. 2015). Trust in scientists depends
interpreted. A panel of volcano experts usually on how these scientists are portrayed to society
comprises seismologists, geochemists, geode- and actually how scientific aspects of volcanol-
sists, geologists, petrologists, and not all of them ogy are presented to the public. This should be a
necessarily have a background in probabilistic system with two-way feedback (Christie et al.
approaches, especially when such approach is 2015). Many countries lack volcano-education
based on the integration of opinions of all mem- among communities, but people know that there
bers of such a group. One idea to cope with this are some scientists that ‘know what they are
problem is to train the group members in how the doing’. People feel protected, but this is a false
probabilistic methodology works and how results feeling of safety, propped up by ignorance of the
should be interpreted, while another member of real situation. When disaster strikes, scientists are
the group (the so-called “PROB-runner”) objec- often the first to blame. If the community will be
tively uses the information provided by the expert involved fully in the mitigation and preparedness
panel to produce the probabilistic forecast. In this process (e.g., Gregg et al. 2004; Rouwet et al.
way, the members of the group can interpret the 2013; Dohaney et al. 2015), they may be guided
numerical outcomes accounting for the associated by ‘compassion’ and will understand that vol-
uncertainty without questioning and second- canologists cannot stop an eruption and protect
guessing the output because they are already people, and eventually the blame or trust too often
aware of the process (Newhall and Pallister 2015; falls on the elected authorities. Elected authorities
Constantinescu et al. 2016). should be the liaison between science and the
The whole idea of the elicitation approach is general public. Trust is inherent when you are
to allow your mind to explore each possibility aware of the problem and the person dealing with
without influencing one of the possible outcomes it. Trust in scientists may grow because they
just because one expert believes more in one successfully predict an eruption, but sustained
outcome than the other. It is some sort of letting growth in trust is due to multi-yearly exposure of
go. People don’t like to admit they might be the scientific staff to the public (Christie et al.
wrong, so the Event Tree and Cooke elicitation 2015). This involves long years of planning,
approaches allow them to anonymously change investing and engagement in educational cam-
their views upon elicitation. If one is capable of paigns. Trust is something that comes in time and
admitting fallibility and look at the big picture involves a feedback loop between people and the
with an open mind, allowing all possibilities to scientists.
unfold, then full discussion can occur and Within the current scope of this opinion chap-
resulting estimates of probabilities will eventu- ter, the probabilistic method often seems to serve
ally have lower uncertainties. In the end, the idea as a more transparent way of bridging between the
of probabilities is that all scenarios are possible scientists and the elected authorities (decision
to happen, some with a larger probability others makers). First, the use of probabilities inherently
with a lower one; all have probabilities (and implies some uncertainty, which in scientifically
related uncertainties) and nothing should be dis- literate society is essential for public trust. Second,
missed simply because ‘I strongly believe it can’t elicitation tools reflect a joint-opinion of a group
be, so I don’t agree’. of experts rather than of one. There is a need for
Deterministic Versus Probabilistic Volcano Monitoring … 43

training and a full understanding of probabilistic efficiently interact with the researchers that run
results by the officials. The role of probabilistic the probabilistic models (e.g., VUELCO simu-
tools should, in our opinion, not intervene in lation exercises). The latter need more detailed
communication protocols with the lay public, but feedback and input information for single
should be rather “restricted” to transmit informa- parameters at the various nodes of the event-tree.
tion to decision-makers, representing part of the Despite refining the probabilistic model during
voice of the group of volcanologists. The efforts in pre-crisis by expert elicitations, in the heat of the
communicating towards the lay public, in order to moment of the crisis, a wrong interpretation of a
build “trust” amongst the population, should be numerical value provided by determinists will
decoupled from the background of the involved sometimes lead to disastrous numerical outcomes
scientist (deterministic or probabilistic). Com- in the probabilistic models. The PROB-runner
munication protocols are independent of the cannot be blamed for not being an expert in all
applied scientific method, and researchers should fields in volcanology, incorporated through
become more skilled to transmit their information parameters in e.g., BET or HASSET.
openly towards the public, with the awareness of Three solutions to this crucial issue are pro-
the uncertainty their information contains (Hicks posed: (1) probabilistic model runners should
and Few 2015). actively take part in the “closed-door” discus-
sions by the experts from the various fields,
before incorporating numerical values in proba-
4.4 Towards Collaborative Volcano bilistic models, (2) a separate team of experts that
Monitoring profoundly know the needs and functioning of
probabilistic models should flank the
A major accomplishment of the probabilistic PROB-runners during crisis. Both realities are
method is to have increased harmony among the not yet accomplished, and/or (3) an event tree
various deterministic research environments. structure, put forward by a facilitator between the
This new dynamic favours the refining of pre- deterministic and probabilistic research teams
vious conceptual models that originate from should serve as the base to guide scientific dis-
deterministic research, as the reference frame has cussion and get fast to the nucleus of the crisis
become more complete. (Newhall and Pallister 2015). Future simulation
Nevertheless, the probabilistic research exercises on volcanic crisis situations should
approach is not yet fully accepted by the deter- focus on this training approach, in order to be
ministic community due to criticism and anxiety prepared when real the crisis strikes.
to be “replaced” by the probabilistic method Moreover, the outcomes of probabilistic
(VUELCO simulations Colima, Campi Flegrei, models have to be included in the final reports
Cotopaxi and Dominica). This concern is transmitted to authorities, and be respected as
unnecessary, since the first requisite for the one of the many monitoring tools, without
probabilistic method to function is the availabil- decreasing or increasing their weight and value
ity of data, information, a priori believes and within the still deterministic-dominated general
models, originating from deterministic research. opinion. Since a probabilistic framework offers a
More input information for the probabilistic measure of the uncertainty, any interpretation
method means significant decreases in the epis- should not be taken for granted, neither decision
temic uncertainty of probabilistic outcomes. makers should make decisions based solely on a
Moreover, during volcanic crisis situations, probability. Probabilities should be considered as
deterministic researchers still stick to the “round an addition to the information upon which deci-
table” approach and the lack of time inhibits to sions are made, and not as a decisive factor.
44 D. Rouwet et al.

5 Take Home Ideas date, probabilistic forecasts have not been rigor-
ously evaluated to know whether they are an
Probabilistic forecasting has become an inherent improvement over traditional, non-probabilistic
part within a multi-facet view of research and vol- forecast methods. For sure, they do better than
cano monitoring; neither deterministic, nor proba- traditional methods at estimating uncertainty. We
bilistic methods, can or should stand alone. More still need tests on whether they are more accurate,
than being a means to transmit information between and more useful for decision makers than older
volcanologists and decision making authorities, methods (Newhall and Pallister 2015).
probabilistic models should also be based on, and In the future, reports should include the forecast
promote, deterministic research that can be written of probabilistic models; a monitoring architecture
up after “round table” discussions (Fig. 1). should expand beyond the classical “seismo-
Incorporation of probabilistic models in vol- deformation-gas” setup and become “seismo-
cano monitoring has many advantages: (1) pro- deformation-gas-probability” setup (in random
tecting against oversimplified, over-confident order of importance) (Fig. 1). Probabilistic models
forecasts. Even though decision makers may ini- cannot stand alone, as they need the input and
tially have difficulties to understand uncertainties feedback from deterministic research. “Proba-
and prefer black-on-white numbers, Y/N forecasts, bilists” should not communicate their numerical
they will soon come to appreciate probabilities if outcomes directly to the decision-making authori-
they are represented in an understandable way; ties: it is better to convey the opinion of an entire
(2) creating harmony amongst the volcanological group. Probabilistic methods can serve as “mod-
community because probabilities will reflect the erator” among the various disciplines, while expert
general view of the monitoring team, and (3) le- elicitations are the “glue” between the deterministic
gally protecting the entire monitoring team by and probabilistic approaches (Fig. 1). Probabilistic
probabilities and their uncertainties as forecasts are methods should knock down walls and stimulate
perfectly traceable and reproducible, if disaster discussion and coherence amongst the various
strikes after “erroneous forecasts”. However, to research branches (seismologists, geodesists,

Fig. 1 From dichotomic monitoring setups (deterministic Mexico (November 2011, D.R.). Before interpretation of
vs. probabilistic forecast, “or” setups) to an “and” strategy. the data (monitoring data, eruptive history or any a priori
The picture shows VUELCO target volcano Popocatépetl, model) the volcano is considered a “black box”
Deterministic Versus Probabilistic Volcano Monitoring … 45

geochemists, petrologists, geologists). This requires in Kona communities from Mauna Loa and Hualālai
time, effort and an open-mind by all involved volcanoes, Hawai’i. J Volcanol Geotherm Res
130:179–196
parties/volcanologists in volcano monitoring. Haynes K, Barclay J, Pidgeon N (2008) The issue of trust
and its influence on risk communication during a
volcanic crisis. Bull Volcanol 70:605–621. doi:10.
References 1007/s00445-007-0156-z
Hicks A, Few R (2015) Trajectories of social vulnerability
during the Soufrière Hills volcanic crisis. J Appl
Aspinall WP (2006) Structured elicitation of expert judge- Volcanol 4:10. doi:10.1186/s13617-015-0029-7
ment for probabilistic hazard and risk assessment in Lindsay J, Marzocchi W, Jolly G, Constantinescu R,
volcanic eruptions. In: Mader HM, Coles SG, Con- Selva J, Sandri L (2010) Towards real-time eruption
nor CB, Connor LJ (eds). Statistics in volcanology. forecasting in the Auckland Volcanic Field: applica-
IAVCEI Publications. ISBN: 978-1-86239-208-3. tion of BET_EF during the New Zealand National
pp 15–30 Disaster Exercise ‘Ruaumoko’. Bull Volcanol 72:185–
Aspinall WP (2010) A route to more tractable expert 204. doi:10.1007/s00445-009-0311-9
advice. Nature 463(21):294–295 Marzocchi W, Woo G (2007) Probabilistic eruption
Aspinall WP, Woo G, Voight B, Baxter PJ (2003) forecasting and the call for an evacuation. Geophys
Evidence-based volcanology: application to vol canic Res Lett 34:L22310. doi:10.21029/22007GL031922
crises. J Volcanol Geotherm Res 128:273–285 Marzocchi W, Bebbington M (2012) Probabilistic erup-
Bartolini S, Cappello A, Martí J, Del Negro C (2013) tion forecasting at short and long time scales. Bull
QVAST: a new quantum GIS plugin for estimating Volcanol. doi:10.1007/s00445-012-0633-x
volcanic susceptibility. Nat Hazards Earth Syst Sci Marzocchi W, Sandri L, Gasparini P, Newhall C,
13:3031–3042. doi:10.5194/nhess-13-3031-2013 Boschi E (2004) Quantifying probabilities of volcanic
Becerril L, Bartolini S, Sobradelo R, Marti J, Morales JM, events: The example of volcanic hazard at Mount
Galiendo I (2014) Long-term volcanic hazard assess- Vesuvius. J Geophys Res 109:B11201. doi:10.1029/
ment on El Hierro (Canary Islands). Nat Hazards Earth 2004JB003155
Syst Sci 14:1853–1870. doi:10.5194/nhess-14-1853- Marzocchi W, Sandri L, Selva J (2008) BET_EF: a
2014 probabilistic tool for long- and short-term erup tion
Bretton RJ, Gottsmann J, Aspinall WP, Christie R (2015) forecasting. Bull Volcanol 70:623–632. doi:10.1007/
Implications of legal scrutiny processes (including the s00445-007-0157-y
L’Aquila trial and other recent court cases) for future Moran SC, Newhall C, Roman DC (2011) Failed
volcanic risk governance. J Appl Volcanol 4:18. magmatic eruptions: late-stage cessation of mag ma
doi:10.1186/s13617-015-0034-x ascent. Bull Volcanol 73:115–122. doi:10.1007/
Christie R, Cooke O, Gottsmann J (2015) Fearing the s00445-010-0444-x
knock on the door: critical security studies insights Newhall C, Hoblitt RP (2002) Constructing event trees for
into limited cooperation with disaster management volcanic crises. Bull Volcanol 64:3–20. doi:10.1007/
regimes. J Appl Volcanol 4:19. doi:10.1186/s13617- s004450100173
13015-10037-13617 Newhall C, Pallister J (2015) Using multi data sets to
Constantinescu R, Robertson R, Lindsay JM, Tonini R, populate probabilistic volcanic event trees. In:
Sandri L, Rouwet D, Patrick Smith P, Stewart R (2016) Schroder JF, Papale P (eds) Volcanic hazards, risks,
Application of the probabilistic model BET_UNREST and disasters. Hazards and disasters series. Elsevier,
during a volcanic unrest simulation exercise in Amsterdam, pp 203–232
Dominica, Lesser Antilles. Geochem Geophys Geosyst Phillipson G, Sobradelo R, Gottsmann J (2013) Global
17:4438–4456. doi:10.1002/2016GC006485 volcanic unrest in the 21st century: an ana lysis of the
Cooke RM (1991) Experts in uncertainty: opinion and first decade. J Volcanol Geotherm Res 264:183–196
subjective probability in science. Oxford University Rouwet D, Iorio M, Polgovsky D (2013) A science & arts
Press. ISBN: 9780195064650 sensitization program in Chapultenango, 25 years
Dohaney J, Brogt E, Kennedy B, Wilson TM, Lindsay JM after the 1982 El Chichón eruptions (Chiapas, Mex-
(2015) Training in crisis communication and volcanic ico). J Appl Volcanol 2:6. doi:10.1186/2191-5040-2-6
eruption forecasting: design and evaluation of an Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J,
authentic role-play simulation. J Appl Volcanol 4:12. Tonini R, Papale P (2014) Recognizing and tracking
doi:10.1186/s13617-015-0030-1 hazards related to non-magmatic unrest: a review.
Donovan A, Oppenheimer C, Bravo M (2011) Social J Appl Volcanol 3:17. doi:10.1186/s13617-014-0017-3
studies of volcanology: knowledge generation and Sandri L, Guidoboni E, Marzocchi W, Selva J (2009)
expert advice on active volcanoes. Bull Volcanol. Bayesian event tree for eruption forecasting (BET_EF)
doi:10.1007/s00445-011-0547-z at Vesusvius, Italy: a retrospective forward application
Gregg CE, Houghton BF, Johnston DM, Paton D, to the 1631 eruption. Bull Volcanol 71:729–745.
Swanson DA (2004) The perception of volcanic risk doi:10.1007/s00445-008-0261-7
46 D. Rouwet et al.

Sandri L, Jolly G, Lindsay J, Howe T, Marzocchi W Selva J, Orsi G, Di Vito MA, Marzocchi W, Sandri L
(2012) Combining long-term and short-term proba- (2011) Probability hazard map for future vent opening
bilistic volcanic hazard assessment with cost-benefit at Campi Felgrei caldera. Bull Volcanol, Italy. doi:10.
analysis to support decision making in a volcanic 1007/s00445-011-0528-2
crisis from the Auckland Volcanic Field, New Sobradelo R, Bartolini S, Marti J (2014) HASSET: a
Zealand. Bull Volcanol 74:705–723. doi:10.1007/ probability event tree tool to evaluate future volcanic
s00445-011-0556-y scenarios using Bayesian inference. Bull Volcanol
Sandri L, Thouret J-C, Constantinescu R, Biass S, 76:770. doi:10.1007/s00445-013-0770-x
Tonini R (2014) Long-term multi-hazard assessment Sobradelo R, Marti J (2015) Short-term volcanic hazard
for El Misti volcano (Peru). Bull Volcanol 76:771. assessment through Bayesian inference: retrospective
doi:10.1007/s00445-013-0771-9 application to the Pinatubo 1991 volcanic crisis.
Sandri L, Tonini R, Rouwet D, Constantinescu R, J Volcanol Geotherm Res 290:1–11. doi:10.1016/j.
Mendoza-Rosas AT, Andrade D, Bernard B (2017) jvolgeores.2014.11.011
The need to quantify hazard related to non-magmatic Sparks RSJ (2003) Forecasting volcanic eruptions. Earth
unrest: from BET_EF to BET_UNREST. In: Gotts- Planet Sci Lett 210:1–15. doi:10.1016/S0012-821X
mann J, Neuberg, J, Scheu B (eds) Volcanic unrest: (03)00124-9
from science to society—IAVCEI advances in vol- Sparks RSJ, Aspinall WP (2004) Volcanic activity:
canology. Springer frontiers and challenges in forecasting, pre diction
Selva J, Costa A, Marzocchi W, Sandri L (2010) and risk assessment. The State of the Planet: Frontiers
BET_VH: exploring the influence of natural uncer- and Challenges in Geo physics. Geophysical Mono-
tainties on long-term hazard from tephra fallout at graph 150, IUGG Volume 19. doi:10.1029/150GM28
Campi Flegrei (Italy). Bull Volcanol 72:717–733. Tonini R, Sandri L, Rouwet D, Caudron C, Marzocchi W,
doi:10.1007/s00445-010-0358-7 Suparjan P (2016) A new Bayesian Event Tree tool to
Selva J, Marzocchi W, Papale P, Sandri L (2012) track and quantify unrest and its application to Kawah
Operational eruption forecasting at high-risk vol Ijen volcano. Geochem Geophys Geosyst 17:2539–
canoes: the case of Campi Flegrei, Naples. J Appl 2555. doi:10.1002/2016GC006327
Volcanol 1:5. http://www.appliedvolc.com/
content/1/1/5

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter are
the Creative Commons Attribution 4.0 International included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Probabilistic E-tools for Hazard
Assessment and Risk Management

Stefania Bartolini, Joan Martí, Rosa Sobradelo

and Laura Becerril

Abstract
The impact of a natural event can significantly affect human life and the
environment. Although fascinating, a volcanic eruption creates similar or even
greater problems than more frequent natural events due to its multi-hazard
nature and the intensity and extent of its potential impact. It is possible to live
near a volcanic area and take advantage of the benefits that volcanoes offer, but
it is also important to be aware of the existing threats and to know how to
minimise risks. In this chapter, we present an integrated approach using
e-tools for assessing volcanic hazard and risk management. These tools have
been especially designed to assess and manage volcanic risk, to evaluate long-
and short-term volcanic hazards, to conduct vulnerability analysis, and to
assist decision-makers during the management of a volcanic crisis. The
methodology proposed here can be implemented before an emergency in
order to identify optimum mitigating actions and how these may have to be
adapted as new information is obtained. These tools also allow us identifying
the most appropriate probabilistic and statistical techniques for volcanological
data analysis and treatment in the context of quantitative hazard and risk
assessments. Understanding volcanic unrest, forecasting volcanic eruptions,
and predicting the most probable scenarios, all imply a high degree of inherent
uncertainty, which needs to be quantified and clearly explained when
transmitting scientific information to decision-makers.

Resumen
El impacto de un evento natural puede afectar significativamente la vida
humana y el medio ambiente. Aunque fascinante, una erupción volcánica

S. Bartolini (&)
J. Martí
L. Becerril L. Becerril

Group of Volcanology, SIMGEO (UB-CSIC), e-mail: [email protected]
Institute of Earth Sciences Jaume Almera,
ICTJA-CSIC, Lluís Solé I Sabarís S/N, Barcelona R. Sobradelo
08028, Spain WRN Earth Hub Leader, Analytics Technology &
e-mail: [email protected] Willis Research Network, London, UK
e-mail: [email protected]
J. Martí
e-mail: [email protected]

Advs in Volcanology (2019) 47–61

DOI 10.1007/11157_2017_14
© The Author(s) 2017
Published Online: 15 July 2017
48 S. Bartolini et al.

puede generar problemas parecidos o incluso mayores que otros eventos

naturales más frecuentes, ya que se trata de un fenómeno multipeligro y
con un impacto potencial de gran intensidad y magnitud. Vivir cerca de
una zona volcánica es posible considerando sus innumerables beneficios,
pero sin embargo debemos ser conscientes de la amenaza existente y
tomar las medidas oportunas para minimizar el riesgo. En este capítulo, se
presenta un enfoque integrado que utiliza herramientas para la evaluación
de la peligrosidad y del riesgo volcánico, el análisis de vulnerabilidad, y
para ayudar en la gestión de una crisis volcánica. La metodología aquí
propuesta puede ser implementada antes de una emergencia con el fin de
determinar las medidas óptimas de mitigación y como pueden adaptarse a
medida que se obtiene nueva información. Además, estas herramientas se
basan en las técnicas probabilísticas y estadísticas más adecuadas para el
análisis de datos vulcanológicos y su implementación en el contexto de las
evaluaciones cuantitativas del peligro y riesgo. La interpretación del
unrest, la predicción de las erupciones volcánicas, y la identificación de
los escenarios más probables, implican un alto grado de incertidumbre,
que debe ser cuantificado y claramente explicado para transmitir
correctamente la información científica a los gestores de las crisis
volcánicas.

Keywords


Volcanic risk Volcanic hazard
Vulnerability
Risk management



Decision-making E-tools

pre-unrest phase that includes long-term assess-

Introduction
ment, hazard and risk mapping, and estimation
regarding expected scenarios, volcano monitor-
One of the most important tasks in modern vol-
ing, and emergency planning; (ii) the unrest
canology is to manage volcanic risk and, conse-
phase, which includes short-term assessment,
quently, to minimise it. Forecasting volcanic
alert, communication, and information proce-
eruptions and predicting the most probable sce-
dures, the implementation of emergency mea-
narios are tasks that are subject to high degrees of
sures, and the interpretation of eruption
uncertainty but which need to be quantified and
precursors; (iii) the volcanic event itself, repre-
clearly explained when transmitting scientific
sented by a major change in the state of the
information to decision-makers. Assessing erup-
volcano; decisions can be revised as new infor-
tion risk scenarios in probabilistic ways has
mation is obtained and the evolution of the vol-
become one of the main challenges tackled by
canic event is updated; (iv) the post-event phase
modern volcanology (Newhall and Hoblitt 2002;
characterized by rescue and recovery.
Marzocchi et al. 2004; Aspinall 2006; Neri et al.
Previous studies, focusing attention on the
2008; Sobradelo et al. 2014).
first phase of volcanic crisis management, have
The volcanic management cycle consists of
developed different methodologies for evaluating
four phases (Sobradelo et al. 2015): (i) the
Probabilistic E-tools for Hazard Assessment … 49

hazards. Most of these studies are based on the areas and retrospectively for a number of vol-
use of simulation models and Geographic canic crisis (Martí et al. 2012; Sobradelo et al.
Information Systems (GIS) that enable volcanic 2014; Bartolini et al. 2014a, b; Becerril et al.
hazards such as lava flows, pyroclastic density 2014; Scaini et al. 2014), to date they only exist
currents (PDCs), and ash fallout to be modeled as academic exercises. To convert them into
and visualised (Felpeto et al. 2007; Toyos et al. practical tools ready to be used by Civil Protec-
2007; Cappello et al. 2012; Martí et al. 2012; tion managers and decision-makers, they must be
Becerril et al. 2014; Bartolini et al. 2014a, b). checked and tested, and then adapted to the real
The use of spatio-temporal data and GIS has needs of end users.
now become an essential part of integrated Probabilities are still the best outcome of
approaches to disaster risk management in light scientific forecasting. However, they are not
of the development of new analysis/modelling easily understood. Understanding the potential
techniques. These spatial information systems evolution of a volcanic crisis is crucial for
are used for storage, situation analysis, mod- designing effective mitigation strategies. One of
elling, and visualisation (Twigg 2004). Nor- the main issues when managing a volcanic crisis
mally, these studies present a systematic is how to make scientific information under-
approach based on the estimation of spatial standable for decision-makers and Civil Protec-
probability, that is, a susceptibility analysis tion managers. Thus, we need quantitative
(Martí and Felpeto 2010), a temporal analysis risk-based methods for decision-making under
based on Bayesian inference (Marzocchi et al. conditions of uncertainty that can be developed
2004; Sobradelo et al. 2014), or the evaluation of and applied to volcanology. In order to resolve
hazards (Felpeto et al. 2007; Bartolini et al. this problem and to take a step forward in min-
2014a, b; Becerril et al. 2014) and vulnerability imising risks, we have defined an integrated
(Marti et al. 2008; Scaini et al. 2014). These approach using user-friendly e-tools, which can
studies have been applied in a number of vol- be run on personal computers. They are specifi-
canic areas such as Etna, Sicily (Cappello et al. cally useful for long- and short-term hazard
2013), Tenerife, Spain (Martí et al. 2012), Peru assessment, vulnerability analysis, decision-
(Sandri et al. 2014), the island of El Hierro, making, and volcanic risk management. In this
Spain (Becerril et al. 2014), and Deception chapter, we describe the e-tools designed to
Island, Antarctica (Bartolini et al. 2014a). Other manage and to minimise volcanic risk.
procedures have been employed to assess vol-
canic hazards in Campi Flegrei, Italy (Lirer et al.
2001), Furnas (São Miguel, Azores), Vesuvius, Volcanic Risk: Hazard, Vulnerability,
Italy (Chester et al. 2002), and Auckland, New and Value
Zealand (Sandri et al. 2012).
These studies underline the fact that scientists In general, risk is defined as the probability or
are aware of the relationship between volcanic likely magnitude of a loss (Blong 2000). In
hazards and socio-economic impacts and are in volcanic risk assessment, risk depends on the
the process of developing new approaches and adverse effects of volcanic hazards and can be
models to assess its importance. One positive defined as the product of three main factors:
aspect is that, as a result, there is now a choice of volcanic hazard, vulnerability to those hazards,
freely available models; on the other hand, these and the value of what is at risk.
models are not integrated into a single platform Volcanic hazard is defined as the probability
and have been developed in a variety of different of any particular area being affected by a
programming languages. destructive volcanic event within a given period
Despite the fact that these tools have been of time (Blong 2000). The quantification and
created for application in real situations and have evaluation of the volcanic hazard allow us
been successfully tested in different volcanic determining which areas will be affected by a
50 S. Bartolini et al.

given volcanic event, and to design appropriate buildings, etc.) and productive capacity (facto-
emergency plans and territorial planning. ries, power plants, highways, etc.) exposed to the
A vulnerability assessment uses indicators to destructive events (Blong 2000).
quantify the physical vulnerability of the elements Risk management is a complex process
of each sub-system (buildings, transportation (Fig. 1) since different steps are necessary for
system, urban services, and population) and is a evaluating and minimising risk. It can be thought
measurement of the proportion of the value likely of as the sum of risk assessment—which includes
to be lost as a result of a given event (Blong 2000). risk analysis and risk evaluation—and risk con-
In fact, the values of elements-at-risk that can be trol. Risk analysis aims to improve prevention
directly or indirectly damaged by a given hazard tools through the collection and acquisition of
will vary. Each hazardous phenomenon affects data on hazards and risks, and then to dissemi-
elements and infrastructures in different ways in nate it in the form of maps and scenarios (Thierry
terms of their specific physical vulnerability; this et al. 2015). This phase is characterised by five
in turn will be depend on their number (of build- different steps: hazard identification, hazard
ings, people, etc.), monetary value, surface area, assessment, elements-at-risk/exposure analysis,
and the importance of the elements-at-risk. Indi- vulnerability assessment, and risk estimation
cators are used to determine the specific physical (Van Westen 2013). In particular, it is important
vulnerability of each hazardous phenomenon and to distinguish between long- and short-term
sub-system (Scaini et al. 2014). hazard assessments, which will vary according
The value is the number of human lives at to the expected period of time over which the
stake, together with the capital value (land, process will display significant variations.

Fig. 1 Risk management schematic: steps for evaluating and minimising risk
Probabilistic E-tools for Hazard Assessment … 51

Long-term assessment is based on historical and Finally, the second part of risk management is
geological data, as well as theoretical models, risk control, which consists of the decision-
and refers to the time window available—during making process involved in managing risks
which time the volcanic system shows no signs whose aim is to improve crisis management
of unrest—before an unrest episode occurs. On capabilities and implement risk-mitigation mea-
the other hand, short-term assessment refers to sures using the results of risk assessment as an
the unrest phase, and complementary information input (Western 2013). During this phase mea-
derived from the combination of a long-term sures must be adopted for reducing vulnerability
analysis with real-time monitoring data is needed (people and infrastructure) and developing
to update the status of the volcanic hazard (Blong recovery and resilience capacities after an event
2000). Short-term evaluation helps forecast has taken place.
where and when the eruption may take place and
the most likely eruptive scenarios.
Once the long-term risk analysis is computed, it E-tools for Volcanic Hazard and Risk
is then possible to adopt mitigation measures such Management
as land-use planning and emergency preparations
to reduce the risk. In addition, this long-term In this section, we present different e-tools that
analysis will help manage the volcanic crisis as it have been specifically designed to assess and
will constitute the basis for the short-term analysis manage volcanic risk (Fig. 2). The objective is to
and, combined with a cost-benefit analysis, will combine freely available models to produce a
assist in correct decision-making (Sobradelo et al. new approach for minimising and managing
2015). To evaluate the total risk related to a par- volcanic risk. These e-tools are based on the
ticular volcanic eruption we have to repeat the assumption that the best way to show how
evaluation of the vulnerability and the cost-benefit probabilities work is to use the possible scenarios
analysis (risk evaluation) for each possible hazard and outcomes of volcanic unrest (an increase in
scenario and then sum the results. This will allow volcanic activity that may or may not precede a
us to estimate the impact and the economic losses volcanic eruption) to design an integrated model
that will affect society and the environment, and to that can act as a descriptor of scenarios. The
identify a range of risk management alternatives. effectiveness 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 a
volcanic area and develop appropriate hazard and risk maps
52 S. Bartolini et al.

in different volcanic areas showing pros and Storing Data

contras of these approaches, such as in the case
of El Hierro Island in Canary Islands (Becerril The results of our analyses are as important as
et al. 2014; Sobradelo et al. 2015), Tenerife in the input parameters that are used. Given the
Canary Islands (Scaini et al. 2014), Deception nature, variety, and availability of the data we
Island in Antarctica (Bartolini et al. 2014a), and need to handle, one of the most important
La Garrotxa Volcanic Field in Spain (Bartolini aspects of risk assessment is the management
et al. 2014b). User manuals are available at and exchange of information (De la Cruz-Reyna
https://volcanbox.wordpress.com/ and http:// 1996). The quality of the data will determine the
www.vetools.eu. evaluation of the volcanic risk, which is an
These e-tools are designed to be implemented essential part of risk-based decision-making in
before an emergency in order to identify (i) the land-use planning and emergency management.
optimum mitigating actions and (ii) the most Some of the most relevant issues include how
appropriate probabilistic and statistical tech- and where to store the data, in which format
niques for volcanological data analysis and should they be made available, and how to
treatment in the context of quantitative hazard facilitate its use and exchange. Thus, it is
and risk assessment, and (iii) how appropriate essential to have an appropriate database that is
responses may change as new information is specifically adapted to the task of evaluating and
obtained. They constitute different steps, as managing volcanic risk. For that reason, we
shown in Fig. 3, in hazard assessment and risk designed VERDI (Volcanic managEment Risk
management. Database desIgn) (Bartolini et al. 2014c), a

Fig. 3 Steps in volcanic risk assessment applying e-tools to be implemented before an emergency
Probabilistic E-tools for Hazard Assessment … 53

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

geodatabase with an appropriate architecture for probabilistic models that are the first step in our
volcanic risk assessment and management that risk analysis.
stores the data from which we will extract the
input parameters to run our e-tools. VERDI is a
spatial database structure (Fig. 4) that allows Hazard E-tools
different types of data, including geological,
volcanological, meteorological, monitoring, and Volcanic hazard assessment consists of simulat-
socio-economic information, to be manipulated, ing eruptive scenarios for use in risk-based
organised, and managed. The data contained in decision-making, land-use planning, and emer-
this database are the basis for applying the gency management. They must necessarily be
54 S. Bartolini et al.

based on a good knowledge of the past eruptive input data sets are available for the area, the total
history of the volcano or volcanic area, and will susceptibility map is obtained by assigning dif-
reveal how volcanoes erupted in the past, thereby ferent weights to each of the PDFs, which are
providing clues to how they will erupt in the then combined via a weighted sum and modeled
future. The first step in the quantitative assess- in a non-homogeneous Poisson process. This
ment of volcanic hazards is the spatial probability e-tool has been used to evaluate susceptibility on
of occurrence of a hazard, i.e. where the next the island of El Hierro (Canary Islands) (Becerril
eruption may take place and its extent. This et al. 2014) and on Deception Island (Antarctica)
analysis is based on the development of suscep- (Bartolini et al. 2014a).
tibility maps (Martí and Felpeto 2010), that is, When monitoring data generated during an
the spatial probability of a future vent opening unrest phase are available, the QVAST e-tool can
given the past eruptive activity of a volcano and also be used to update the susceptibility map. In
the simulation of possible eruptive scenarios. fact, seismicity and surface deformation are good
Another important task is to investigate the indicators of magma movement and during vol-
temporal probability, in other words, when the canic unrest variations in shallow volcano-
next eruption will occur in the future and the type tectonic and long-period seismicity, as well as
of scenarios that are most likely to be involved. ground deformation, are observed as the magma
Thus, an evaluation of volcanic hazard enables migrates within the volcanic system (Martí et al.
us to infer where and when the next eruption may 2013). Thus, QVAST is a highly useful tool that
take place and its magnitude. can be applied to both long- and short-term
evaluations.
Spatial Analysis
Susceptibility analysis is the evaluation of the Temporal Analysis
spatial distribution of future vent openings HASSET (Hazard Assessment Event Tree),
(Fig. 5a). This challenging issue is generally developed by Sobradelo et al. (2014), is a
tackled using probabilistic methods that use the probability tool built on an event tree structure
calculation of a kernel function at each data that uses Bayesian inference to estimate the
location to estimate probability density functions probability of occurrence of a future volcanic
(PDFs). Commonly, a Gaussian kernel, describ- scenario. It also evaluates the most relevant
ing a normal distribution, is used to estimate sources of uncertainty in the corresponding
local event densities in volcanic fields, which volcanic system. Event tree structures (Newhall
will give the intensity of a new vent opening. and Hoblitt 2002) constitute one of the most
This method is based on the distance from nearby useful and necessary tools in modern volcanol-
volcanic structures and a smoothing parameter, ogy for assessing the volcanic hazard of future
also known as bandwidth. This factor is the most eruptive scenarios. An event tree is a graphic
important parameter in the kernel function and representation of events in the form of nodes
represents the degree of randomness in the dis- and branches. It evaluates the most relevant
tribution of past events. sources of uncertainty when estimating the
QVAST (QGIS for VolcAnic SuscepTibility), probability of occurrence of a future volcanic
developed by Bartolini et al. (2013), is a new tool event.
designed to generate user-friendly quantitative The objective of this e-tool is to outline all
assessments of volcanic susceptibility (e.g. the relevant possible outcomes of volcanic unrest at
probability of hosting a new eruptive vent). progressively greater detail and to assess the
QVAST allows an appropriate method for eval- hazard of each scenario by estimating its proba-
uating the bandwidth for the kernel function to be bility of occurrence within a future time interval.
selected on the basis of input parameters and the Each node of the event tree represents a step and
shapefile geometry, and can also evaluate the contains a set of possible branches (the outcomes
PDF with the Gaussian kernel. When different for that particular category). The nodes are
Probabilistic E-tools for Hazard Assessment … 55

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

alternative steps from a general prior event, state, only central eruptions), (ii) geothermal or tec-
or condition that move towards increasingly tonic unrest (as opposed to only magmatic
specific subsequent events and a final outcome. unrest), and (iii) felsic or mafic lava composition,
HASSET (Fig. 5b) uses this event tree structure as well as (iv) certain volcanic hazards as pos-
to make estimations of the probabilities for each sible outcomes of an eruption, and (v) the dis-
possibility (branches and nodes) using a statisti- tance reached by each hazard.
cal methodology known as Bayesian Inference A user-friendly interface guides the user
(Newhall and Hoblitt 2002; Marzocchi et al. through all steps and helps
2004; Sobradelo et al. 2014). In particular, and
based on comparisons with previous event trees – enter all the data needed for the analysis;
for volcanic eruptions, HASSET accounts for the – compute the estimated probability for each
possibility of (i) flank eruptions (as opposed to branch in the event tree;
56 S. Bartolini et al.

– compute the total estimated probability and important factor determining the path of a
compare up to five different scenarios. lava flow. The determination of the proba-
bility of each point being invaded by lava is
This tool has been used for determining the performed by computing several random
eruption probability on El Teide (Tenerife, paths with a Monte Carlo algorithm. Fallout
Spain) (Sobradelo and Martí 2010), El Hierro simulation models are advection diffusion
(Canary Islands) (Becerril et al. 2014), and models that assume that away from the vent
Deception Island (Antarctica) (Bartolini et al. the transport of the particles from a Plinian
2014a). column is controlled by the advective effect
In both long-term and short-term evaluations of the wind, by diffusion due to atmospheric
HASSET can be useful for determining the turbulence, and by the settling velocity of the
occurrence probabilities of eruptive scenarios and particles. The model for simulating pyro-
can assist decision-makers assess the required clastic density currents is the energy cone
mitigation actions associated with each scenario model proposed by Sheridan and Malin
and estimate the corresponding potential risk. (1983). The input parameters are the topog-
raphy, the collapse equivalent height (H), and
Simulation Models the collapse equivalent angle (h). The inter-
Simulating eruptive scenarios caused directly section of the energy cone, originating at the
(e.g. lava flows, fallout, surges) and indirectly eruptive source, with the ground surface
(earthquakes, landslides) by an eruption requires defines the distal limits of the flow.
a detailed analysis of the past activity of the – HAZMAP is a free program for simulating
volcano or volcanic area, and must take into the sedimentation of volcanic particles at
account all the possible hazards associated with discrete point sources that predicts the corre-
the eruptive activity. Volcanic hazard can be sponding ground deposits (deposit mode)
assessed via two different types of approaches: (Macedonio et al. 2005). HAZMAP is also
deterministic, which defines a maximum exten- able to evaluate the probability of overcoming
sion area affected by an eruptive episode based a given loading threshold in ground deposits
on deposits generated by past activity, or prob- by using a set of different wind profiles
abilistic, based on the probability that a certain recorded on different days (probability mode).
area will be affected by an eruptive process. In Using a statistical set of recorded wind pro-
order to generate hazard maps (Fig. 5d), it is files (and/or other input parameters), it can
important to understand past eruptive behaviour also be used to draw hazard maps for ashfall
and to employ physical simulation models that deposits. In HAZMAP, settling velocities can
will permit the behaviour of future volcanic be calculated using several models as a direct
activity to be foreseen. In this type of approach, function of particle diameters, densities, and
accurate and detailed geographic and carto- shapes. The advantage of HAZMAP is that it
graphic data are required for high-quality analy- is a simple tool able to predict ashfall during
sis with a GIS. hypothetical or real eruptions of a given
Here, we describe some of the e-tools freely magnitude and wind profile.
available for download that allow volcanic haz- – LAHARZ is a semi-empirical code for cre-
ards to be evaluated (Fig. 5c): ating hazard-zonation maps that depict esti-
mates of the location and extent of areas
– VORIS 2.0.1 is a GIS-based tool, developed inundated by lahars (Schilling 1998). The
by Felpeto et al. (2007) that allows users to input parameters for this model are the Digital
simulate lava flows, fallout, and pyroclastic Elevation Model and the lahar volume, which
density current scenarios. Lava-flow simula- provide an automated method for mapping
tions are based on a probabilistic model that areas of potential lahar inundation. These
assumes that topography is the most hazard zones can be displayed in a GIS with
Probabilistic E-tools for Hazard Assessment … 57

other types of volcano hazard information estimating the expected damage caused by vol-
such as the proximal hazard zone, infras- canic eruptions. VOLCANDAM (Fig. 6) con-
tructure, hydrology, and population, as well sists of three main parts: exposure analysis,
as contours and shaded relief, to produce vulnerability assessment, and the estimation of
volcano hazard-zonation maps. Such maps expected damages. The exposure analysis iden-
show the proximity and intersection of tifies the elements exposed to the potential haz-
potential hazard zones to people and ard and focuses on the relevant assets of the
infrastructures. study area (population distribution, social and
– TITAN2D is a computer program model economic conditions, and productive activities
developed by the University at Buffalo (Patra and their role in the regional economy). The
et al. 2005) that simulates granular flows over vulnerability analysis defines a physical vulner-
digital elevation models, based on a “thin ability indicator for all exposed elements, as well
layer model”. The input for the computer as a corresponding qualitative vulnerability
code includes simulation time, minimum index. Systemic vulnerability considers the pos-
thickness of the final deposit, internal and bed sible relevance of each element in the system and
friction angles, starting coordinates, and the their interdependencies by taking into account all
initial speed and direction of the flow. Addi- exposed and non-exposed elements (people,
tionally, this program allows users to define buildings, transportation network, urban services,
the specific starting pile dimensions or a and productive activities). Damage assessment is
dynamic flow source. The outputs from the performed by associating a qualitative damage
program (represented dynamically) are flow rating to each combination of hazard and vul-
depth and momentum, which yield the nerability, bearing in mind their specific contexts
deposit limit, run-out path, average flow and roles in the system. The way one element can
velocity, inferred deposit thickness, and travel be damaged—and thus lose its functionality—
time. depends in fact on the type of hazardous event
and the characteristics of the element. The result
is damage maps that can be displayed at different
levels of detail, depending on user preferences.
Vulnerability E-tool This tool aims to facilitate territorial planning
and risk management in active volcanic areas.
Once we have obtained hazard maps, the next
step consists of adding population, infrastruc-
tures, and land-use data to evaluate the vulnera- Decision-Making
bility associated with the impact of a determined
hazard. The data required for generating vulner- The evaluation of the “direct costs” and “factors”
ability maps are very complex and varied, and (indirect costs) that have an impact on the eco-
depend on the observation scale. Vulnerability is nomic growth of an area affected by a volcanic
directly dependent on the type of phenomena in event needs to take into account a number of
question and on the socio-economic characteris- elements. A cost-benefit analysis may assist the
tics of the environment. VOLCANDAM is a new decision-making process by evaluating the eco-
e-tool based on the methodology developed by nomic impact of the different scenarios. The
Scaini et al. (2014) that generates maps approach 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
58 S. Bartolini et al.

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

is a Bayesian decision model that applies a gen-

eral, flexible, probabilistic approach to the man- Discussion and Conclusions
agement of volcanic crises by combining the
hazard and risk factors that decision-makers need Modern volcanology is a scientific discipline
for a holistic analysis of a volcanic crisis. These with social and practical applications that derive
factors include eruption scenarios and their from its direct involvement in risk reduction.
probabilities of occurrence, the vulnerability of Today, a multidisciplinary approach is required
populations and their activities, and the costs of to risk assessment since new methodologies are
false alarms and erroneous forecasts. This model constantly being developed and explored, and,
can be implemented before an emergency to for example, geo-spatial technologies are
(i) pinpoint actions for reducing the vulnerability becoming extremely helpful in the disaster-risk
of a particular area, (ii) identify the optimum management. A defined and solid methodology
mitigating actions during an emergency and how for assessing volcanic hazard and managing risk
these may change as new information is obtained, is fundamental in both long-term evaluations and
(iii) assess after an emergency the effectiveness of in unrest phases when monitoring information is
a mitigating response and, in light of results, available. Furthermore, e-tools can be progres-
(iv) how to improve strategies before another sively modified and implemented in light of the
crisis occurs. BADEMO (BAyesian DEcision outcomes obtained in order to integrate as many
MOdel) is part of this integrated approach, and models as possible into the assessment and
enables the previous analysis of the distribution management of volcanic risk.
of local susceptibility and vulnerability to erup- The possibility of foreseeing an event such as
tions to be combined with specific costs and a volcanic eruption is more effective—and
potential losses. Indeed, BADEMO should be guarantees greater economic savings—than act-
seen as a tool for improving communication ing after a disaster. The potential of simple
between the monitoring scientists who provide e-tools such as those presented in this chapter lies
volcanological information and those responsible in the fact that they are freely available and have
for deciding which action plans and mitigating been developed on accessible and dynamic
strategies should be put into practice. graphical user interfaces. The main advantage of
Probabilistic E-tools for Hazard Assessment … 59

using these e-tools is that new data or new model conducted previous long-term hazard assessment
results can be easily incorporated into the pro- —among many other tasks—to properly manage
cedures for updating the hazard assessment. By a volcanic crisis. They allow scientists and
contrast, the use of expensive commercial e-tools managers to understand the characteristics of the
hampers the exchange of information and com- volcano or volcanic area and its past eruptive
plicates their testing in situ in volcanic fields. history, and to infer the possible eruptive sce-
Furthermore, non-free tools may hinder effective narios that may occur in the future. With this
risk assessment since they are often beyond the previous information, short-term hazard assess-
means of advisory and management groups with ment can be conducted when volcanic unrest
limited financial, technological, and manpower starts and hazard maps can be drawn up and
resources (Leidig and Teeuw 2015). However, alert levels be defined; nevertheless, it is
this does not mean that anybody can use the important to always bear in mind that the best
e-tools discussed here for, on the contrary, they form of protection is the evacuation of the
require expertise in volcanic hazards and related population at risk. Volcanic monitoring is an
issues and their use implies scientific knowledge essential part of short-term assessment and so
that will avoid incorrect outcomes. should be performed by experts and based on a
In this chapter we have presented different good understanding of volcanic processes.
statistical methodologies and e-tools for inter- Despite the possibility of conducting cost-benefit
preting volcanic data and assessing long- and analysis, which can help maintain the economic
short-term volcanic hazards and vulnerability, order, the security and health of the population
and for carrying out cost-benefit analyses. Sta- should always be the main concern.
tistical analysis enables us to extract information One of the purposes in the near future is to
about the future behaviour of a volcano by create a new software platform (VolcanBox, see
looking at the geological and historical activity VETOOLS European Project—www.vetools.eu)
of the volcanic system (VERDI database, Bar- with a user-friendly interface. This platform will
tolini et al. 2014c). The QVAST tool (Bartolini contain different e-tools that, via a homogeneous
et al. 2013) can be used to analyse past activity and systematic methodology, will help minimise
and to calculate the possibility that new vents risk. However, the feasibility and applicability of
will open (volcanic susceptibility), while the each tool will have to be analysed by different
Bayesian event tree statistical method HASSET groups of experts with experience in regions
(Sobradelo et al. 2014) can be applied to calcu- possessing different volcanological and
late eruption recurrence. Using these calcula- socio-economic scenarios. This evaluation must
tions, we can identify a number of significant also bear in mind potential end users (i.e. Civil
scenarios using GIS-based e-tools (i.e. VORIS Protection agencies), not only to test the ability
2.0.1, HAZMAP, …) and evaluate the potential of existing tools but also to understand
extent of the main volcanic hazards expected to decision-makers’ needs and requirements, and
occur in volcanic areas. The results obtained train them in the use of these tools. If we are to
allow us to generate volcanic hazard maps for reduce volcanic risk we must ensure that scien-
different levels of hazards, evaluate vulnerability tists, managers, and decision-makers are all fully
(VOLCANDAM e-tools, Scaini et al. 2014), prepared to confront this phenomenon since the
conduct cost-benefit analysis (BADEMO e-tools, best way to guarantee risk reduction is to possess
Sobradelo et al. 2015), and, finally, manage good knowledge of its causes.
volcanic risk.
During a volcanic crisis, emergency plans Acknowledgements This work was supported by the
European Commission (FP7 Theme: ENV.2011.1.3.3-1;
must be put into practice and so different gov- Grant 282759: VUELCO). The English text was corrected
ernment departments need to be prepared in by Michael Lockwood.
advance. Therefore, it is important to have
60 S. Bartolini et al.

References and felsic eruptions on Tenerife (Canary Islands).

J Volcanol Geotherm Res 195:69–77
Marti J, Spence R, Calogero E, Ordoñez A, Felpeto A,
Aspinall WP (2006) Structured elicitation of expert Baxter P (2008) Estimating building exposure and
judgment for probabilistic hazard and risk assessment impact to volcanic hazards in Icod de los Vinos,
in volcanic eruptions. In: Mader HM et al (eds) Statis- Tenerife (Canary Islands). J Volcanol Geotherm Res
tics in volcanology. Special Publication of IAVCEI, 178:553–561
Geological Society of London Martí J, Sobradelo R, Felpeto A, García O (2012)
Bartolini S, Cappello A, Martí J, Del Negro C (2013) Eruptive scenarios of phonolitic volcanism at
QVAST: a new Quantum GIS plugin for estimating Teide-Pico Viejo volcanic complex (Tenerife, Canary
volcanic susceptibility. Nat Hazards Earth Syst Sci 13 Islands). Bull Volcanol 74:767–782. doi:10.1007/
(11):3031–3042 s00445-011-0569-6
Bartolini S, Geyer A, Martí J, Pedrazzi D, Aguirre-Díaz G Martí J, Pinel V, López C, Geyer A, Abella R, Tárraga M,
(2014a) Volcanic hazard on deception Island (South Blanco MJ, Castro A, Rodríguez C (2013) Causes and
Shetland Islands. J Volcanol Geotherm Res, Antarc- mechanisms of the 2011–2012 El Hierro (Canary
tica). doi:10.1016/j.jvolgeores.2014.08.009 Islands) submarine eruption. J Geophys Res Solid
Bartolini S, Bolós X, Martí J, Riera E, Planagumà LL Earth 118(3):823–839
(2014b) Hazard assessment at the Quaternary La Marzocchi W, Sandri L, Gasparini P, Newhall C,
Garrotxa Volcanic Field (NE Iberia). Nat Hazards. Boschi E (2004) Quantifying probabilities of volcanic
doi:10.1007/s11069-015-1774-y events: the example of volcanic hazard at Mount
Bartolini S, Becerril L, Martí J (2014c) VERDI: a new Vesuvius. J Geophys Res 109. doi:10.1029/
volcanic management risk database design. J Volcanol 2004JB003155
Geotherm Res 288:132–143 Neri A, Aspinall W, Cioni R, Bertagnini A, Baxter P,
Becerril L, Bartolini S, Sobradelo R, Martí J, Morales JM, Zuccaro G, Andronico D, Barsotti S, Cole P,
Galindo I (2014) Long-term volcanic hazard assess- Ongaro TE, Hincks T, Macedonio G, Papale P,
ment on El Hierro (Canary Islands). Nat Hazards Earth Rosi M, Santacroce R, Woo G (2008) Developing
Syst Sci 14:1853–1870 an event tree for probabilistic hazard and risk assess-
Blong R (2000) Volcanic hazards and risk management. ment at Vesuvius. J Volcanol Geotherm Res 178
In: Sigurdsson H et al (eds) Encyclopedia of volca- (3):397–415
noes. Academic, San Diego, pp 1215–1227 Newhall CG, Hoblitt RP (2002) Constructing event trees
Cappello A, Neri M, Acocella V, Gallo G, Vicari A, Del for volcanic crisis. Bull Volcanol 64:3–20
Negro C (2012) Spatial vent opening probability map Sandri L, Jolly G, Lindsay J, Howe T, Marzocchi W
of Etna volcano (Sicily, Italy). Bull Volcanol (2012) Combining long-and short-term probabilistic
74:2083–2094. doi:10.1007/s00445-012-0647-4 volcanic hazard assessment with cost-benefit analysis
Cappello A, Bilotta G, Neri M, Del Negro C (2013) to support decision making in a volcanic crisis from
Probabilistic modelling of future volcanic eruptions at the Auckland Volcanic Field, New Zealand. Bull
Mount Etna. J Geophys Res Solid Earth 118:1925– Volcanol 74(3):705–723
1935. doi:10.1002/jgrb.50190 Sandri L, Thouret JC, Constantinescu R, Biass S,
Chester DK, Dibben CJL, Duncan AM (2002) Volcanic Tonini R (2014) Long-term multi-hazard assessment
hazard assessment in Western Europe. J Volcanol for El Misti volcano (Peru). Bull Volcanol 76(2):1–26
Geotherm Res 115:411–435 Scaini C, Felpeto A, Martí J, Carniel R (2014) A
De La Cruz-Reyna S (1996) Long term probabilistic GIS-based methodology for the estimation of potential
analysis of future explosive eruptions. In: Scarpa R, volcanic damage and its application to Tenerife Island,
Tilling RI (eds) Monitoring and mitigation of volcanic Spain. J Volcanol Geotherm Res 278–279:40–58
hazards. Springer, Berlin, pp 599–629 (IAVCEI/ Schilling SP (1998) LAHARZ: GIS program for auto-
UNESCO Volume) mated mapping of lahar-inundation hazard zones. US
Felpeto A, Martí J, Ortiz R (2007) Automatic GIS-based Geological Survey Open-File Report, pp 98–638
system for volcanic hazard assessment. J Volcanol Sheridan MF, Malin MC (1983) Application of
Geotherm Res 166:106116 computer-assisted mapping to volcanic hazard evalu-
Leidig M, Teeuw R (2015) Free software: a review, in the ation of surge eruptions: Vulcano, Lipari. J Volcanol
context of disaster management. Inter J Appl Earth Geotherm Res 17:187–202
Observ Geoinfo 42:49–56 Sobradelo R, Martí J (2010) Bayesian event tree for
Lirer L, Petrosino P, Alberico I (2001) Volcanic hazard long-term volcanic hazard assessment: application to
assessment at volcanic fields: the Campi Flegrei case Teide-Pico Viejo stratovolcanoes, Tenerife, Canary
history. J Volcanol Geotherm Res 101(1–4):55–75 islands. J Geophys Res 115. doi:10.1029/
Macedonio G, Costa A, Longo A (2005) A computer 2009JB006566
model for volcanic ash fallout and assessment of Sobradelo R, Bartolini S, Martí J (2014) HASSET: a
susbsequent hazard. Comput Geosci 31:837–845 probability event tree tool to valuate future volcanic
Martí J, Felpeto A (2010) Methodology for the compu- scenarios using Bayesian inference presented as a
tation of volcanic susceptibility. An example for mafic plugin for QGIS. Bull Volcanol 76:770
Probabilistic E-tools for Hazard Assessment … 61

Sobradelo R, Martí J, Kilburn C, López C (2015) Twigg J (2004) Disaster risk reduction: mitigation and
Probabilistic approach to decision-making under preparedness in development and emergency pro-
uncertainty during volcanic crises: retrospective appli- gramming. Overseas Development Institute, Human-
cation to the El Hierro (Spain) 2011 volcanic crisis. itarian Practice Network, London, Good Practice
Nat Hazards 76(2):979–998 Review no. 9
Thierry P, Neri M, Le Cozannet G, Jousset P, Costa A Van Westen CJ (2013) Remote sensing and GIS for
(2015) Preface: approaches and methods to improve natural hazards assessment and disaster risk manage-
risk management in volcanic areas. Nat Hazards Earth ment. In: Shroder J (editor in Chief), Bishop MP
Syst Sci 15:197–201 (ed) Treatise on geomorphology. vol 3, Remote
Toyos GP, Cole PD, Felpeto A, Martí J (2007) A Sensing and GIScience in Geomorphology. Academic
GIS-based methodology for hazard mapping of small Press, San Diego, pp 259–298
pyroclastic density currents. Nat Hazards 41:99–112

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 The Need to Quantify Hazard Related
to Non-magmatic Unrest: From
BET_EF to BET_UNREST

Laura Sandri, Roberto Tonini, Dmitri Rouwet,

Robert Constantinescu, Ana Teresa Mendoza-Rosas,
Daniel Andrade and Benjamin Bernard

Abstract
Most volcanic hazard studies focus on magmatic eruptions and their
accompanying phenomena. However, hazardous volcanic events can also
occur during non-magmatic unrest, defined as a state of volcanic unrest in
which no migration of magma is recognised. Examples include tectonic
unrest, and hydrothermal unrest that may lead to phreatic eruptions.
Recent events (e.g. Ontake eruption, September 2014) have demonstrated
that the successful forecasting of phreatic eruptions is still very difficult. It
is therefore of paramount importance to identify indicators that define the
state of non-magmatic unrest. Often, this type of unrest is driven by
fluids-on-the-move, requiring alternative monitoring setups, beyond the
classical seismic-geodetic-geochemical architectures. Here we present a
new version of the probabilistic model BET (Bayesian Event Tree), called
BET_UNREST, specifically developed to include the forecasting of
non-magmatic unrest and related hazards. The structure of BET_UNREST
differs from the previous BET_EF (BET for Eruption Forecasting) by
adding a dedicated branch to detail non-magmatic unrest outcomes.
Probabilities are calculated at each node by merging prior models and past

L. Sandri (&)
D. Rouwet A.T. Mendoza-Rosas

Sezione di Bologna, Istituto Nazionale di Geofisica e Instituto de Geofísica, Universidad Nacional
Vulcanologia, Via Donato Creti 12, Bologna, Italy Autónoma de México. C. Universitaria, Coyoacán
e-mail: [email protected] 04510, México, D.F., Mexico
R. Tonini Present Address:
Sezione di Roma 1, Istituto Nazionale di Geofisica e A.T. Mendoza-Rosas
Vulcanologia, Via di Vigna Murata 605, Rome, Italy CONACYT-Centro de Ingeniería y, Desarrollo
Industrial, Av. Playa pie de la cuesta 702, Desarrollo
R. Constantinescu
San Pablo, Querétaro, Qro. 76125, Mexico, C.P.,
Seismic Research Centre, University of West Indies,
Mexico
Gordon street, Saint Augustine, Trinidad & Tobago
D. Andrade
B. Bernard
Present Address:
Escuela Politécnica Nacional, Instituto Geofísico,
R. Constantinescu
Quito, Ecuador
School of Geosciences, University of South Florida,
4202 E. Fowler Avenue, NES 107, Tampa, FL
33620-5550, USA

Advs in Volcanology (2019) 63–82

DOI 10.1007/11157_2017_9
© The Author(s) 2017
Published Online: 12 August 2017
64 L. Sandri et al.

data with new incoming monitoring data, and the results can be updated
any time new data has been collected. Monitoring data are weighted
through pre-defined thresholds of anomaly, as in BET_EF. The
BET_UNREST model is introduced here, together with its software
implementation PyBetUnrest, with the aim of creating a user-friendly,
open-access, and straightforward tool to support short-term volcanic
forecasting (already available on the VHub platform). The BET_UNREST
model and PyBetUnrest tool are tested through three case studies in the
frame of the EU VUELCO project.

Resumen extendido
La mayoría de los estudios sobre amenazas volcánicas están enfocados en
las erupciones magmáticas y fenómenos relacionados. Sin embargo,
fenómenos volcánicos peligrosos pueden también ocurrir durante una fase
de “unrest” no-magmático, definido por el estado de unrest volcánico en el
cual no se reconoce la migración de un magma. Ejemplos de esto son
unrest tectónico (capaz de causar preocupación independientemente del
resultado posterior) y unrest hidrotermal, que pueden resultar en
erupciones freáticas. Eventos recientes (e.g. la erupción de Ontake en
septiembre 2014) han demostrado que las erupciones freáticas siguen
siendo difícilmente previsibles. Por estas razones, es de extrema
importancia identificar señales que permitan definir un estado de unrest
no-magmático. Muchas veces, este tipo de unrest es provocado por fluidos
en movimiento, y requiere la instalación de un sistema de monitoreo
alternativo, más allá de la clásica arquitectura sismo-geodético-química.
En este capítulo, presentamos la nueva versión del modelo probabilístico
BET (Arbol de Eventos Bayesiano, por sus siglas en inglés), llamado
BET_UNREST, específicamente desarrollado para incluir la previsión de
unrest no-magmático y sus peligros relacionados. La estructura de
BET_UNREST difiere de la versión anterior BET_EF (BET para
Previsión de Erupciones, por sus siglas en inglés), añadiendo una rama
dedicada para detallar los resultados potenciales de unrest no-magmático.
Las probabilidades están calculadas para cada nodo juntando modelos a
priori y datos pasados con los datos nuevos, provenientes del monitoreo.
Los datos de monitoreo están ponderados mediante umbrales predefinidos
de anomalía, como es el caso en BET_EF. Este capítulo ilustra el modelo,
y su herramienta, con tres casos de estudio, en el marco del proyecto EU
VUELCO:

(i) un análisis retrospectivo para el volcán Popocatépetl, en donde no

hay necesidad de la rama hidrotermal, debido al carácter mag-
mático; Popocatépetl ha permanecido en estado de unrest desde
diciembre 1994 hasta el presente. Para esta aplicación,
BET_UNREST fue corrido usando la base de datos de la UNAM
(1997–2012), con una aplicación retrospectiva para prever
The Need to Quantify Hazard Related to Non-magmatic Unrest … 65

erupciones mayores (columnas eruptivas >8 km) durante el período

abril-junio 2013.
(ii) una aplicación basada en un ejercicio de simulacro en el Cotopaxi; en
este caso se probó con BET_UNREST de manera retrospectiva, pero,
esta vez, usando datos creados específicamente para el simulacro, junto
a datos de entrada basados en la historia real del volcán preparados
antes del simulacro. Presentamos la previsión de erupciones magmáti-
cas resultantes del simulacro mismo.
(iii) el simulacro en tiempo casi-real organizado en el marco de VUELCO
en Dominica (mayo 2015). El sistema volcánico de Dominica es el
prototipo para BET_UNREST debido a su carácter hidrotermal.
Actividad freática/freatomagmática ocurrió realmente durante el simu-
lacro, lo cual de hecho era bastante probable según BET_UNREST (la
probabilidad media de unrest hidrotermal fue de 0.73, mientras la
probabilidad media de una erupción hidrotermal fue de 0.32). También
se produjó un mapa de probabilidades para la apertura de ventos
eruptivos en caso de erupciones magmáticas y freáticas.

Con estos ejercicios, estamos convencidos de haber llevado BET un paso más
cerca hacia una implementación completa en situaciones de crisis. Al final,
BET_UNREST funcionó como se esperaba. Sin embargo, es importante ser
consiente de algunos puntos críticos que han resultado de estas aplicaciones,
incluso realizar más pruebas para mejorar su diseño y comprobar su utilidad
en casos reales en el futuro. BET_UNREST se introdujo junto a su imple-
mentación digital PyBetUnrest con el objetivo de crear un instrumento de fácil
uso, libre y de acceso directo (disponible en el sitio web Vhub) para ayudar en
la evaluación de la amenaza volcánica a corto plazo.

Keywords
Volcanic unrest
Forecasting
Hydrothermal
Magmatic
Bayesian
inference

Palabras clave
Unrest volcánico
Previsión
Hidrotermal
Magmático
Inferencia
Bayesiana

change in the volcanic system with respect to its

1 Introduction background behaviour, leading to cause for
concern. Unrest can be due to several factors and
Monitoring activities represent the main source depends on the local characteristics of each vol-
of information to understand the behaviour of canic system, making it very difficult to find
volcanic systems on short time-scales and, pos- general features or patterns (Phillipson et al.
sibly, during emergency crises. In this frame- 2013). Unrest may be followed by volcanic
work, one of the main challenges of volcano eruptions due to the movement of magma, but
monitoring is the identification and characteri- can also be associated with other dangerous
sation of the phase defined as “unrest”, which phenomena: indeed, in addition to
consists of a relevant physical or chemical magma-related hazards (e.g., tephra fallout, lava
66 L. Sandri et al.

flows, ballistics), hydrothermal and tectonic In the last decade it has become crucial to
activities, without evidence for provide forecasts of the possible outcomes of
“magma-on-the-move”, can also lead to danger- volcanic unrest, to give quantitative support and
ous outcomes (i.e., flank collapses, gas emis- scientific advice to decision makers (e.g., Woo
sions, phreatic explosions, lahars). Such 2008; Marzocchi and Woo 2007, 2009). Because
hazardous events related to non-magmatic unrest of this, event tree schemes have been proposed
are not easy to track and, in volcanic hazard (e.g., Newhall and Hoblitt 2002; Marzocchi et al.
evaluations, are sometimes underestimated 2004), and a few probabilistic tools based on
(Rouwet et al. 2014). For instance, many volca- event trees and Bayesian inference have been
noes pass through a phase of hydrothermal unrest developed (e.g., BET_EF, Marzocchi et al. 2008;
for years, decades or even centuries. Due to this HASSET, Sobradelo et al. 2013) with the ability
long-term behavioural similarity, it is often dif- to quantify the probability of different possible
ficult to recognise how hydrothermal unrest can outcomes related to magmatic unrest. However,
lead to related hazards in the short-term. Where the need for recognising and tracking the evo-
the driving agent and the main eruptive product lution of any type of volcanic unrest, and to
is not magma, but water (liquid or vapour) and quantify the probability linked to non-magmatic
occasionally liquid sulphur, or gas, this type of unrest as well, have led us, within the VUELCO
unrest can lead to non-magmatic eruptions. On project, to the development of a new probabilistic
the other hand, non-eruptive hydrothermal unrest model, able to forecast both magmatic and
can also promote volcanic hazards after pro- non-magmatic hazardous events related to vol-
longed gas emissions, acidic fluid infiltration into canic unrest: BET_UNREST. The BET_UNR-
aquifers, soils and the hydrologic network, or EST model is based on an event tree, whose
deformation induced by a rising fluid front (see structure is extended with respect to the previous
Rouwet et al. 2014). schemes such as BET_EF (see the generalisation
In this light, although most volcanic hazard from BET_EF to BET_UNREST in Fig. 1,
assessments focus only on magmatic eruptions as highlighted in red) by adding a specific branch to
potential hazard sources, hazardous events can detail the track and outcome of non-magmatic
also occur during non-magmatic unrest, which in unrest. Nonetheless, BET_UNREST adopts from
this chapter is defined as a state of volcanic BET_EF the Bayesian inferential paradigm and
unrest in which no migration of magma is the ability to account both for long-term data
recognised. Examples of non-magmatic unrest (typically from the geological record) and
include the tectonic (which causes concern short-term information from monitoring
independently on how it evolves and eventually networks.
ends), and hydrothermal unrest types; the latter In this chapter, we briefly present the
may eventually lead to phreatic eruptions. Recent BET_UNREST model and its implementation in
occurrences of phreatic eruptions (e.g. Ontake the PyBetUnrest software tool (Tonini et al.
eruption, September 2014, Japan) have demon- 2016), made with the aim of providing a
strated that they are still very hard to anticipate user-friendly, open-access, and straightforward
from classical observations based on tool to handle probabilistic forecasts and visu-
seismic-geodetic-geochemical monitoring archi- alise results, and that has already been included
tectures. For these reasons, it is of paramount on the Vhub platform (https://vhub.org/
importance to identify indicators that define the resources/betunrest). The new event tree and
state of non-magmatic unrest. Often, this type of tool are applied here as illustrative examples to
unrest is driven by “fluids-on-the-move”, the VUELCO target volcanoes Popocatépetl
requiring alternative and innovative monitoring (Mexico), Cotopaxi (Ecuador) and Dominica
setups, beyond the classical ones. (West Indies).
The Need to Quantify Hazard Related to Non-magmatic Unrest … 67

Fig. 1 The new event tree as defined for the BET_UNREST model (on top) and its visual implementation in the
software PyBetUnrest (on bottom). The red branch corresponds to the previous BET_EF model
68 L. Sandri et al.

2 BET_UNREST Model – the posterior pdf, at each node, is achieved by

and PyBetUnrest Tool statistically combining, through Bayes’ theo-
rem, a prior probability distribution (usually
As with all the previous BET models (e.g., coming from theoretical models and/or expert
BET_EF, for short- and long-term eruption judgement) and information from the avail-
forecasting, Marzocchi et al. 2008; BET_VH, for able data relevant for that node.
long-term volcanic hazard associated to any
potential hazardous phenomenon accompanying As in BET_EF, the probability ½hk  at each
an eruption, Marzocchi et al. 2010; Tonini et al. node k is actually described by a statistical
2015; BET_VHst, a model that merges the pre- mixing of two pdfs, describing respectively the
vious two, Selva et al. 2014), BET_UNREST 
fMg
“so-called” long-term ½hk  and short-term
performs probabilistic assessments in the frame fMg
½hk  regimes of the volcano as follows:
of volcanic hazard analysis, based on an event
tree scheme. The main novelty in the fMg 
fMg
BET_UNREST event tree is the introduction, ½hk  ¼ ck ½hk  þ ð1  ck Þ½hk 
with respect to the BET_EF tree, of a new branch
where ck represents the weight in the interval
(Fig. 1) for exploring and forecasting the out-
[0,1] depending on the degree of unrest (Mar-
comes of non-magmatic unrest (Rouwet et al.
zocchi et al. 2008). With such mixing,
2014). Due to the resemblance of BET_UNR-
BET_UNREST switches between the two
EST to other BET models from a methodological
“regimes”. In practice:
and computational point of view, here we will
only give a brief overview. The papers by Mar-
• When anomalies with respect to the volcano’s
zocchi et al. (2004, 2008) provide a more
background activity are not observed at time
detailed description.
t = t0, BET_UNREST relies on the so-called
BET_UNREST probabilities are evaluated by
long-term information to assign the proba-
a Bayesian inferential procedure, in order to
bilities (hereinafter also referred to as back-
quantify both the aleatory and epistemic uncer-
ground probabilities) at the various branches.
tainty characterising the impact of volcanic 
fMg
eruptions in terms of eruption forecasting and/or Such background probabilities (i.e., ½hk )
hazard assessment. Such a procedure allows are based on theoretical models and infor-
merging all the available information, such as mation from the geological and eruptive
models, a priori beliefs, past data from volcanic record of the volcano studied, or of similar
records and, when available, real-time monitor- volcanoes, and describe the long-term fre-
ing data in order to include, in principle, all the quencies of magmatic or non-magmatic
knowledge about the considered volcanic system. unrest, and subsequent outcomes at these
In general, the Bayesian inference procedure volcanic systems.
at the basis of BET_UNREST assigns a proba- • When a clear state of unrest of whatever
bility to each node, providing a framework nature is detected at t = t0 by BET_UNREST,
where: the probabilistic assignment at all the suc-
cessive nodes is based mainly on the moni-
– probabilities are expressed through a proba- toring information. In practice, monitoring
bility density function (pdf), and not as a data are transformed into subjective pdfs (i.e.,
fMg
single number, to account for a ½hk ) relative to the occurrence of magmatic
best-evaluation value (for example the mean or non-magmatic unrest and the following
of the probability density function, repre- branches. Actually, at some nodes, monitor-
senting a degree of aleatory uncertainty) and ing data are not considered as relevant (for
for a measure of the epistemic uncertainty example, in forecasting the size of an
(the dispersion of the pdf);
The Need to Quantify Hazard Related to Non-magmatic Unrest … 69

eruption, magmatic or not), and here • Magmatic eruption: probability (P(MEr)) of a

BET_UNREST continues to rely on theoret- magmatic eruption, given magmatic unrest;
ical models and long-term frequencies. the following sub-branches mirror the
• When, at time t = t0, BET_UNREST BET_EF structure, so we point the reader to
observes a “degree of unrest” (of whatever Marzocchi et al. (Marzocchi et al. 2008) for
nature) without it being completely clear, the them;
statistical mixing provides a resulting pdf • Non-magmatic unrest: this is the comple-
which accounts for both the regimes, giving mentary of the Magmatic unrest branch, so by
the short-term regime a weight equal to the definition is the probability of non-magmatic
degree of unrest, and to the long-term regime unrest, given an unrest;
its complement. • Hydrothermal unrest: probability (P(HU)) of
hydrothermal unrest, given a non-magmatic
In this way, during a phase of unrest, the past unrest;
data have less (null, in the case of complete • Tectonic unrest: this is the complementary of
unrest) importance. The short term hazard/ the Hydrothermal unrest branch, so it
eruption forecasting depends exclusively on the describes the probability (P(TU)) of a tectonic
translation of observed anomalies into pdfs unrest, given a non-magmatic unrest;
describing all the branches of the event tree. This • Hydrothermal eruption: probability (P(HEr))
is done, separately at each node, by weighting of a hydrothermal eruption, given a
monitoring data through pre-defined thresholds hydrothermal unrest;
of anomaly (Marzocchi et al. 2008) and con- • Vent of hydrothermal eruption: here we
verting the resulting “degree of anomaly” into a explore the spatial probability of vent opening
best-evaluation probability, to which a degree of in a hydrothermal eruption, given a
variance is associated (Fig. 2). This is a very hydrothermal eruption occurring; this node is
simple and intuitive procedure, in which the an extension with respect to the event tree
basic assumptions are: proposed in Rouwet et al. (2014);
• Size of hydrothermal eruption: probability of
1. the first anomaly detected is the most an explosive hydrothermal eruption, given a
informative hydrothermal eruption occurring from a
2. subsequent anomalies contribute less and less specific vent; its complementary branch is the
to the increase of the degree of anomaly effusive hydrothermal eruption.
3. strong non-linear coupling among anomalies
are neglected. In order to keep the structure of BET_UNR-
EST as simple as possible, an effort has been
made to maintain, where possible, a dichotomic
At each node, BET_UNREST evaluates the branching into complementary (i.e., exhaustive
following probabilities (see also Fig. 1) by and mutually exclusive) events. This is why the
means of Bayesian inference (we give the acro- Unrest node does not branch directly into mag-
nyms used throughout the chapter to indicate the matic, hydrothermal and tectonic, but first it
probability at each node in brackets): branches into magmatic-or-not. This allows a
simplification in the evaluation of short-term
• Unrest: probability (P(U)) of unrest in the probabilities. In particular, with this type of
time period [t0; t0 + |], given the monitoring ramification, the user defines which monitoring
observations at time t = t0; the time window | measurements (plus thresholds and weight)
is defined by the user; affect the pdf of one of the two branches; the
• Magmatic unrest: probability (P(MU)) that pdf of the complementary branch then comes
the unrest is due to “magma-on-the-move”, automatically.
given the unrest;
70 L. Sandri et al.

Fig. 2 This figure explains how monitoring measures are a weighted average (xi is the weight of the i-th parameter)
transformed into a best-evaluation probability at a given in order to obtain the total degree of anomaly (b). Then the
node of the event tree. First, a monitoring measure xi is total degree of anomaly is transformed into an average
translated in a degree of anomaly zi according to a selected probability using a predefined function, in BET_UNR-
anomaly function l(
) (a). In the above example, a EST, we use the function in (c). The parameters, weights,
measure below x1 is considered background, above x2 is and thresholds are selected by the user, possibly through
anomalous, and in between it has a certain degree of expert opinions’ elicitation. Figure modified from (Mar-
anomaly. After collecting the degree of anomaly for all zocchi and Bebbington 2012)
parameters considered at the node, we combine them using

The new BET_UNREST model is applied VUELCO Deliverable 7.3 (at http://vhub.org)
here with its software implementation PyBe- and in Tonini et al. (2016).
tUnrest presented in Tonini et al. (2016), which So far BET_UNREST and PyBetUnrest have
aims to provide an open and usable tool to bridge not yet been blindly tested in real-time during an
between the scientific community and decision actual volcanic crisis, but only retrospectively
makers, with a graphical user interface which (Tonini et al. 2016) at Kawah Ijen (Indonesia),
allows the exploration of the event tree and the for the time period 2010–2012 (after a learning
visualisation of the results (see Fig. 1). This period based on the observations from 2000 to
solution was also implemented in the VHub 2010). The term “blindly” signifies that the rules
cyber-infrastructure (http://vhub.org/resources/ of BET_UNREST (the long-term pdfs, and the
betunrest). In the present PyBetUnrest tool only monitoring parameters, thresholds and weight at
one file needs to be adapted when new moni- the different nodes) are set before the beginning
toring information is gathered. This structure of the application, on different data (the learning
makes PyBetUnrest extremely fast and dataset), and then the model is applied untouched
user-friendly during crisis situations. More on the to new data (the voting dataset), typically cov-
technical background of the BET_UNREST ering a different time period (as in the case of
model and PyBetUnrest tool can be found in the Tonini et al. 2016).
The Need to Quantify Hazard Related to Non-magmatic Unrest … 71

In the next section of this chapter, results and three eruptions with an eruption column >8 km
performances of the new model and tool will be have occurred. No Plinian eruptions have occurred
discussed and validated by analysing the unrest during the 1994-ongoing eruption cycle, and thus
crises for VUELCO target volcanoes Popocaté- none of the past Plinian eruptions have been
petl, Cotopaxi and Dominica through blind monitored. For practical purposes, we thus define a
applications of BET_UNREST. The latter two major eruption for Popocatépetl as an eruption
applications show the results of the VUELCO with an eruption column >8 km, as they are
crisis simulation exercises held in Quito recorded during the current monitoring period.
(November 2014) and Dominica (May 2015). These eruptions have caused ash fall in the
Puebla-Mexico City metropolitan area, thus hav-
ing an impact on human activity. We aim at finding
3 BET_UNREST Applications precursory signals for major eruptions (>8 km,
VEI 3) for the period 1997–2012 (the learning
3.1 Popocatépetl, Mexico: period), and test the BET_UNREST retrospec-
A Retrospective tively, using monitoring data of the volcanic
Application Based activity observed during 2013 (the voting period).
on the Popo-DataBase The time window, |, is defined as 1 month.
In Table 1 we report the activity carried out
Here we apply the BET_UNREST model to 24/7 with regards to monitoring at Popocatépetl,
Popocatépetl Volcano (Mexico), based on a cata- available as short-term information for unrest,
log of monitored parameters of the 1994-ongoing origin of unrest and eruption. However, for the
eruptive period. Popocatépetl volcano awakened time period 1994–2012, the available data (as
in December 1994, after almost 48 years of vol- listed in Mendoza-Rosas, VUELCO deliverable
canic quiescence. Since 1994, Popocatépetl vol- 5.1), are restricted mainly to seismicity (VT,
cano has been one of the most active volcanoes in tremor, number of events) and visual observa-
the world, and magmatic activity has been nearly tions (i.e. number of eruptions, column height).
constant. This fact raises the need to first redefine No real-time SO2 flux is available for our pur-
the concept of volcanic unrest for Popocatépetl, as pose, and deformation data would need further
BET_UNREST, at the Unrest node, requires processing. Regarding past data (long-term
indicative parameters to verify if the given volcano information for unrest, origin of unrest, and
is in a state of unrest, or not. In stricto sensu, eruption), there have been 13 unrest episodes,
Popocatépetl has remained at least in a state of and constant unrest since December 1994 (so, a
unrest, or even magmatic or eruptive unrest, since priori probability to be in a state of unrest for the
1994, as its common manifestations are dome next month is about 85%). Out of the 13 unrest
growth and vulcanian eruptive phases. The con- episodes, 6 were due to magma-on-the-move
tinuous state of unrest is reflected by the decision (magmatic unrest), of which 3 lead into a mag-
to never decrease the level of alert from orange to matic major eruption. The monitoring parameters
green (traffic light, De la Cruz-Reyna and Tilling listed in Table 2, along with respective thresh-
2008). Nevertheless, many of these eruptions are olds and weight, have been identified in the
of no cause of concern (so, no unrest in lato sensu), UNAM (Universidad Nacional Autónoma de
neither for volcanologists nor for population. On México) database for the period 1997–2012, and
the other hand, a practical scope of the used to set BET_UNREST for Popocatépetl. The
BET_UNREST application at Popocatépetl is to volcano is a stratocone with a higher probability
forecast major eruptions, which can be considered of an eruption to occur from the central vent. For
a deviation from its current background activity. the period of observation (1997–2012) all erup-
During the past 23,000 years, nine Plinian erup- tions were magmatic and occurred at the central
tions occurred at Popocatépetl (Mendoza-Rosas crater. The a priori spatial distribution of vent
and De la Cruz-Reyna 2008), while, since 1994, opening is assigned as in Table 3. As a prior
72 L. Sandri et al.

Table 1 Activity carried Observations 4 cameras for visual observations

out 24/7 as regards
monitoring at Popocatépetl 5 three-component seismic stations
5 BB seismic stations
1 video camera + microwaves
1 doppler radar
3 biaxial inclinometers
Geochemical observations (3 sites)
Routine actions Automatic alarm for anomalies in seismicity
Cell phone messages to personnel
Comité Técnico Cientifico Asesor UNAM/CENAPRED
Reports by SMS to population

model to define the size/style of magmatic data available. Results of P(MEr) for the retro-
eruptions we take the power law from Simkin spective application period (weekly updated) are
and Siebert (1994). As past data we take the presented in Fig. 3. For the whole period, P
Mendoza-Rosas and De la Cruz-Reyna (2008) (MEr) of a major eruption (>8 km eruption col-
catalog for the past 23,000 years, and assume it umn) was <1% per month.
to be complete for VEI >= 2 (Table 3).
We retrospectively applied BET_UNREST
for the voting period April–June 2013, in which 3.2 Cotopaxi, Ecuador: Retrospective
respectively 10, 11 and 2 eruptions of 2, 3 and Application Inspired
4 km-high columns were observed. No major by the VUELCO Simulation
eruption occurred. Observed anomalies include Exercise in Quito
ash eruptions up to 130/day (all with columns
<4 km), seismic tremor, incandescence in the A volcanic unrest simulation exercise for Coto-
crater/dome, and VT events (but no shallow paxi volcano (5897 m.a.s.l.) was performed on
event with depth <5 km). There was no anoma- November 13th, 2014 in Quito, Ecuador. The
lous deformation, no dome growth, and no SO2 ice-capped stratovolcano, with an andesitic to

Table 2 Monitoring parameters set for BET_UNREST at Popocatépetl

Parameters
Unrest – # exhalations with ash (<4 km) > 20/day
– Tremor Y/N
– Increase VT Y/N
Magmatic unrest – Incandescence dome Y/N, weight 2
– Duration tremor >6000 s, weight 1
– SO2 >2000 t/d, weight 1
Magmatic eruption – Dome growth Y/N
– SO2 >9000 t/d
– Tectonic EQ > M5.5 (along the coast/arc Michoacán-Chiapas) Y/N
– Incandescent debris Y/N
– Change in # tremor Y/N
– VT depth <5 km
– Increase # VT >M2 Y/N
– Duration tremor >30,000 s (inertia 2 months) Y/N
– Increase # ash eruptions >2000–4000 (inertia 2 months)
– Deformation Y/N
The Need to Quantify Hazard Related to Non-magmatic Unrest … 73

Table 3 Left Part: Spatial probability of vent opening for magmatic eruptions assigned for BET_UNREST at
Popocatépetl: best guess a priori values. No past data are used. Right Part: Parameters of the magmatic eruption size
distribution assigned for BET_UNREST at Popocatépetl: best guess a priori values and past data
Spatial probability of vent opening in magmatic Size of magmatic eruption
eruptions
Vent location A priori probability (best guess values; Size A priori (best guess values; Past data
equivalent number of data = 1) equivalent number of data = 1)
Central vent 0.99 VEI 1 0.83 975
North flank 0.0025 VEI 2 0.14 13
East flank 0.0025 VEI 3 0.023 3
South flank 0.0025 VEI 4 0.0038 7
West flank 0.0025 VEI  5 0.0008 2

Fig. 3 Time history of probability (expressed in percentage) to have a magmatic eruption in the retrospective analysis
at Popocatépetl

rhyolitic composition, is one of the most active A still unstable version of PyBetUnrest was
and hazardous volcanoes in Ecuador. Historic set up (along with parameters and thresholds at
eruptions at Cotopaxi produced large lithic-rich each node for Cotopaxi volcano derived from
pyroclastic flows, ash flows, lava flows as well as monitoring information) before the simulation
large lahars (Barberi et al. 1995; Hall and Mothes exercise, based on the available data in the lit-
2008; Biass and Bonadonna 2011). Some lahars erature up to the beginning of the simulation (the
reached the Pacific Ocean at >200 km distance learning period stopped with the beginning of the
(Aguilera et al. 2004; Pistolesi et al. 2013). exercise), in order to preliminarily test its value
Recent unrest periods at Cotopaxi occurred in in decision support by providing near-real time
1975–1976 and 2001–2002 and were charac- probabilities of (i) the occurrence of unrest,
terised by increased fumarolic activity, elevated (ii) the origin and nature of unrest and (iii) erup-
seismicity and edifice deformation (Molina et al. tive activity. However, during the simulation, the
2008). Fumarolic activity is a concern due to the reports from the “volcano team” did not reflect
heat transfer that may affect the ice cover the real eruptive and unrest history of Cotopaxi,
resulting in non-eruptive debris flows or lahars. as the past activity for the simulation was
74 L. Sandri et al.

“invented”. A different setting of BET_UNREST (MU) increases, along with a decrease in the
(and consequently of PyBetUnrest) on site was associated uncertainty.
not possible due to the lack of time and the still (4) Phase 3 (Bulletin 3): the observed anomalies
premature customisability of the tool. This in this phase were: an increase in VT and LP
obliged us to set up and run the old BET_EF tool events, occurrence of tremor, appearance of
during the exercise (Constantinescu et al. 2015). new fumaroles, an increase in SO2 emission,
Obviously, this prevented us from providing and an increase in the crater thermal anom-
probabilistic assessment of non-magmatic events aly. As a consequence, the P(MU) is similar
during the exercise at Cotopaxi: this would have to Bulletin 2, but the P(HU) increases
been possible with BET_UNREST, enabling the slightly, due to the new fumaroles.
calculation of probabilities for hydrothermal (5) Phases 4 and 5 (Bulletins 4 and 5): the
unrest and hydrothermal eruptions (P(HU) and P observed anomalies in these phases were
(HEr)). Nevertheless, the unrest scenario pro- similar, and included: intense fumarolic
posed by the “volcano team” (Bulletins 1–5) did activity, occurrence of hybrid seismic events,
not emphasise a significant state of hydrothermal an increase in SO2 emission, and an increase in
unrest, which, on the one hand, made our output the crater thermal anomaly. As a consequence,
less biased in not providing an evaluation for P P(MEr) increases from 0.21 (phase 3) to 0.57,
(HU) and P(HEr); but on the other hand this combined with a lower uncertainty.
simulation was probably not the best case to test
BET_UNREST.
Here, we will re-run BET_UNREST and
PyBetUnrest at Cotopaxi retrospectively for the
unrest phases described in the five bulletins 3.3 Dominica, West Indies, Lesser
provided by the “volcano team” during the sim- Antilles: VUELCO
ulation exercise and using the BET_UNREST Simulation Exercise,
setup prepared prior to the simulation based on Dominica, May 2015
the real past activity of the volcano (Table 4).
The time window | was set to 1 month. In Dominica is characterised by hydrothermal
Table 5 we show the probabilities resulting from activity manifested as thermal springs (up to
the run of the code, after each bulletin: boiling temperature), boiling-temperature
fumarolic emissions (e.g. Valley of Desolation)
(1) Phase 0: The background activity of Coto- and a crater lake, known as ‘Boiling Lake’, with
paxi (NO anomalies): results are based on a particular hydrodynamic behaviour (Fournier
the past activity of Cotopaxi, with all et al. 2009; Joseph et al. 2011; Rouwet et al.
observation within background limits. 2017). No high-temperature manifestations occur
(2) Phase 1 (Bulletin 1): the observed anomalies on the island, so no clear evidence of active
in this phase were limited to an increase in magmatic degassing exists at the present time.
seismic activity compared to background The simulation exercise, and consequently the
level. Such an increase is indicative, accord- BET_UNREST application, for the VUELCO
ing to pre-set parameters, of magma-on-the- target island of Dominica mainly focused on an
move (P(MU) = 0.68). The considerable unrest scenario for the southern part of the island.
uncertainty is summarised by the 10th to 90th The purpose of the exercise was to test the
percentiles confidence interval. tracking/assessment of an unrest period, and the
(3) Phase 2 (Bulletin 2): the observed anomalies decision making process undertaken by the sci-
in this phase were: a drastic increase in entific advisory group and local authorities.
seismicity, an increase in SO2 emission (5 Due to the hydrothermal character of
times background levels), and a crater ther- Dominica, the application of BET_UNREST is
mal anomaly. As a consequence, the mean P highly suited. Before the simulation exercise, the
The Need to Quantify Hazard Related to Non-magmatic Unrest … 75

Table 4 Monitoring parameters set for BET_UNREST at Cotopaxi

Node-parameter# Parameter and threshold(s) (Y/N indicates a Boolean observation)
Unrest-parameter 1 LP/month (205–335) (Garcia-Aristazabal 2010)
Unrest-parameter 2 VT/month (24–32) (Garcia-Aristazabal 2010)
Unrest-parameter 3 M Tectonic EQ (3–4)
Unrest-parameter 4 SO2 (Y/N)
Magmatic unrest-parameter 1 EQ depth (>4.5–5.5 km)
Magmatic unrest-parameter 2 Deep VLP (Y/N)
Magmatic unrest-parameter 3 T fumarole (>119 °C)
Magmatic unrest-parameter 4 Appearance of acidic gas (Y/N)
Magmatic unrest-parameter 5 VT/month (>32)
Magmatic unrest-parameter 6 Increased deformation (Y/N)
Magmatic unrest-parameter 7 VLP + LP together (Y/N)
Magmatic unrest-parameter 8 Harmonic LP tremor (Y/N)
Magmatic unrest-parameter 9 SO2 flux (t/d) (>100–350)
Magmatic eruption-parameter 1 sudden stop (Y/N)
Magmatic eruption-parameter 2 SO2 flux (t/d) (>2000–2500)
Magmatic eruption-parameter 3 Tornillos (Y/N)
Hydrothermal unrest-parameter 1 New fumarole (Y/N)
Hydrothermal unrest-parameter 2 Anomalous glacier volume decrease (defrosting) (Y/N)
Hydrothermal unrest-parameter 3 LP/month (>205–335) (Garcia-Aristazabal 2010)
Hydrothermal eruption-parameter 1 Increase in T of fumarole (>120–200 °C)
Hydrothermal eruption-parameter 2 Increase in extension of fumarolic field (Y/N)
Hydrothermal eruption-parameter 3 Inflation of fumarolic field (Y/N)
Hydrothermal eruption-parameter 4 Landslides in hydrothermal areas (Y/N)
Hydrothermal eruption-parameter 5 New/extension of alteration areas (Y/N)

PyBetUnrest tool was set for Dominica, based on is listed in Table 7 (parameters and thresholds
(1) existing literature of the past volcanic activ- identified prior to the exercise onset, see above).
ity; (2) insights on the current hydrothermal Further details on the Dominica simulation
activity; (3) discussion-based expert elicitation exercise and on the BET_UNREST application
sessions (4 sessions at SRC and 1 at are given in Constantinescu et al. (2016).
INGV-Bologna); and (4) exchanges with local During the simulation exercise (May 14–15,
experts in order to fine-tune the code with the 2015) three phases of changes in volcanic
monitoring parameters. We remark that all of this activity, each with a duration of six months, were
was done prior to the start of the simulation distributed by the “volcano team” to the opera-
exercise (the learning period stopped at the tors of the unrest crisis. The reports included four
beginning of the simulation exercise, as for types of observations: (1) seismic bulletin,
Cotopaxi), and again no hindsight tuning was (2) GPS, (3) geothermal monitoring data, and
made. The long-term setup of PyBetUnrest is (4) other observations.
done by filling up a configuration file that The translation of the reported bulletins into
includes the a priori and past data specifically for the values for the selected parameters in the
Dominica, whose main information is sum- BET_UNREST for Dominica setup were repor-
marised in Table 6. The short-term information ted back to the team of experts in real-time
76 L. Sandri et al.

Table 5 Resulting P(U) P(MU) P(MEr) P(HU) P(HEr)

probabilities from
retrospective application of Phase 0 Mean 0.005 0.002 0.0005 0.001 0.0006
BET_UNREST at (Background) 10th prctile 0.0013 0.0002 0 0 0
Cotopaxi
50th prctile 0.004 0.001 0.0002 0.0006 0.0002
90th prctile 0.009 0.004 0.001 0.003 0.001
Phase 1 Mean 1 0.68 0.18 0.08 0.02
10th prctile 1 0.07 0 0 0
50th prctile 1 0.84 0.02 0.001 0
90th prctile 1 1 0.69 0.30 0.04
Phase 2 Mean 1 0.83 0.22 0.05 0.013
10th prctile 1 0.27 0 0 0
50th prctile 1 1 0.04 0 0
90th prctile 1 1 0.75 0.13 0.008
Phase 3 Mean 1 0.80 0.21 0.13 0.07
10th prctile 1 0.14 0 0 0
50th prctile 1 1 0.04 0.002 0.0003
90th prctile 1 1 0.72 0.54 0.22
Phase 4 and 5 Mean 1 0.81 0.57 0.12 0.07
10th prctile 1 0.23 0.02 0 0
50th prctile 1 1 0.65 0.0004 0.0002
90th prctile 1 1 1 0.49 0.28

Table 6 Set up of BET_UNREST at Dominica in terms of long-term information

A priori mean (equivalent n data in brackets) Past data
Unrest 0.5 (1) Past data (successes) = 14
Past data (total) = 608
Magmatic 0.5 (1) Past data (successes) = 13
past data (total) = 14
Magmatic eruption 0.58 from Phillipson et al. (2013) (1) Past data (successes) = 0
Past data (total) = 13
Magmatic vent location file file
Hydrothermal vent location file file
Size distribution (Magmatic) Dome extrusion: 0.83 Dome extrusion: 0
Small explosive: 0.14 Small explosive: 5
Large explosive: 0.03 Large explosive: 2
(1)
Some of the data are too many to be listed (this is indicated by the label “file” in the table). They can be provided in the
form of files on request

during the simulation. In Table 8 we provide the probability distributions, across all the time
probabilities resulting from the run of the code periods spanned by the simulation exercise in
after each bulletin. In Fig. 4 we also provide the Dominica. For each bulletin, among the output
time evolution of some of the most relevant information from PyBetUnrest, there were two
The Need to Quantify Hazard Related to Non-magmatic Unrest … 77

Table 7 Monitoring parameters set for BET_UNREST at Dominica

Node–parameter# Parameter and threshold(s) (Y/N indicates a boolean observation)
Unrest-parameter 1 Increased CO2 flux above background (Y/N)
Unrest-parameter 2 Increase in T of hot springs and/or fumaroles (Y/N)
Unrest-parameter 3 Changes in H2O/CO2 (Y/N)
Unrest-parameter 4 Appearance of new fumaroles and/or hot springs (Y/N)
Unrest-parameter 5 Vegetation die back (Y/N)
Unrest-parameter 6 Appearance of LPs and hybrid EQs (Y/N)
Unrest-parameter 7 Large regional tectonic event (M > 7) (Y/N)
Unrest-parameter 8 Number of VTs [if >10/day for two weeks]
Unrest-parameter 9 Detectable ground deformation (Y/N)
Magmatic unrest-parameter 1 Increase in C/S, or decrease after increase (Y/N)
Magmatic unrest-parameter 2 Detectable SO2, HCl, HF (Y/N)
Magmatic unrest-parameter 3 Extreme increase in T [>300 °C]
Magmatic unrest-parameter 4 Any VLPs (Y/N)
Magmatic unrest-parameter 5 No. of LPs after significant VT swarms (#/day) (>5–10)
Magmatic unrest-parameter 6 Consistent increase in No. of VTs for 1 month (Y/N)
Magmatic unrest-parameter 7 Deep VTs [>8 km] (#/week) (4–5)
Magmatic unrest-parameter 8 Detectable radial deformation (localized-coherent signal) (Y/N)
Magmatic unrest-parameter 9 Surface deformation (island wide, >6 cm in over 6 months) (Y/N)
Magmatic eruption-parameter 1 Decreasing C/S after increase (Y/N)
Magmatic eruption-parameter 2 Increase in Cl, Br, F content in hot springs/pools (Y/N)
Magmatic eruption-parameter 3 Decrease in H2O/CO2 and/or H2S/SO2 and/or SO2/HCl (Y/N)
Magmatic eruption-parameter 4 Phreatic activity (Y/N)
Magmatic eruption-parameter 5 Large thermal anomaly [incandescense] (Y/N)
Magmatic eruption-parameter 6 Landslides in hydrothermal areas (Y/N)
Magmatic eruption-parameter 7 Acceleration of VTs, LPs, hybrids [weekly] (Y/N)
Magmatic eruption-parameter 8 Presence of harmonic tremor (Y/N)
Magmatic eruption-parameter 9 Shallowing of VTs hypocenters in the ediffice or shallow depths [<3 km]
(Y/N)
Magmatic eruption-parameter 10 Sudden reversal of activity (Y/N)
Hydrothermal unrest-parameter 1 Anomalous behavior of Boiling Lake [overflow, lower or higher T than
usual, no return of lake, etc.] (Y/N)
Hydrothermal unrest-parameter 2 Changes in hydrothermal features (Y/N)
Hydrothermal unrest-parameter 3 Increase in B and/or NH4 concentration in waters (Y/N)
Hydrothermal unrest-parameter 4 Increase in CH4/CO2 (fumaroles) (Y/N)
Hydrothermal unrest-parameter 5 Increase in T of fumaroles (Y/N)
Hydrothermal eruption-parameter 1 Increase in T of fumaroles (fuzzy 120–200 °C)
Hydrothermal eruption-parameter 2 riSe of water level in pools/overflow of BL (Y/N)
Hydrothermal eruption-parameter 3 Increase in extension of fumarolic field (Y/N)
Hydrothermal eruption-parameter 4 Muddy pools (Y/N)
(continued)
78 L. Sandri et al.

Table 7 (continued)
Node–parameter# Parameter and threshold(s) (Y/N indicates a boolean observation)
Hydrothermal eruption-parameter 5 Boiling/bubbling of pools that previously didn’t (Y/N)
Hydrothermal eruption-parameter 6 Inflation of fumarolic field (Y/N)
Hydrothermal eruption-parameter 7 Landslides in hydrothermal areas (Y/N)
Hydrothermal eruption-parameter 8 New/extension of alteration areas (Y/N)

Table 8 Resulting P(U) P(MU) P(MEr) P(HU) P(HEr) P(TU)

probabilities from real-time
application of Phase 1 mean 1 0.26 0.06 0.62 0.42 0.12
BET_UNREST at 10th prctile 1 0 0 0.05 0.01 0
Dominica during VUELCO
50th prctile 1 0.06 0 0.73 0.32 0
simulation exercise
90th prctile 1 0.85 0.22 1 0.95 0.5
Phase 2 mean 1 0.82 0.53 0.13 0.03 0.05
10th prctile 1 0.29 0.01 0 0 0
50th prctile 1 1 0.56 0.001 0 0
90th prctile 1 1 1 0.54 0.06 0.06
Phase 3 mean 1 0.70 (0.24) 0.17 (0.07) 0.08 0.02 0.22
10th prctile 1 0.09 0 0 0 0
50th prctile 1 0.87 0.02 0 0 0.08
90th prctile 1 1 0.68 0.27 0.03 0.80
In bracket estimates of mean values without including HCl anomaly in Phase 3

maps of the spatial probability of vent opening:

one for the case of magmatic eruption, and one 4 Discussion and Implications
for hydrothermal eruption (Fig. 4). We believe for Unrest Tracking
this could be particularly useful, for example in a
volcanic system like Dominica, where there are This chapter presents the need for an updated
numerous areas showing hydrothermal activity, BET model and tool that is able to account for
thus increasing the uncertainty on the position of the non-magmatic nature of some volcanic unrest
a possible phreatic event. episodes, which can often go under-estimated,
The parameter “detectable SO2, HCl, HF” if not totally neglected. The new model
created confusion and opened up a scientific (BET_UNREST) and tool (PyBetUnrest) allow
discussion. For the sake of transparency, we the tracking of unrest phases at volcanic systems
provide the mean values of P(MU) and P(MEr) and enables short-term volcanic forecasts. It has
including, or not, the HCl anomaly (Table 8). been fully developed within the VUELCO pro-
Beyond the scientific implications of this issue, ject, during which time it has been applied
this concern reflected the sensitivity of to some of the project’s target volcanoes. In
BET_UNREST to the interpretation of some general, when we are able to distinguish
parameters. When relatively few monitoring magma-on-the-move (Rouwet et al. 2014) from
parameters are provided, the weight of a single the monitoring observations the new model
anomaly can be high: this is somehow a measure basically “collapses” to BET_EF (or, better, the
of the epistemic uncertainty. assessment of the probabilities related to
The Need to Quantify Hazard Related to Non-magmatic Unrest … 79

Fig. 4 Average values (top left) obtained by BET_UNR- confidence interval between 10th and 90th percentiles.
EST during the three phases of Dominica exercise for P On bottom left, a snapshot of PyBetUnrest tool shows the
(MU), P(HU), P(MEr) and P(HEr). Asterisk points are the spatial probability of vent opening during Phase 1,
alternative average values for P(MU) and P(MEr) without localising the most probable position of the phreatic
considering HCl as detectable. On the right column the eruption
same probabilities are shown together with their

magmatic outcomes provided by the two models The chapter illustrates the development and
coincide). On the other hand, if we are not able to implementation of BET_UNREST model and
identify a magmatic “active role” in the unrest PyBetUnrest tool through three different
(from the available monitoring observations), applications:
BET_UNREST is still able to provide the prob-
abilities of hazardous events that accompany (i) the pure retrospective analysis at Popoca-
non-magmatic volcanic unrest, rather than tépetl volcano, where there is no com-
neglecting them. As discussed in Rouwet et al. pelling need for a hydrothermal branch
(2014), a very difficult case is presented by due to the current magmatic nature of the
phreatomagmatic eruptions that, sometimes, can unrest episodes. Popocatépetl has
occur without any precursors indicating magma remained in unrest from December 1994
movement. This is surely an important limit to to present and, for this application,
overcome which requires further efforts to detect BET_UNREST and PyBetUnrest were run
subtle changes in the very short-term (hours to using the UNAM Data Base for the
minutes) by improving monitoring techniques. learning period 1997–2012, with a
80 L. Sandri et al.

retrospective application aiming to fore- one can always apply and “populate” the
cast major eruptions (column heights BET_UNREST model in any “volcanic” cir-
greater than 8 km) for the April–June cumstance. The uncertainty on the results pro-
2013 volcanic activity. vided by BET_UNREST, and consequently their
(ii) the application based on a simulation practical use, will however be strongly dependent
exercise at Cotopaxi. Here we tested the on the available information and data used to set
BET_UNREST retrospectively, but, this up the models rules. If only a few pieces of
time, using the invented data provided evidence are available, the models results will be
during the VUELCO simulation exercise, characterised by a large uncertainty, and thus
in addition to data based on the real past might be not very helpful for decision-makers.
history of the volcano. As more and more knowledge is gathered,
(iii) the almost real-time simulation exercise BET_UNREST output probabilities will become
organised by the VUELCO project more attractive from a practical point of view,
in Dominica (May 2015). The volcanic since their uncertainty will be increasingly small.
system of Dominica presents a “prototype” This is an intrinsic feature of the Bayesian
setting for BET_UNREST due to inferential procedure at the basis of the model.
its hydrothermal character. Phreatic/
phreatomagmatic activity occurred during
the simulation, coinciding with high asso- References
ciated probabilities from BET_UNREST
(the average values P(HU) = 0.73 and Aguilera E, Pareschi MT, Rosi M, Zanchetta G (2004)
P(HEr) = 0.32). We also positively tested Risk from Lahars in the Northern Valleys of Cotopaxi
the feasibility of providing different maps Volcano (Ecuador). Nat Hazards 33:161–189
Barberi F, Coltelli M, Frullani A, Rosi M, Almeida E
of the spatial probability of vent opening in
(1995) Chronology and dispersal characteristics of
case of magmatic or phreatic eruption. recently (last 5000 years) erupted tephra of Cotopaxi
(Ecuador): implications for long-term eruptive fore-
As mentioned in previous sections, we casting. J Volcanol Geotherm Res 69:217–239
Biass S, Bonadonna C (2011) A quantitative uncertainty
implemented the BET_UNREST model into
assessment of eruptive parameters derived from tephra
PyBetUnrest software tool using a graphical user deposits: the example of two large eruptions of
interface aiming to provide a fast, open and Cotopaxi volcano, Ecuador. Bull Volcanol 73:73–90.
user-friendly tool, which extends the usage of doi:10.1007/s00445-010-0404-5
Constantinescu R, Rouwet D, Gottsmann J, Sandri L,
BET_UNREST to volcanologists with different
Tonini R (2015) Tracking volcanic unrest at Cotopaxi,
expertise. The PyBetUnrest tool reached a Ecuador: the use of BET_EF tool during an unrest
mature and usable version during the Dominica simulation exercise. Geophys Res Abs, 17-EGU
simulation and its first stable release has been 2015–2251
Constantinescu R, Robertson R, Lindsay JM, Tonini R,
uploaded to Vhub cyber-infrastructure.
Sandri L, Rouwet D, Patrick Smith P, Stewart R
With these exercises we strongly believe we (2016) Application of the probabilistic model
have brought BET a step closer to a full and BET_UNREST during a volcanic unrest simulation
proper implementation during a crisis situation. exercise in Dominica, Lesser Antilles, Geochem
Geophys Geosyst 17:4438–4456, doi:10.1002/
The PyBetUnrest tool eventually worked as 2016GC006485
expected, but it is important to take advantage of De la Cruz-Reyna S, Tilling RI (2008) Scientific and
the lessons learned during these applications and public responses to the ongoing volcanic crisis at
pursue more tests that will improve its design and Popocatépetl Volcano, Mexico: importance of an
effective hazards-warning system. J Volcanol
prove its usefulness in real-case scenarios. Geotherm Res 170:121–134
As a final comment, we would like to remark Fournier N, Witham F, Moureau-Fournier M, Bardou L
that, as with any other event tree model (e.g. (2009) Boiling Lake of Dominica, West Indies:
BET models by Marzocchi et al. 2004, 2008, high-temperature volcanic crater lake dynamics. J Geo-
phys Res 114(B02203). doi:10.1029/2008JB005773
2010; HASSET model by Sobradelo et al. 2013),
The Need to Quantify Hazard Related to Non-magmatic Unrest … 81

Garcia-Aristazabal A (2010) Analysis of eruptive and Newhall CG, Hoblitt RP (2002) Constructing event trees
seismic sequences to improve the short- and long-term for volcanic crises. Bull Volcanol 64:3–20. doi:10.
eruption forecasting. PhD Università deli Studi di 1007/s004450100173
Bologna, pp 167 Phillipson G, Sobradelo R, Gottsmann J (2013) Global
Hall M, Mothes P (2008) The rhyolitic–andesitic eruptive volcanic unrest in the 21st century: an analysis of the
history of Cotopaxi volcano. Ecuador Bull Volcanol first decade. J Volcanol Geotherm Res 264:183–196
70(6):675–702 Pistolesi M, Cioni R, Rosi M, Cashman KV, Rossotti A,
Joseph EP, Fournier N, Lindsay JM, Fischer TP (2011) Aguilera E (2013) Evidence for lahar-triggering
Gas and water geochemistry of geothermal systems in mechanisms in complex stratigraphic sequences: the
Dominica, Lesser Antilles island arc. J Volcanol post-twelfth century eruptive activity of Cotopaxi
Geotherm Res 206:1–14. doi:10.1016/j.jvolgeores. Volcano. Ecuador Bull Volcanol 75:698. doi:10.1007/
2011.06.007 s00445-013-0698-1
Marzocchi W, Bebbington M (2012) Probabilistic erup- Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J,
tion forecasting at short and long time scales. Bull Tonini R, Papale P (2014) Recognizing and tracking
Volcanol 74:1777–1805. doi:10.1007/s00445-012- hazards related to non-magmatic unrest: a review.
0633-x J Appl Volcanol 3:17. doi:10.1186/s13617-014-0017-3
Marzocchi W, Woo G (2007) Probabilistic eruption Rouwet D, Hidalgo S, Joseph EP, González-Ilama G
forecasting and the call for an evacuation. Geophys (2017) Fluid geochemistry and volcanic unrest:
Res Lett 34:L22310. doi:10.1029/2007GL031922 dissolving the haze in time and space. In: Gottsmann J,
Marzocchi W, Woo G (2009) Principles of volcanic risk Neuberg, J, Scheu B (eds) Volcanic Unrest: from
metrics: theory and the case study of Mount Vesuvius Science to Society—IAVCEI Advances in Volcanol-
and Campi Flegrei. Italy J Geophys Res 114:B03213. ogy, Springer, Berlin
doi:10.1029/2008JB005908 Selva J, Costa A, Sandri L, Macedonio G, Marzocchi W
Marzocchi W, Sandri L, Gasparini P, Newhall CG, (2014) Probabilistic short-term volcanic hazard in
Boschi E (2004) Quantifying probabilities of volcanic phases of unrest: a case study for tephra fallout.
events: the example of volcanic hazard at Mount J Geophys Res 119:8805–8826
Vesuvius. J Geophys Res 109:B11201 doi:10.1029/ Simkin T, Siebert L (1994) Volcanoes of the world, 2nd
2004JB003155 edn. Geoscience Press for the Smithsonian Institution,
Marzocchi W, Sandri L, Selva J (2008) BET_EF: a Tucson, p 349
probabilistic tool for long- and short-term eruption Sobradelo R, Bartolini S, Marti J (2013) HASSET: a
forecasting. Bull Volcanol 70:623–632 probability event tree tool to evaluate future volcanic
Marzocchi W, Sandri L, Selva J (2010) BET_VH: a scenarios using Bayesian inference. Bull Volcanol
probabilistic tool for long-term volcanic hazard 76:770. doi:10.1007/s00445-013-0770-x
assessment. Bull Volcanol 72:705–716 Tonini R, Sandri L, Thompson MA (2015) PyBetVH: a
Mendoza-Rosas AT, De la Cruz-Reyna S (2008) A Python tool for probabilistic volcanic hazard assess-
statistical method linking geological and historical ment and for generation of Bayesian hazard curves
eruption time series for volcanic hazard estimations: and maps. Comput Geosci 79:38–46
applications to active Polygenetic volcanoes. J Vol- Tonini R, Sandri L, Rouwet D, Caudron C, Marzocchi W,
canol Geotherm Res. doi:10.1016/j.jvolgeores.2008. Suparjan (2016) A new Bayesian Event Tree tool to
04.005 track and quantify unrest and its application to Kawah
Molina I, Kumagai H, García-Aristizábal A, Nakano M, Ijen volcano. Geochem Ge-ophys Geosyst 17:2539–
Mothes P (2008) Source process of very-long-period 2555, doi:10.1002/2016GC006327
events accompanying long-period signals at Cotopaxi Woo G (2008) Probabilistic criteria for volcano evacua-
Volcano, Ecuador. J Volcanol Geotherm Res tion decisions. Nat Hazards 87–97. doi:10.1007/
176:119–133 s11069-007-9171-9
82 L. Sandri et al.

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Groundwater flow and volcanic
unrest

Alia Jasim, Brioch Hemmings, Klaus Mayer

and Bettina Scheu

Abstract explore how these conditions can be modified

Hydrology around active volcanoes is strongly by volcanic unrest and we identify feedbacks
controlled by the interaction between ground- between dynamic hydrothermal behaviour and
water, and the fluids, dissolved elements and on-going unrest. The interaction between these
heat associated with magmatic intrusion. The systems, and therefore the associated monitor-
chemical and mechanical processes associated ing signals, are the result of complex ground-
with magmatic unrest can result in observable water-volcano coupling within multi-phase
changes in the hydrothermal system. Conse- flow system in evolving lithologies. Nonethe-
quently, observations of chemical and physical less, detailed monitoring of hydrothermal and
hydrothermal variations may provide insights hydrological behaviour can provide insights
into the state of volcanic activity. Additionally, into unrest and the evolution of hazards at
the interaction between hydrological and vol- restless volcanoes.
canic systems leads to the presence of
high-temperature, pressurised, and often acidic
fluids, which add to, and intensify, the volcanic Keywords
hazard. In the following chapter we present the

Hydrothermal system Groundwater
major components of, and controls on, mag-


Fluid flow Permeability Unrest-monitoring
matic hydrothermal systems focusing on the
mutual perturbation between the groundwater
flow system and the volcanic system. We
1 Resumen

A. Jasim (&)
Brevemente resumimos nuestra comprensión de
University of Bristol, Earth Science School, los sistemas magmáticos hidrotermales y discu-
Wills Memorial Building, Queen’s Road, timos las mayores incógnitas y sus implicaciones
Clifton BS8 1RJ, UK en el vigilancia volcánica. También proveemos
K. Mayer
B. Scheu directrices adicionales para la recolección de
Department of Earth and Environmental Sciences, datos a usarse en la calibración de la variabilidad
Ludwig-Maximilians-Universität München (LMU),
Theresienstrasse 41/III, 80333 Munich, Germany
del sistema de aguas subterráneas, alrededor de
volcanes activos, como un paso crucial para
B. Hemmings
GNS Science, 1 Fairway Drive, Lower Hutt,
desacoplar las señales magmáticas de las pura-
New Zealand mente hidrotermales.

Advs in Volcanology (2019) 83–99

DOI 10.1007/11157_2018_33
© The Author(s) 2017
Published Online: 23 May 2018
84 A. Jasim et al.

La interacción entre los sistemas hidrológico y períodos de calma. Esta evolución controla la
volcánico es un elemento importante durante posible respuesta a la perturbación ter-
reactivación volcánica. Los cambios en el com- modinámica y química asociada con la iniciación
portamiento hidrológico de un volcán activo, de la reactivación volcánica. Dadas las intrin-
como la elevación del nivel del agua subterránea, cadas retroalimentaciones entre el magma, la
la descarga de manantiales, los cambios de tem- hidrología, y los cambios repentinos de los sis-
peratura y de la química, pueden ser indicadores temas involucrados, únicamente mediciones de
preliminares de evolución de la actividad volcá- alta frecuencia (de horas a semanas) de la tem-
nica. Las interacciones hidrológicas pueden tam- peratura, pH, conductividad eléctrica del agua,
bién alterar y aumentar el peligro volcánico profundidad del nivel del agua subterránea, del
existente. Las interacciones físicas y químicas contenido de REE (Elementos de Tierras Raras),
entre la roca encajante y los diferentes tipos de RFEs (Elementos de Formación de Rocas) y gas
fluido pueden modificar los caminos de desgasi- disuelto, conjuntamente con mediciones geofísi-
ficación, generando distribuciones de presión cas, pueden aclarar la evolución del sistema
dinámicas dentro del edificio volcánico. Aún los magmático, la apertura/cierre de fracturas y la
procesos lentos, como el desarrollo creciente de dinámica estacional del agua subterránea.
zonas de alteración permanentes, pueden mani-
festarse como un peligro dinámico asociado con
una reactivación continua o futura, ya que las 2 Introduction
rocas altamente cristalinas son hidrotérmicamente
alteradas produciendo arcillas débiles secun- Much of the research relating to the interaction
darias. Discutimos los principales parámetros que between hydrological and volcanic systems has
controlan las reacciones y sus efectos en la dis- focused on the role of hydrothermal systems in
tribución de la alteración en ambientes volcánicos. the development of economic mineral deposits.
Debido a la introducción del calor de la fuente hydrothermal systems have formed vast
en el sistema del agua saturada, se presentan ore-deposits around the world, most of them
peligros adicionales. Esto frecuentemente conll- clearly result from the interaction between mag-
eva a explosiones freáticas y freato-magmáticas. matic and meteoric fluids (Hedenquist and
La presencia de paquetes de gases bajo la Lowenstern 1994).
superficie además incrementa este peligro. El The interaction between hydrological and
balance entre el ingreso de agua freca fría, la volcanic systems is an important element in
desgasificación y la disipación de calor, está volcanic unrest. Changes in hydrological beha-
críticamente relacionado con la abilidad del sis- viour, such as water table elevation, spring dis-
tema para transmitir fluidos, el mismo que charge, temperature and chemistry, at an active
evoluciona en función tanto de los procesos volcano can provide early indications of changes
químicos (ej., las reacciones de dilución/precip- in volcanic activity. Hydrological interactions
itación mineral) como físicos (ej., fracturamiento can also alter and augment the existing volcanic
y compactación de la roca), produciendo así hazard. Chemical and physical interactions
propiedades hidrológicas de la roca fuertemente between host rocks and different fluid types can
dependientes de la escala (ej., porosidad, per- modify fluid degassing pathways, generating
meabilidad y conductividad térmica). dynamic pressure distributions within a volcanic
En resumen, las señales físicas y químicas, o edifice. Additional hazards are also presented by
la perturbación hidrológica asociada con la the introduction of a heat source into a water
reactivación magmática, son complejas y saturated system, this frequently results in dan-
dependientes del sitio. Las contribuciones de los gerous phreatic and phreatomagmatic explosions.
diferentes componentes del fluido, y sus inter- Understanding the controls on hydrological and
acciones con caminos de flujo existentes, pueden hydrothermal behaviour in volcanic settings is
determinar cómo un sistema evoluciona en essential for understanding the array of hazards
Groundwater flow and volcanic unrest 85

presented by volcanic unrest. Continued devel- the depth and fluctuation of the water table.
opment of this understanding is also providing A number of studies suggest a correlation
new volcano monitoring opportunities. Despite between the fluctuation of the water table and
the clear relevance and importance of hydrolog- elevated seismicity. Both the reduction in effec-
ical and volcanic interactions in relation to vol- tive stress due to a seasonal increase in hydro-
canic unrest, the dynamics of this interaction static pore pressure, and the snow unloading,
remain poorly constrained. lead to a seasonal peak of seismicity (Saar and
Manga 2003; Christiansen et al. 2005). Further-
more, Mason et al. (2004) identify seasonal
3 Hydrothermal System peaks in the eruption rate of volcanoes, which
may be due to the load/unload seasonal stress
Although insulated or distal, cool groundwater cycle imposed by the hydrological cycle.
aquifers can respond to volcanic perturbation, the The definition of water table implies a water
clearest manifestation of volcanic and hydrologi- saturated medium below it. However, crater lakes
cal interactions is a hydrothermal system. Fumar- (Fournier et al. 2009) and caldera settings (Bruno
oles, often visible on active volcanoes, represent et al. 2007; Jasim et al. 2015) often have portion
the surface expression of this hydrothermal of the water table sustained by a two phases
system. system (liquid and gas). Similarly, the conden-
A magmatic hydrothermal system is com- sation of magmatic gases (primarily vapour)
posed of three main elements: a host rock (or often feeds the groundwater reservoir (Chiodini
reservoir), which contains a circulating fluid, set et al. 2001). While the location of the water table
in motion by an igneous heat source (Fig. 1). is important, the conventional definition does not
While the difference in relief between stratovol- really apply in volcanic settings, especially in
canoes and calderas can lead to contrasting high-relief stratovolcanoes, in which it is often
hydrological systems, the lower limit of any unclear to what extent the edifice is water satu-
hydrological system is commonly defined as the rated. Specifically, the local water table may
brittle-ductile transition zone. Within this zone, differ from the regional water table, with highly
fluid pressures transition from hydrostatic to dynamic elevation changes controlled by both
lithostatic as rock permeability becomes severely meteoric and volcanic processes.
reduced (Fournier 1999). When this region is Many studies suggest the presence of high
subjected to high strain rates, fracturing may elevation springs, however it is not clear whether
occur, leading to episodic influxes of mass and they are fed by the regional water table or by
heat to the hydrothermal system (Bodnar et al. perched saturated layers high up on the cone
2007). While agreement exists on the definition (Cabrera and Custodio 2004; Custodio 2007;
of lower limit of a volcanic hydrological system, Cruz and Oliveira Silva 2001; Hemmings et al.
the same is not true for the upper limit. 2015a; Ingebritsen and Scholl 1993; Join et al.
We consider the water table as the upper limit 2005; Peterson 1972). Access to wells on the
of the hydrological system. However, the earth’s flank of volcanoes and geophysical imaging
surface could equally be considered part of the methods can help resolve the hydrogeology
system. This adds the further complexity of flow behind such springs (Finn et al. 1987, 2001;
within the unsaturated (vadose) zone (Hemmings Aizawa et al. 2008). However, in either scenario,
et al. 2015a). Furthermore, the upper limit of the the development of saturated flow units at high
hydrological system closely depends on the elevation has a particular relevance to forecasting
precipitation regime. Precipitation is a function volcanic hazards such as lahar, landslides and
of geography, including both latitude and eleva- flank collapses, which can be sudden, unpre-
tion. Together with surface processes it deter- dicted and deadly. Such mass wasting events all
mines the recharge dynamics of the aquifer and involve the displacement of material from the
86 A. Jasim et al.

Fig. 1 Conceptual model of a hydrothermal system in, a high-relief volcano (modified after Goff and Janik 2000) and
b caldera (modified after Kuhn 2004)
Groundwater flow and volcanic unrest 87

upper part of the volcano to the surrounding gas saturation (Sg) as the fraction of a represen-
valleys. They can be triggered by gravitational tative bulk volume of the porous medium filled
instability (often in response to heavy rainfall), by the gas phases, such that (Eq. 2)
increase in pore pressure, reduction of rock
strength, volcano-tectonic earthquake or intru- Sg þ Sw ¼ 1; ð2Þ
sion of magma. The depressurisation induced by
mass movement on the volcanic edifice may also The capillary pressure Pc (in Pa) is due to the
result in the sudden reactivation of the magmatic pressure difference between the two phases
system. (Eq. 3), hence
The waters circulating within volcanic sys-
P c ¼ Pg  P w ð3Þ
tems are often high-temperature, sometimes
supercritical fluids. They interact with the host
and is a unique function of water saturation (Sw).
rock through chemical reactions that result in
Finally, due to the competing flow of the two
hydrothermal alteration. The evolution and
phases, the relative permeability (kr) is less than
dynamics of hydrological and volcanic interac-
or equal to the single phase (usually water) per-
tions are strongly controlled by fluid flow. This,
meability, k (m2), of the medium. We define the
in-turn, is a function of the pressure, temperature,
relative permeability of the gas phase (krg) and
fluid composition, and, critically, the system’s
the relative permeability of the liquid phase (krw)
ability to transmit fluids.
as (Eqs. 4 and 5)

kg
3.1 Fluid flow krg ¼ ð4Þ
k
kw
Laminar flow through saturated porous media is krw ¼ ð5Þ
k
described by Darcy’s Law (Eq. 1),
where kg and kw are the effective permeabilities
k dp for each of the two fluids. Again the relative
q¼ ð1Þ
l dl permeabilities are assumed to be a unique func-
tion of water saturation (Sw). Hence, we define
where q is specific discharge (m/s), k is perme- the flow velocity vector (m/s) of the gas phase
ability of the porous media (m2), µ is fluid vis- along a direction D (m) as (Eq. 6)
cosity (Pa s), and dP/dl is the hydraulic head, or
pressure, gradient (Pa/m) along length l (m). The krg

vg ¼ k rPg  qg grD ð6Þ
fluid viscosity is usually approximated as that of lg
water (liquid or vapour). However, the combined
effect of topography and the presence of a deep and similarly the velocity (m/s) of the water
source of heat and fluids likely produce vast (Eq. 7)
unsaturated (or two-phase) portions within the
volcano. Dissolved air and gas near the surface, krw
vw ¼ k ðrPw  qw grDÞ ð7Þ
the phase transition from water (liquid) to vapour lw
due to the temperature gradient, and the decom-
pression of upwelling fluids are some of the where l and q are respectively the viscosity
processes that produce two-phase (liquid and (Pa s) and the density (kg/m3) of the gas (g) and
gas) fluid flow regions within a volcanic system. liquid (w) and g is the acceleration (m/s2) due to
To extend the Darcy’s equation to two phase gravity. Similarly, for non-laminar flow further
flow we define the liquid saturation (Sw) as the terms (e.g., Forchheimer term) can be added to
fraction of a representative bulk volume of the Darcy’s equation of fluid flow to consider the
porous medium filled by water and, similarly, the inertial effects due to turbulence.
88 A. Jasim et al.

3.1.1 Permeability and Porosity channels or fractures, as is common in volcanic

Permeability and porosity are the primary settings. Such flow pathways themselves are an
parameters controlling flow and storage of fluids active component of a dynamic system. Their
in the subsurface (Manning and Ingebritsen ability to transmit fluids and therefore the role
1999). The permeability also controls the heat they play in hydrothermal circulation can be
regime of the hydrothermal system: high per- modified by physical changes - the opening or
meabilities (  10−14 m2) favour advective closing of fractures in response to stress-field
transfer of heat away from the igneous source, changes - and chemical alteration, which can
resulting in low-temperature vapour-dominated both enhance and obstruct fluid flow, as dis-
systems. Contrastingly, low-permeabilities cussed in the next section.
(<10−16 m2) favour slow heat conduction, also
producing low-temperature systems. The hottest
hydrothermal plumes reside in intermediate per- 3.2 Chemical Reactions
meabilities, around 10−15 m2 (Hayba and
Ingebritsen 1997). Whilst permeability (k) is an Chemical alteration is an important process
important parameter, it is often one of the least within a hydrothermal system. Dissolution can
well constrained. It can vary over 17 orders of reduce cohesion and weaken a volcanic edifice.
magnitude, from  10−20 m2 in intact crystalline This can lead to catastrophic flank collapses and
rocks to  10−9 m2 in porous and fractured debris avalanches (Reid 2004). Such events can
basalt (Table 1). depressurise the magmatic system and trigger an
Porosity (/), the ratio of voids over total eruption. Conversely, chemical precipitation and
volume (voids and solid) in a given material, deposition processes can cause plugging in areas
defines the storage capacity of the rocks; it also of intense mineralisation, this can promote pres-
affects permeability and effective thermal con- surisation, which can also lead to flank collapse,
ductivity. In many porous and fractured media, as pore pressures increase within the edifice.
there is a positive correlation between / and k, as Such pressurisation can also generate violent
permeability is simply the interconnected pore steam-driven explosions (Ingebritsen et al. 2010).
network. Clays and volcanic tuff are unusual in Chemical dissolution and precipitation processes
that they can have high porosity values but low also alter the host-rock permeability. These pro-
permeabilities, at least in part due to their very cesses are sensitive to the thermodynamic con-
small particle sizes, propensity to bridge pores ditions and the proportions of different fluid
and form aggregates (Neuzil 1994). Permeability components (Pirajno 2010). Feedbacks between
in such lithology is greatly enhanced by the physical and chemical flow behaviour within a
presence of discontinuities such as fissures, hydrothermal system can modify the physical
joints, shears and faults that can connect other- and chemical characteristics of its surface
wise isolated pores. expression - hydrothermal fluid and fumarolic
Porosity and permeability of volcanic units discharge. Therefore, monitoring the hydrother-
are primarily controlled by the type of volcanic mal discharge can provide clues about the state
product (e.g., lava, pyroclastic density current, of volcanic unrest and can even be used to help
ash fall) and depositional environment. Inherent predict volcanic eruptions (Cronan et al. 1997).
heterogeneities between deposits can be To maximise the value of chemical analysis of
enhanced by subsequent compaction, fracturing hydrothermal discharge as a volcanic monitoring
and chemical alteration. The resulting hydroge- tool, and to fully understand the hazard presented
ology is complex; hydrological rock properties by hydrological and magmatic interactions it is
(porosity, permeability and thermal conductivity) necessary to quantify the rates, spatial distribu-
are strongly scale dependent, particularly when tion and physical effects of chemical alteration
fluid flow is focussed along high permeability within a hydrothermal system.
Groundwater flow and volcanic unrest 89

Table 1 Measured Rock type Permeability (m2) Porosity (%)

permeability and porosity
ranges for various rock Min Max Range
types Unconsolidated rocks
Gravel 10−10 10−7 25–40
−13 −9
Clean sand 10 10 5–50
Silty sand 10−14 10−10
Silt, loess 10−16 10−12 35–50
−20
Unweathered clay a
10 10−15 40–80
Consolidated rocks
Shale 10−20 10−16 0–10
−20 −17
Unfractured metamorphic and igneous 10 10 0–5
Sandstone 10−17 10−13 5–35
−16 −13
Limestone and dolomite 10 10 0–20
Fractured igneous and metamorphic 10−15 10−13 0–10
−14 −9
Permeable basalt 10 10 0–25
Karst limestone 10−13 10−9 5–50
Fractured basalt 5–50
Basalt near surfaceb 10−14 10−12
Basalt at 1 km depthb 10−18 10−10
−20
Andesite c
10 10−18 0.2–0.3
d −18 −14
Thermometamorphic 10 10 2–17
Campi Flegrei trachy-phonolite
Tuff, surfacee 10−16 10−15 48–52
−17
Tuff, depth e
10 10−15 19–52
d −18 −14
Chaotic tuff/tuffites < 1 km depth 10 10 6–40
Chaotic tuff/tuffites > 1 km depthd 10−18 10−14 5–36
f −16 −16
Tuffites (HT altered) 10 10 0.05–0.07
Lavad 10−18 10−14 7–25
Montserrat andesite
Pyroclastic flow depositg 10−18 10−13
Lavag 10−17 10−13
−14
Lahar deposit g
10 10−13
Sedimentary rocks are given for comparison. Hydrothermal systems in limestone usually
develop as skarn deposit. Data from Freeze and Cherry (1979). aNeuzil (1994),
b
Ingebritsen et al. (2006), cPetrov et al. (2005), dPiochi et al. (2014), ePeluso and Arienzo
(2007), fGiberti et al. (2006), gHemmings et al. (2015a)

3.2.1 Reaction Controlling Parameters Pirajno 2010) and appears to vary between dif-
Many factors may influence hydrothermal alter- ferent case studies.
ation, including temperature, pressure, rock type,
fluid flux, fluid composition, and time. The rel- 3.2.2 Fluid Composition
ative importance of each of these has been much The dominant form of chemical alteration (dis-
discussed in the literature (Gifkins et al. 2005; solution and/or precipitation) is principally a
90 A. Jasim et al.

function of the composition of the circulating result from extensive interaction with the reser-
hydrothermal fluids. Meteoric water and seawater voir rocks and may cause silica or carbonate
are the main sources of fluid in hydrothermal supersaturation at surface condition;
systems with an additional and dynamic contri- (ii) acid-sulphate waters which result from the
bution of magmatic fluids. The composition of condensation of volcanic gases into the shallower
hydrothermal fluid critically affects the groundwater system and are often depleted in
mineral-fluid equilibria and therefore the con- alkali and Cl but enriched in metals; (iii) bicar-
centration of rock forming elements (RFEs) such bonate waters which usually show thermody-
as Silica (SiO2), Sodium (Na), Potassium (K), namic equilibrium with the reservoir rock and are
Calcium (Ca) and Magnesium (Mg). The common at the edge of magmatic-hydrothermal
mineral-fluid equilibria and solubility of RFEs, as systems (Ellis and Mahon 1977; Giggenbach and
well as sulphate (SO4 2 ), chloride (Cl−) and Soto 1992; Goff and Janik 2000).
bicarbonate (HCO3 ), is affected by temperature,
pressure and water/rock (W/R) ratio as well as the 3.2.3 Acidity of Hydrothermal Fluids
composition of the host rock. These factors also Upper regions of hydrothermal systems are often
affect kinetic rate of chemical alteration. In addi- characterized by steam-heated fumarolic alter-
tion, the relative mobility of elements depends on ation due to the presence of acidic, sulphate-rich
the characteristics of fluid flow, the number of fluids (Rye 2005). These fluids may cause
phases (liquid and gas) and chemical condition leaching of the host rocks, resulting in an
along the flow path including pH, redox condition, increase in both rock porosity and permeability.
sulfidation state, availability of ligands. Eventually extreme acidic fluids (pH < 2) gen-
The chloride-sulphate-bicarbonate ternary erate the development of vuggy silica and
diagram by Giggenbach and Soto (1992) pro- thereby facilitate faster gas escape in the shallow
vides a tool to identify water end-members zone (Mayer et al. 2016 and references therein).
(Fig. 2). It represents graphically the classifica- In the presence of abundant sulphate ions and
tion of thermal water suggested by Ellis and Al-rich host rocks, within a lesser acidic envi-
Mahon (1977) based on major ions, which ronment (pH > 2), the formation of alunite
identifies (i) neutral alkali-chloride waters which dominates (alunitic alteration, Pirajno 2010).

Fig. 2 Classification ternary

diagram of thermal waters
from Giggenbach and Soto
(1992) based on the content of
major solutes: Cl−, SO4 2 and
HCO3 . The vertexes
represent the alkali-chloride,
sulphate and bicarbonate
water type end members,
whilst the arrows show major
differentiation processes
Groundwater flow and volcanic unrest 91

Fig. 3 Conceptual model for

the formation of near surface
high-sulfidation alteration. pH
and composition control the
development of alteration
zones with increasing distance
to the main fumarolic conduit.
A highly permeable, acidic
core characterized by
amorphous silica is laterally
replaced by a zone of alunite
and amorphous silica.
Successive neutralization of
the fluids promotes the
formation of kaolinite. Rock
permeability as well as the
degree of alteration increase
toward the center of the
hydrothermal activity (Mayer
et al. 2016)

Often the fluids undergo progressive neutraliza- W/R ratios, are termed “closed systems”, or “rock
tion as they flow away from the degassing vents. dominated systems”. In this case, the secondary
This results in a sequence of alteration facies minerals depend nearly entirely on temperature
(Fig. 3) from silicic to advanced argillic to and are of similar chemical composition to the
intermediate argillic (Fulignati et al. 1998). original rocks, although possibly in a hydrated
In similar environments the distribution of form (typical of propylitic alteration, see below).
kaolinite and alunite may also be affected by the The alteration minerals often resemble those due
presence of groundwater. Alunite preferentially to metamorphism (Giggenbach 1984).
forms at or above the groundwater table where Most hydrothermal systems are, however,
atmospheric oxygen could oxidize H2S to “open systems”, or “liquid dominated systems”,
H2SO4, which is required for the formation of which are characterised by high W/R ratios. In
alunite (Mutlu et al. 2005). these cases, the fluid composition has a greater
importance. The minerals that precipitate in open
3.2.4 Water/Rock Ratio systems are the result of alteration by mobile
The amount of water and the rock surface area fluids of constantly changing composition. In
available for reactions are two of the primary these systems permeability is highly influential
controls on alteration. Therefore, the Water/Rock and systematic spatial patterns of alteration
(W/R) ratio and rock porosity and permeability zoning are common.
will determine the type and extent of alteration.
W/R ratios range between 0.1 and 0.4 (Henley 3.2.5 Rock Type
and Ellis 1983) while porosity can vary from 0 Many studies support the assumptions that the
to  80% (Freeze and Cherry 1979; Neuzil chemical and mineralogical composition of the
1994). Static systems, with low porosity and low original rock will change the composition of the
92 A. Jasim et al.

equilibrium solution and therefore the rate of indi- 1985). Sharp pressure gradients can occur
vidual dissolution/precipitation reactions (e.g. between high and low permeability portions of a
Pirajno 2010). Consequently, various investiga- hydrothermal system, for example within the flow
tions have attempted to assess the most easily system feeding fumaroles. Major processes occur
altered minerals in the host rocks (e.g. Browne at the interface of liquid-gas phases, such as
1984). Glass, followed by olivine are the least stable massive precipitation of minerals (Lu and Kieffer
phases at surface conditions, as such basalts are 2009). In addition, high pressures can cause rock
likely to alter more rapidly than felsic rocks. compaction, thus reduce permeability and drive
Numerous studies of basalt dissolution have shown pressure solution. Conversely, rapid increase in
that the presence of glass can increase dissolution fluid pressure can promote fracturing leading to
rates (Wolff-Boenisch et al. 2006; Berger et al. increases in permeability.
1994; Stefansson and Gislason 2001; Zakharova
et al. 2007; Hu et al. 2010; Gudbrandsson et al. 3.2.7 Temperature
2011). However, experimental results on the alter- Temperature, on the other hand, controls the
ation of volcanic materials are biased by the dom- general alteration patterns of hydrothermal sys-
inant use of basalt as starting material. tems because it is the main control on mineral
Even though basaltic glass dissolves relatively solubility (Giggenbach 1988; Oelkers et al.
rapidly, basalts are still low-silica magmas, and 2009). For example, metal chlorides and alkaline
therefore silica concentrations remain higher in minerals are more soluble at high temperatures,
felsic rocks. Browne (1978) showed that, at while gypsum, anhydrite, calcite and dolomite
temperatures above 280 °C, the host rock com- show retrograde solubility below  100 °C
position has a negligible effect on alteration (Frazer 2014). Silica solubility increases as
minerals. Indeed, he gathered evidence of the temperatures rise, until  300 °C. With further
same stable alteration assemblages in basalts, temperature increases, silica solubility decreases
sandstones, rhyolites and andesites. Most acces- (Fournier 1985). These types of thermodynamic
sible hydrothermal systems, however, are at relationships in single-phase hydrothermal sys-
temperatures below 280 °C and in such systems tems are relatively well constrained, and are
Browne (1978) reports high-silica zeolites in reviewed in detail by Oelkers et al. (2009).
rhyolitic volcanoes, and low-silica zeolites in
basaltic and andesitic systems.
4 Hydrothermal Systems
3.2.6 Pressure and Unrest
Pressure generally has a secondary role in hy-
drothermal alteration (Robb 2005). An important Many of the conditions that control fluid flow
exception is the role of pressure in controlling and chemical alteration are modified by the
boiling in hydrothermal environments. Boiling at re-activation of the magmatic system, and evolve
depth leads to low-salinity vapour and during volcanic unrest. For example, the intro-
high-salinity brine. This phase separation is duction of fresh magma into the deep portions of
responsible for transport and deposition of ele- an active hydrothermal system can critically
ments that are key to ore mineralization (Henley change the pressure and temperature conditions
and Berger 2013). In the upper  400 m, pres- within the system, thus leading to the develop-
sure is due to the weight of the hot, possibly ment of gas pockets in the subsurface (Jasim
vapour rich, water column leading to pressure et al. 2015). This can rapidly lead to phreatic
gradient below hydrostatic. At the margin of the eruptions. Thus it is crucial to expand unrest
hydrothermal system mixing between cold and tracking to include monitoring of non-magmatic
hot water is enhanced by the pressure difference hazards (Sandri et al. 2017—this volume).
(Henley 1985). At greater depth, pressures exceed Seismicity, gases and ground deformation are
hydrostatic, feeding the upper reservoir (Henley usually monitored around active and restless
Groundwater flow and volcanic unrest 93

volcanoes (Sparks 2003). Gravity anomaly ligands Cl− and F−, from magmatic degassing,
studies (Gottsmann et al. 2008; Coco et al. 2016) can mobilize metals and F− greatly enhances the
and acoustic waves (Ferrazzini and Aki 1987) dissolution rates of aluminium silicates (Oelkers
also provide insight into subsurface processes. and Gislason 2001; Wolff-Boenisch et al. 2004)
Measurements of water chemistry composition with detrimental effect on rock mechanical
from hydrothermal manifestation such as boiling properties. However, along the upwelling flow
pools, crater lakes and thermal springs are also path, mixing with surface waters and chemical
routinely conducted (Varekamp et al. 2001, reactions with the rock often occur, overprinting
2009; Federico et al. 2002; Tassi et al. 2003). the magmatic signature. Magmatic gases and
The increase of Rare Earth Elements (REE)/Cl, vapour either mix with deep circulating water
RFEs (e.g., Ca, Mg, K)/Cl and the increase in (Giggenbach 1988) or they condense to form
SiO2 concentration are indicative of either in situ thermal waters (Rye 1993). In both cases,
intrusion of fresh magma within the hydrother- they rapidly dissociate and form a strong acidic
mal reservoir or exposure to water/rock interac- solution which causes cation leaching of the host
tion of fresh rock due to hydrofracturing rock leading to advanced argillitic alteration
(Varekamp et al. 2008). A pH drop and tem- (Giggenbach 1988; Symonds et al. 2001). Prior to
perature increase in spring water can also be mineral precipitation, the resulting waters are
indicative of an increase in magmatic activity. enriched in Si, Na+, K+, Mg2+ and Ca2+, and other
However the majority of springs in volcanic metals, proportionally to their concentrations in
environment are fed by the regional groundwater the host rock (rock congruent dissolution) until
reservoirs, thus representing the cooler water the solution is more or less neutralised. This is
inflow of the hydrothermal system (Jasim 2016). therefore called the “primary neutralisation
Thermal waters are focused on fluid upwelling zone”. Continued acid leaching in K-feldspar
pathways such as faults and fractures (Curewitz systems leads invariably to the formation of alu-
and Karson 1997; Hemmings et al 2015b; Jasim nite, an important component of high-sulfidation
et al. 2015). epithermal systems, where it can replace entire
The influx of magmatic fluids can manifest as masses of rocks. For example, alunite can be
changes in chemical compositions of hydrother- formed indirectly from K-feldspar through the
mal discharges. In particular the strong field formation of K-mica and kaolinite (Eqs. 8a–8c).

3KAlSi3 O8 þ 2H þ ! KAl3 Si3 O10 ðOHÞ2 þ 2K þ þ 6SiO2

ð8aÞ
K-feldspar K-mica

KAl3 Si3 O10 ðOHÞ2 þ 2H þ þ 3H2 O ! 3Al2 Si2 O5 ðOHÞ4 þ 2K þ

ð8bÞ
K-mica Kaolinite

3Al2 Si2 O5 ðOHÞ4 þ 2K þ þ 6H þ þ 4SO4 2 ! 2KAl3 ðSO4 Þ2 ðOHÞ6 þ 6SiO2 þ 3H2 O

ð8cÞ
Kaolinite Alunite
94 A. Jasim et al.

Most field and experimental observations pressurisation, hydrothermal outflow and phrea-
imply that permeability in hydrothermal systems tic explosions. Even slow processes such as the
tends to decrease with time. However, most of incremental development of pervasive alteration
these systems remain active for long periods of zones can manifest as a dynamic hazard in
time (typically 103–106 years, Ingebritsen et al. response to continued or future unrest, as strong
2010). Mechanisms must exist by which perme- crystalline rocks are hydrothermally altered into
able pathways are maintained and or developed to weak secondary clays.
allow continuing circulation of hydrothermal
fluids and ongoing alteration. Such mechanisms
include: (i) the periodic re-organisation of flow 5 Monitoring and Signals
patterns related to spatial variations in dissolution
and precipitation behaviour (Ritchie and Pritch- The dynamic hydrological and hydrothermal
ard 2011); (ii) the dissolution of minerals because response to volcanic unrest means that, bore-
of pulses of acidic fluids (Plumlee 1999); holes, springs, fumaroles, crater lakes and geo-
(iii) feedback between permeability reduction, physical imaging of the hydrological system can
fluid pressure and rock mechanics resulting in provide a rare window into the state of a volcano
hydrofracturing, shear dislocation, mineral dis- and the evolution of volcanic hazard. Hydrolog-
solution and the opening of flow pathways (Bar- ical monitoring itself is multi-parametric; insights
nes 2015; Weis 2015); (iv) pulsating volcanic can be gained from exploring physical and
activity causing fracturing, periodic variations in chemical patterns. For instance, the effect of
temperature, pressure and the composition of the groundwater on volcanic gases changes accord-
circulating fluids (Bodnar et al. 2007); (v) Cool- ing to their solubility. As such, a larger propor-
ing of the magmatic sills and dikes may lead to tion of SO2, HCl and HF emitted from magma
thermal cracking (Cathles et al. 1997) and the remain in solution in water compared to CO2 and
thermal expansion of the rock; both of which H2S. Hence, SO2, HCl and HF can be detected at
cause increases in fracture density (Chen et al. the surface, only during intense magmatic
1999) enhancing rock permeability; and activity or after drying of degassing pathways
(vi) stressed induced fracturing (Tapponnier and (Symonds et al. 2001). In the absence of active
Brace 1976). Furthermore, fluid pathways often degassing, the isotopic ratio of the gases dis-
ease the movement of magma towards the surface solved in groundwater2 such as 3He/4He (posi-
as shown by the 1975–1984 volcano-tectonic tive) and d13C (negative) can be indicative of a
crisis at Krafla caldera (Iceland), which lead to the magmatic source (Sorey et al. 1998; Allard et al.
emplacement of fault-controlled pseudodikes at 1997; Federico et al. 2002).
shallow depths (<100 m) and eruptive events Fluctuations of the water table/spring dis-
(Opheim and Gudmundsson 1989). charge have also been frequently recorded before
The evolution of a hydrothermal system the onset of magmatic activity and are often
involves the interplay between a number of interpreted as the effect of opening and closing of
mechanisms, physical and chemical, that operate fractures during the intrusion of fresh magma
at very different timescales ranging from seconds (Tanguy 1994; Shibata and Akita 2001; Newhall
to hundreds of years. Where volcanic unrest et al. 2001). Alternatively, the effect of the water
results in rapid modification of the hydrothermal phase transition from liquid to gas at relatively
system, hazards associated with unrest can shallow (<2 km) depth may also cause uplift of
manifest rapidly, with limited warning or pre- the water table (Jasim et al. 2015). Water levels
cursory activity. Dynamic changes in perme- in boreholes can be relatively easily monitored
ability, related to rapid mineral precipitation or and have been observed to respond to tectonic
opening of fractures can immediately modify and volcanic perturbations in a range of volcanic
flow pathways. This can result in dramatic settings (e.g. Usu Volcano, Japan; Kilauea Vol-
changes in heat and fluid flow, near surface cano, Hawai’i; Koryajskii Volcano, Kamchatka).
Groundwater flow and volcanic unrest 95

Level changes have been attributed to thermal volcano, Mexico would make a better monitoring
pressurisation, compression of water saturated target than near-neutral, high discharge springs.
rocks and opening of fractures in response to the Potential chemical indicators of unrest in these
intrusion of magma. However, the magnitude springs include increase in relative concentration
and even the sign of this hydrological response is of Mg, and increase in Cl/B and Cl/Br ratios.
a complex function of the nature of the thermal In summary, the physical and chemical sig-
and mechanical perturbation, the orientation and nals or hydrological perturbation associated with
connectivity of permeable pathways and even the magmatic unrest are complex and site dependant.
design of the well itself. Thus, interpreting such Relative contributions of different fluid compo-
signals in relation to magmatic unrest requires nents and their interaction with existing flow
some prior understanding of the hydrological pathways, can determine how a system evolves
features involved. during quiescent periods. This evolution dictates
Spring discharge fluctuations are harder to the likely response to thermodynamic and
measure than well water level changes, especially chemical perturbation associated with the initia-
on the flanks of volcanoes experiencing unrest, tion of volcanic unrest. Given the intricate feed-
and are therefore less well documented. Decline backs between magma, hydrology and hazards
in non-thermal spring discharge on Centre Hills, and the sudden changes of the systems involved,
Montserrat, were observed prior to the onset of only high-frequency (hour-week) monitoring of
volcanic activity at the adjacent Soufrière Hills temperature, pH, electrical conductivity of water,
Volcano in 1995. This was followed by an depth of the water table, REE, rock forming
increase after the cessation of the second eruptive elements and dissolved gas coupled with geo-
phase in 2004 (Hemmings et al. 2015a). The physical monitoring can untangle the evolution
mechanism behind such fluctuation is unclear, it of the magmatic system, the opening/closure of
may relate to fracture dynamics associated with fractures and the seasonal groundwater dynamic.
magmatic pressurisation (and depressurisation).
Regular temperature measurement and chemical
analysis of spring systems and hydrological lakes 6 Open Questions—Important
in volcanic settings can provide insights into the Unknowns
differences and changes in flow pathways related
to magmatic perturbation. We have established that circulating hydrother-
Chemical analysis of thermal springs and mal fluids are highly reactive and may result in
fumaroles are more common precipitation of alteration products or dissolution
hydrological/hydrothermal monitoring strategies of the host rock, both of which may cause
employed at active volcanoes. Changes in porosity, and permeability changes. However,
chemical composition and isotopic concentra- the precise nature of this alteration varies with
tions are often related to changes in the relative fluid chemistry, rock mineralogy and thermody-
contribution of magmatic fluids to other ground- namic conditions. This uncertainty in alteration
water species. Although there are general indi- makes predicting the impact of water/rock
cators for increased magmatic fluid contribution interaction (WRI) on porosity and permeability,
to discharging hydrothermal fluids, effective use and therefore on fluid flow, particularly chal-
of spring temperature, chemistry and discharge lenging. The background fluid flow regime is a
data as volcanic unrest monitoring tools requires a critical part of the local expression of heat-flow
good understanding of the underlying composi- as well as pressure distribution that surrounds a
tion of the hydrological and hydrothermal fea- magmatic system. As such it may exercise an
tures and the likely sensitivity to different important control over the dynamics of the
perturbation scenarios, in specific volcanic areas. magmatic system that is currently poorly under-
For example, Taran et al. (2008) proposed that stood. Data constraining the time scales over
lower flowing, acidic springs at El Chichón which hydrothermal alteration occurs, related to
96 A. Jasim et al.

data gathered from long term monitoring of Liquid liquid state of matter (e.g., water)
coupled magmatic-hydrothermal systems, are Gas gas state of matter (e.g., vapour)
thus crucial to inform ongoing interpretations
and further predictions of areas experiencing Water table level below which water saturation
magmatic-hydrothermal unrest. occurs

Acknowledgements The authors thank Shaul Hurwitz

for his constructive comments during the review of this
manuscript and Pablo Palacios for carefully editing the
extended abstract. This project has received funding from
References
the European Union’s Seventh Program for research,
technological development, and demonstration under Aizawa K (2008) Classification of self-potential anoma-
grant agreement no. 282759 (VUELCO). K. Mayer and lies on volcanoes and possible interpretations for their
B. Scheu also acknowledge the support of a PROCOPE subsurface structure. J Volcanol Geoth Res 175:253–
grant (Hot Hydrothermal Volcanic Systems; project-ID 268
57130387), funded and implemented by the Deutscher Allard P, Jean-Baptiste P, D’Alessandro W, Parello F,
Akademischer Austauschdienst (DAAD) in Germany, and Parisi B, Flehoc C (1997) Mantlederived helium and
the Ministry of Foreign and European Affairs (MAE) and carbon in groundwaters and gases of Mount Etna,
the Ministry of Higher Education and Research (MESR) Italy. Earth Planet Sci Lett 148:501–516
in France. K. Mayer and B. Scheu acknowledge the Barnes H (2015) Hydrothermal processes: the develop-
support of the ERC Advanced Investigator Grant ment of geochemical concepts in the latter half of the
(EVOKES—no. 247076). twentieth century. Geochem Perspect 4:1–93
Berger G, Claparols C, Guy C, Daux V (1994) Dissolu-
tion rate of a basalt glass in silica-rich solutions:
implications for long-term alteration. Geochim Cos-
Glossary mochim Acta 58:4875–4886
Bodnar RJ, Cannatelli C, De Vivo B, Lima A, Belkin HE,
Milia A (2007) Quantitative models for magma
Hydrothermal system A groundwater system degassing and ground deformation (bradyseism) at
that has an area of recharge, an area of dis- Campi Flegrei, Italy: implications for future eruptions.
charge, and a heat source. When a magma Geology 35:791–794
supplies the heat source and volatiles, the Browne PRL (1978) Hydrothermal alteration in active
hydrothermal system is termed a magmatic geothermal fields. Annu Rev Earth Planet Sci 6:
229–250
hydrothermal system Browne PRL (1984) Subsurface stratigraphy and
Hydrothermal alteration hydrothermal alter- hydrothermal alteration of Eastern section of the
Olkaria geothermal field, Kenya. In: Proceedings of
ation is a complex process involving chemi-
the 6th New Zealand geothermal workshop, vol 1,
cal, mineralogical, and textural changes, due pp 33–41
to the interaction of hot aqueous fluids and the Bruno PPG, Ricciardi GP, Petrillo Z, Di Fiore V,
host rocks through which they circulate Troiano A, Chiodini G (2007) Geophysical and
hydrogeological experiments from a shallow
Permeability Connected pore space of a rock or hydrothermal system at Solfatara Volcano, Campi
lithology, controlling fluid flow within a Flegrei, Italy: response to caldera unrest. J Geophys
reservoir Res: Solid Earth 112:1–17
Cabrera MC, Custodio E (2004) Groundwater flow in a
Porosity Ratio of voids over the total volume of volcanic-sedimentary coastal aquifer: Telde area, Gran
the rock with respect to a reference Canaria, Canary islands, Spain. Hydrogeol J 12:
rock-volume 305–320
Cathles LM, Erendi AHJ, Barrie T (1997) How long can a
Water/rock interaction (WRI) The set of hydrothermal system be sustained by a single intrusive
chemical reactions between aqueous fluids event? Econ Geol 92:766–771
and rocks. These reactions modify both the Chen Y, Xiaodong W, Fuqing Z (1999) Experiments on
thermal fracture in rocks. C Sci Bull 17:1610–1612
chemistry of the circulating fluid and the
Chiodini G, Frondini F, Cardellini C, Granieri D,
mineralogy of the host rock Marini L, Ventura G (2001) CO2 degassing and
Fluid generic term for either liquid or gas or energy release at Solfatara volcano, Campi Flegrei,
Italy. J Geophys Res 106:16213–16221
both
Groundwater flow and volcanic unrest 97

Christiansen LB, Hurwitz S, Saar MO, Ingebritsen SE, Fulignati P, Gioncada A, Sbrana A (1998) Geologic
Hsieh P (2005) Seasonal seismicity at western United model of the magmatic hydrothermal system of
States volcanic centers. Earth Planet Sci Lett 240: vulcano (Aeolian Islands, Italy). Mineral Petrol
307–321 62:195–222
Coco A, Gottsmann J, Whitaker F, Rust A, Currenti G, Giberti G, Yven B, Zamora, M, Vanorio T (2006)
Jasim A, Bunney S (2016) Numerical models for Database on laboratory measured data on physical
ground deformation and gravity changes during vol- properties of rocks of Campi Flegrei volcanic area
canic unrest: simulating the hydrothermal system (Italy). In: Zollo, Capuano, Corciulo (eds) Geophysical
dynamics of an active caldera. Solid Earth Discussions exploration of the Campi Flegrei (Southern Italy)
7:557–577 Caldera’ interiors: data, methods and results, vol 1,
Cronan DS, Johnson AG, Hodkinson RA (1997) pp 179–192
Hydrothermal fluids may offer clues about impending Gifkins CC, Herrmann W, Large RR (2005) Altered
volcanic eruptions. Eos 78:341–345 volcanic rocks: a guide to description and interpreta-
Cruz JV, Oliveira Silva M (2001) Hydrogeologic frame- tion. Ph.D. thesis, University of Tasmania, Australia
work of Pico Island, Azores, Portugal. Hydrogeol J (275)
9:177–189 Giggenbach WF (1984) Mass transfer in hydrothermal
Curewitz D, Karson JA (1997) Structural settings of alteration systems—A conceptual approach. Geochim
hydrothermal outflow: fracture permeability main- Cosmochim Acta 48:2693–2711
tained by fault propagation and interaction. J Volcanol Giggenbach WF (1988) Geothermal solute equilibria.
Geoth Res 79:149–168 Derivation of Na-K-Mg-Ca geoindicators. Geochim
Custodio E (2007) Groundwater in volcanic hard rocks. Cosmochim Acta 52:2749–2765
In: Krasny and Sharp (eds) Groundwater in fractured Giggenbach WF, Soto RC (1992) Isotopic and chemical
rocks. IAH Selected Paper Series, vol 9, pp 95–108 composition of water and steam discharges from
Ellis AJ, Mahon WAJ (1977) Chemistry and geothermal volcanic-magmatic-hydrothermal systems of the Gua-
systems. Academic Press (392) nacaste Geothermal Province, Costa Rica. Appl
Federico C, Aiuppa A, Allard P, Bellomo S, Jean-Baptiste Geochem 7:309–332
P, Parello F, Valenza M (2002) Magma-derived gas Goff F, CJ Janik (2000) Geothermal systems. In:
influx and water-rock interactions in the volcanic Sigurdsson, Houghton, McNutt, Rymer, Stix
aquifer of Mt. Vesuvius, Italy. Geochimica et Cos- (eds) Encyclopedia of volcanoes, vol 1, pp 817–834
mochimica Acta 66:963–981 Gottsmann J, Camacho AG, Martí J, Wooller L, Fernán-
Ferrazzini V, Aki K (1987) Slow waves trapped in a dez J, Garcia A, Rymer H (2008) Shallow structure
fluid-filled infinite crack: implications for volcanic beneath the Central Volcanic Complex of Tenerife
tremor. J Geophys Res 92:9215–9223 from new gravity data: implications for its evolution
Finn C, Williams DL (1987) An aeromagnetic study of and recent reactivation. Phys Earth Planet Inter
Mount St. Helens. J Geophys Res: Solid Earth 168:212–230
92:10194–10206 Gudbrandsson S, Wolff-Boenisch D, Gislason SR, Oelk-
Finn CA, Sisson TW, Deszcz-Pan M (2001) Aerogeo- ers EH (2011) An experimental study of crystalline
physical measurements of collapse-prone hydrother- basalt dissolution from 2 < pH < 11 and temperatures
mally altered zones at Mount Rainier volcano. Nature from 5 to 75 C. Geochim Cosmochim Acta 75:5496–
409:600–603 5509
Fournier RO (1985) The behavior of silica in hydrother- Hayba D, Ingebritsen S (1997) Multiphase groundwater
mal solutions. In: Berger and Bethke (eds) Geology flow near cooling plutons. J Geophys Res 102:12235–
and geochemistry of epithermal systems. Reviews in 12252
Economic Geology, vol 2, pp 45–61 Hedenquist J, Lowenstern J (1994) The role of magmas in
Fournier RO (1999) Hydrothermal processes related to the formation of hydrothermal ore deposits. Nature
movement of fluid from plastic into brittle rock in the 370:519–527
magmatic-epithermal environment. Econ Geol Hemmings B, Whitaker F, Gottsmann J, Hughes A
94:1193–1211 (2015a) Hydrogeology of montserrat, review and new
Fournier N, Witham F, Moreau-Fournier M, Bardou L insights. J Hydrol: Reg Stud 3:1–30
(2009) Boiling Lake of Dominica, West Indies: Hemmings B, Gooddy D, Whitaker F, Darling GW,
high-temperature volcanic crater lake dynamics. J Geo- Jasim A, Gottsmann J (2015b) Groundwater recharge
phys Res 114:1–17 and flow on Montserrat, West Indies: insights from
Frazer AM (2014) Advances in understanding the evolu- groundwater dating. J Hydrol: Reg Stud 4:611–622
tion of diagenesis in carboniferous carbonate plat- Henley RW (1985) The geothermal framework of
forms: insights from simulations of palaeohydrology, epithermal deposits. Rev Econ Geol 2:1–24
geochemistry, and stratigraphic development. PhD Henley RW, Berger BR (2013) Nature’s refineries metals
Thesis, University of Bristol, UK (261) and metalloids in arc volcanoes. Earth Sci Rev
Freeze R, Cherry J (1979) Groundwater. Prentice Hall 125:146–170
(604)
98 A. Jasim et al.

Henley R, Ellis A (1983) Geothermal Systems Ancient Oelkers EH, Benezeth P, Pokrovski GS (2009) Thermo-
and Modern: a Geochemical Review. Earth Sci Rev dynamic databases for water-rock interaction. In:
19:1–50 Oelkers and Schott (eds) Thermodynamics and kinet-
Hu SM, Zhang RH, Zhang XT, Huang WB (2010) ics of water-rock interaction. Reviews in Mineralogy
Experimental study of water-basalt interactions in and Geochemistry, 70, 1–37
Luzong volcanic basin and its applications. Acta Opheim JA, Gudmundsson A (1989) Formation and
Petrologica Sinica 26:2681–2693 geometry of fractures, and related volcanism, of the
Ingebritsen SE, Scholl MA (1993) The hydrogeology of Krafla fissure swarm, northeast Iceland. Geol Soc Am
Kilauea volcano. Geothermics 22:255–270 Bull 101:1608–1622
Ingebritsen S, Ward S, Neuzil C (2006) Groundwater in Peluso F, Arienzo I (2007) Experimental determination of
geologic processes. Cambridge University Press (564) permeability of Neapolitan Yellow Tuff. J Volcanol
Ingebritsen SE, Geiger S, Hurwitz S, Driesner T (2010) Geoth Res 160:125–136
Numerical simulation of magmatic hydrothermal sys- Peterson FL (1972) Water development on tropic volcanic
tems. Rev Geophys 48:1–33 islands-type example: Hawaii. Ground Water 10:18–23
Jasim A (2016) Exploring the complexity of groundwater Petrov VA, Poluektov VV, Zharikov AV, Velichkin VI,
flow in volcanic terrains: a combined numerical, Nasimov RM, Diaur NI, Terentiev VA,
experimental and field data approach. Ph.D. thesis, Shmonov VM, Vitovtova VM (2005) Deformation
University of Bristol, UK (199) of metavolcanics in the Karachay Lake area, Southern
Jasim A, Whitaker FF, Rust AC (2015) Impact of Urals: petrophysical and mineral-chemical aspects.
channelized flow on temperature distribution and fluid Geol Soc London, Spec Publ 240:307–322
flow in restless calderas: insight from Campi Flegrei Piochi M, Kilburn CRJ, Di Vito MA, Mormone A,
caldera, Italy. J Volcanol Geoth Res 303:157–174 Tramelli A, Troise C, De Natale G (2014) The
Join JL, Folio JL, Robineau B (2005) Aquifers and volcanic and geothermally active Campi Flegrei
groundwater within active shield volcanoes. Evolution caldera: an integrated multidisciplinary image of its
of conceptual models in the Piton de la Fournaise buried structure. Int J Earth Sci 103:401–421
volcano. J Volcanol Geoth Res 147:187–201 Pirajno F (2010) Hydrothermal processes and mineral
Kuhn M (2004) Reactive flow modeling of hydrothermal systems. Springer, Berlin (1250)
systems. Springer, Berlin (264) Plumlee GS (1999) The environmental geology of mineral
Lu X, Kieffer S (2009) Thermodynamics and mass deposits. In: Plumlee and Logsdon (eds) The Envi-
transport in multicomponent, multiphase H2O systems ronmental Geochemistry of Mineral Deposits, Part A.
of planetary interest. Annu Rev Earth Planet Sci Processes, Techniques, and Health Issues: Society of
37:449–477 Economic Geologists. Reviews in Economic Geology,
Manning C, Ingebritsen S (1999) Permeability of the 6, 71–116
continental crust: Implications of geothermal data and Reid ME (2004) Massive collapse of volcano edifices
metamorphic systems. Rev Geophy 37:127–150 triggered by hydrothermal pressurization. Geology
Mason BG, Pyle DM, Dade WB, Jupp T (2004) 32:373–376
Seasonality of volcanic eruptions. J Geophys Res: Ritchie LT, Pritchard D (2011) Natural convection and
Solid Earth 109:1–12 the evolution of a reactive porous medium. J Fluid
Mayer K, Scheu B, Montanaro C, Yilmaz TI, Isaia R, Mech 673:286–317
Aßbichler D, Dingwell DB (2016) Hydrothermal Robb L (2005) Introduction to ore-forming processes.
alteration of surficial rocks at Solfatara (Campi Blackwell Science Ltd (384)
Flegrei): Petrophysical properties and implications Rye RO (1993) The evolution of magmatic fluids in the
for phreatic eruption processes. J Volcanol Geoth Res epithermal environment: the stable isotope perspec-
320:128–143 tive. Economica Geologica 88:733–753
Mutlu H, Sariiz K, Kadir S (2005) Geochemistry and Rye RO (2005) A review of the stable-isotope geochem-
origin of the Şaphane alunite deposit, Western Ana- istry of sulphate minerals in selected igneous envi-
tolia, Turkey. Ore Geol Rev 26:39–50 ronments and related hydrothermal systems. Chem
Neuzil CE (1994) How permeable are clays and shales? Geol 215:5–36
Water Resour 30:145–150 Saar MO, Manga M (2003) Seismicity induced by
Newhall CG, Albano SE, Matsumoto N, Sandoval T seasonal groundwater recharge at Mt. Hood, Oregon.
(2001) Roles of groundwater in volcanic unrest. Earth and Planetary Science Letters 214:605–618
J Geol Soc Philippines 56:69–84 Sandri L, Tonini R, Rouwet D, Constantinescu R,
Oelkers EH, Gislason SR (2001) The mechanism, rates Mendoza-Rosas AT, Andrade D, Bernard B (2017)
and consequences of basaltic glass dissolution: I. An The need to quantify hazard related to non-magmatic
experimental study of the dissolution rates of basaltic unrest: from BET_EF to BET_UNREST. This volume
glass as a function of aqueous Al, Si and oxalic acid Shibata T, Akita F (2001) Precursory changes in well
concentration at 25 C and pH = 3 and 11. Geochim water level prior to the March, 2000 eruption of Usu
Cosmochim Acta 65:3671–3681 volcano, Japan. Geophys Res Lett 28:1799–1802
Groundwater flow and volcanic unrest 99

Sorey ML, Evans WC, Kennedy BM, Farrar CD, Varekamp JC (2008) The volcanic acidification of
Hainsworth LJ, Hausback B (1998) Carbon dioxide glacial Lake Caviahue, Province of Neuquen, Argen-
and helium emissions from a reservoir of magmatic tina. J Volcanol Geoth Res 178:184–196
gas beneath Mammoth Mountain, California. J Geophy Varekamp JC, Ouimette AP, Herman SW, Bermudez A,
Res: Solid Earth 103:15303–15323 Delpino D (2001) Hydrothermal element fluxes from
Sparks RSJ (2003) Forecasting volcanic eruptions. Earth Copahue, Argentina: A ”beehive” volcano in turmoil.
Planet Sci Lett 210:1–15 Geology 29:1059–1062
Stefansson A, Gislason SR (2001) Chemical weathering Varekamp JC, Ouimette AP, Herman SW, Flynn KS,
of basalts, Southwest Iceland: effect of rock crys- Bermudez A, Delpino D (2009) Naturally acid waters
tallinity and secondary minerals on chemical fluxes to from Copahue volcano, Argentina. Appl Geochem
the ocean. Am J Sci 301:513–556 24:208–220
Symonds RB, Gerlach TM, Reed MH (2001) Magmatic Weis P (2015) The dynamic interplay between saline fluid
gas scrubbing: implications for volcano monitoring. flow and rock permeability in magmatic-hydrothermal
J Volcanol Geoth Res 108:303–341 systems. Geofluids 15:350–371
Tanguy JC (1994) The 1902-1905 eruptions of Montagne Wolff-Boenisch D, Gislason SR, Oelkers EH, Putnis CV
Pelee, Martinique: anatomy and retrospection. J Vol- (2004) The dissolution rates of natural glasses as a
canol Geoth Res 60:87–107 function of their composition at pH 4 and 10.6, and
Tapponnier P, Brace WF (1976) Development of temperatures from 25 to 74 C. Geochim Cosmochim
stress-induced microcracks in Westerly granite. Int J Acta 68:4843–4858
Rock Mech Min Sci 13:103–112 Wolff-Boenisch D, Gislason SR, Oelkers EH (2006) The
Taran Y, Rouwet D, Inguaggiato S, Aiuppa A (2008) effect of crystallinity on dissolution rates and CO2
Major and trace element geochemistry of neutral and consumption capacity of silicates. Geochim Cos-
acidic thermal springs at El Chichon volcano, Mexico. mochim Acta 70:858–870
J Volcanol Geoth Res 178:224–236 Zakharova EA, Pokrovsky OS, Dupre B, Gaillardet J,
Tassi F, Vaselli O, Capaccioni B, Macias JL, Nencetti A, Efimova LE (2007) Chemical weathering of silicate
Montegrossi G, Magro G (2003) Chemical composi- rocks in Karelia region and Kola peninsula, NW
tion of fumarolic gases and spring discharges from El Russia: assessing the effect of rock composition,
Chichon volcano, Mexico: Causes and implications of wetlands and vegetation. Chem Geol 242:255–277
the changes detected over the period 1998-2000.
J Volcanol Geoth Res 123:105–121

Open Access This chapter is licensed under the terms of chapter's Creative Commons license, unless indicated
the Creative Commons Attribution 4.0 International otherwise in a credit line to the material. If material is not
License (http://creativecommons.org/licenses/by/4.0/), included in the chapter's Creative Commons license and
which permits use, sharing, adaptation, distribution and your intended use is not permitted by statutory regulation
reproduction in any medium or format, as long as you or exceeds the permitted use, you will need to obtain
give appropriate credit to the original author(s) and the permission directly from the copyright holder.
source, provide a link to the Creative Commons license
and indicate if changes were made.The images or other
third party material in this chapter are included in the
 Experimental Simulations of Magma
Storage and Ascent

C. Martel, R.A. Brooker, J. Andújar, M. Pichavant,

B. Scaillet and J.D. Blundy

Abstract
One of the key issues in utilizing precursor signals of volcanic eruption is
to reliably interpret geophysical and geochemical data in terms of magma
movement towards the surface. An important first step is to identify where
the magma is stored prior to ascent. This can be studied through
phase-equilibrium experiments designed to replicate the phase assemblage
and compositions of natural pyroclasts or by measuring volatiles in melt
inclusions from previous eruptions. The second crucial step is to
characterize the magmatic conditions and processes that will guide the
eruption style. This may be addressed through controlled dynamic
decompression or deformation experiments to examine the different rates
that govern the kinetics of syn-eruptive degassing, crystallization, and
strain. Comparing the compositional and textural characteristics of these
experimental products with the natural samples can be used to retrieve
magma ascent conditions. These experimental simulations allow interpre-
tation of direct observations and in situ measurements of syn-eruptive
processes leading to more accurate forecasting of future eruptive
scenarios.

1 Linking Geophysical
and Geochemical Warning
Signals to Magmatic Processes
C. Martel (&)
J. Andújar
M. Pichavant

B. Scaillet
Institut Des Sciences de La Terre D’Orléans, A key objective in volcanology is to forecast
Université D’Orléans-CNRS-BRGM, Orléans, eruptions, i.e. to establish when, where, and how
France an eruption will occur and what magnitude it will
e-mail: [email protected]
be. The prerequisite of such forecasting is to
R.A. Brooker
J.D. Blundy (i) detect reliable precursory signals of magma
School of Earth Sciences, University of Bristol,
Wills Memorial Building, Queens Road, Bristol BS8 ascent to subsurface and (ii) anticipate the erup-
1RJ, UK tion style in order to inform the crisis

Advs in Volcanology (2019) 101–110

DOI 10.1007/11157_2017_20
© The Author(s) 2017
Published Online: 02 August 2017
102 C. Martel et al.

management strategy. To reach these objectives, 2 Magma Storage

the intensification of the geophysical and geo-
chemical signals associated with an unrest epi- How and where magma is stored before an
sode has to be interpreted in terms of magma eruption are enduring and complex questions,
movements. This is a far from trivial task particularly given the range of hypotheses cov-
because (i) seismicity has to distinguish signals ering single versus multiple storage regions or
related to magma movements from those of rock dyke feeder systems. Key parameters in inter-
fracturing and/or gas percolation (see Chapter preting the precursory geophysical and geo-
“Volcano Seismology: detecting unrest in wiggly chemical signals are the depth of storage and the
lines”), (ii) ground deformation has to precisely volatile content dissolved in the magma, as the
track magma motion towards the surface (see exsolution of these provides an important driving
Chapter “Volcano geodesy and multiparameter force for explosive eruptions. Magma consists
investigations”), and (iii) the flux and the spe- principally of phenocrysts (crystals larger than
ciation of emitted gas at the surface has to be 50–100 µm) coexisting with a silicate melt con-
interpreted in terms of magma ascent and taining dissolved volatile species (e.g. H2O, CO2,
degassing (see Chapter “Volcanic gases and low S species, F, Cl, etc.…). To assess the magma
temperature volcanic fluids”). For any of these storage conditions, one has to determine the
monitoring signals, their interpretation in terms parameters (i.e. the pressure, temperature, redox
of impending eruption requires knowing at what state, volatile content) that govern equilibrium
depth beneath the volcano magma is stored between melt and the phenocrysts. This is
(magma storage conditions) and how it pro- accomplished by comparing the phase assem-
gresses toward the surface (magma ascent con- blage and compositions of the natural products to
ditions). A pertinent approach to investigate the those obtained by phase-equilibrium experiments
conditions of magma storage and ascent consists where all these parameters are controlled.
of comparing petrological and textural studies of
previously erupted products to the results of
experimental simulations carried out under real- 2.1 Decoding Natural Pyroclasts
istic magma conditions. During the last decades
the development of powerful analytical and Phenocrysts. Mineral compositions are sensitive
experimental tools has led to great advances in to various intensive (pressure, temperature, ƒO2)
this type of investigation. Of course, this is an a and extensive (host liquid composition and
posteriori approach (using previously erupted volatile content) variables. In some cases, the
products) that relies on considering past eruptive compositions can be used as geothermobarome-
behaviour of a given volcanic system as a guide ters (or hygro/oxy/chemo-meters) to retrieve
to future activity. For this reason, it is necessary first-order parameters of crystallization. For
to understand the fundamental magmatic pro- instance, we can calculate crystallization tem-
cesses at any particular volcano and in the long perature from the composition of orthopyroxene-
term, build up a record that links the pre-eruptive clinopyroxene pairs (e.g. Lindsley and Andersen
signals with eruptive products. In this way we 1983) or amphibole-plagioclase pairs (e.g.
can successfully use the warning signals to Holland and Blundy 1994), temperature and
forecast or even start to predict the timing and oxygen fugacity from Fe–Ti oxide pairs (e.g.
style of an imminent eruption. Ghiorso and Evans 2008) or pressure from
Experimental Simulations of Magma Storage and Ascent 103

amphibole composition (e.g. Ridolfi and Renzulli timescales of reservoir dynamics, such as resi-
2012). Such calculations require two main cri- dence times between the last magma recharge and
teria: (i) the thermobarometers must have been eruption (e.g. Saunders et al. 2012). Indeed, the
calibrated experimentally for conditions and compositional zoning of some phenocrysts
compositions relevant to those of the target witness cycles of recharge events prior to the final
samples and (ii) the analysed phenocrysts must eruption (e.g. Druitt et al. 2012). Diffusion
be identified as part of the phenocryst assem- chronometry provides a means to infer the crystal
blage in chemical equilibrium with the melt in residence time in the reservoir prior to eruption
the reservoir under pre-eruptive conditions. This (e.g. Costa and Morgan 2010), but again requires
can achieved by comparison with experiments experimental calibration of the rates involved
that simulate a range of variables. Establishing under various conditions.
different ‘equilibrium’ events also becomes cru- Glass inclusions. Glass (or “melt”) inclusions
cial as magma mixing (and associated reheating are aliquots of melt that are trapped in crystals,
prior to eruption or further crystallization outside usually during stages of rapid crystal growth
the storage zone) very often blur the pre-intrusion/ (Fig. 1a). If trapped in phenocrysts crystallizing
eruption equilibrium conditions. Moreover, many in the magma chamber, these inclusions will be
eruptions involve the entrainment of crystals that the only witness of the composition of the melt in
did not grow from the erupted magma and are equilibrium with the phenocrysts prior to erup-
consequently out of equilibrium with the melt in tion, provided that they remained sealed after
which they occur. These are often termed xeno- entrapment (no volatile leak, no crystallization or
crysts or antecrysts and increasingly recognised as post-entrapment interaction with the host min-
important aspects of magma’s crystal cargo (e.g. eral). Indeed, after leaving the storage region, the
Streck 2008; Kilgour et al 2013). The processes of melt initially surrounding the phenocrysts is
mixing, reheating, and potentially decompression likely to degas and crystallize microlites upon
recorded in these crystals, may be used to provide ascent, therefore deviating significantly from its

(a) (b)

Fig. 1 Melt inclusion (MI) volatile contents from the decompression effect during eruption. In b the maximum
Morne aux Diables complex (MAD), northern Dominica, amount of H2O and CO2 that can be dissolved in a melt at
as determined by ion-microprobe measurements. a Close any given depth follows pressure dependent isopleths that
up of inclusions shows they were either slightly vesicu- can be determined experimentally (in this case from
lated, during capture (fluid-saturated) or became so during Tamic et al. 2001). This represents the minimum pressure
ascent. The right hand crystal contains a very large melt (depth of entrapment of a MI). These trends are controlled
inclusion around the bright oxide inclusions that has by the lower solubility of CO2 compared to H2O, such
vesiculated more extensively, possibly due to cracks that there is a rapid drop in CO2 before H2O starts to be
developing and exposing the melt to the full lost (coloured arrows)
104 C. Martel et al.

pre-eruptive composition and character. The two cool. Therefore, imprudent use of geothermo-
main objectives of studying glass inclusions in barometers may yield large pressure and tem-
phenocrysts are (i) mineral-melt thermobarome- perature ranges that cannot be easily reconciled
try (clinopyroxene-melt or plagioclase-melt; e.g., with a single/specific episode of equilibrium
Putirka 2008; Water and Lange 2015) and crystallization. In order to retrieve the storage
(ii) pre-eruption volatile contents and speciation conditions of the resident magma, i.e. prior to
that can be converted into gas saturation pressure deep magma mixing or before possible modifi-
using experimentally-derived volatile solubility cation within the volcanic conduit, a detailed
laws and which represent an end-member in the petrological study is necessary to identify pre-
calculations of the volcanic degassing budget cisely the different stages of perturbation and
(dissolved vs. released gases). This latter their characteristics in terms of phase assemblage
approach requires that the magma be demon- and chemical composition. Experimental petrol-
strably volatile-saturated at the time of melt ogy is one of the tools that helps to unravel the
inclusion entrapment, for example by looking at various magmatic processes at work and their
relationships between dissolved volatile contents relative impact on magma chemistry and mag-
and trace elements in inclusions (Blundy and matic evolution.
Cashman 2008). If the magma was not saturated,
the calculated pressure generally represents a
minimum depth prior to volatile volatile- 2.2 Phase-Equilibrium Experiments
saturated, the calculated pressure generally rep-
resents a minimum pressure estimate. As an Phase-equilibrium experiments use natural (or
example, the glass inclusions in phenocrysts analogue) products as starting material that are
from pyroclasts of the last eruption of Morne aux subjected to high-pressure (HP) and
Diables, Dominica, have been analysed by ion high-temperature (HT) in various devices under
microprobe. The data show about 6 to 8 wt% controlled conditions of pressure, temperature,
dissolved H2O, up to 3000 ppm CO2, together oxygen fugacity, and volatile content. Such
with some chlorine and fluorine. Available experiments are powerful tools to simulate real-
H2O–CO2 solubility models based on experi- istic magmatic conditions for the crustal reser-
ments for comparable compositions (e.g. Tamic voirs that feed volcanic systems. Experimental
et al. 2001) indicate melt entrapment during equipment ranges from cold-seal pressure ves-
phenocryst crystallization at pressures as high as sels, internally-heated pressure vessels, and
400–500 MPa (depth of <22 km) for sample piston-cylinder apparatus, depending on the
DC139 and shallower depth for sample DC08 investigated conditions. The principle is to
although it is also possible this magma originally reproduce the natural assemblage, proportion,
contained even more CO2 than is recorded by any and chemical compositions of the phenocrysts
of the analysed melt inclusions. (Figure 1b). and equilibrium coexisting melt in the magma
Small vesicles in the inclusions of Fig. 1a suggest storage region. This is then compared with the
there was exsolution of volatiles during ascent. natural samples in order to retrieve the
The possible disequilibrium between such vesicles pre-eruptive crystallization conditions (Fig. 2).
and melt produced during very rapid ascent is The first prerequisite for such an approach is a
discussed in Chapter “Magma degassing: the detailed petrological and mineralogical study of
diffusive fractionation model and beyond”. the erupted samples in order to identify the
Complexity of open-systems. An eruption is magmatic processes potentially perturbing equi-
often triggered by the injection of new magma librium crystallization in the reservoir. Indeed,
into the reservoir, which reheats, mingles, and the relevance of the experimental study relies on
mixes with the resident magma. Upon ascent to accurate petrological knowledge that dictates the
the surface, the two batches may interact to choice of the starting material and run procedure
varying degrees and can further crystallize and (Pichavant et al. 2007). The second prerequisite
Experimental Simulations of Magma Storage and Ascent 105

Fig. 2 Phase equilibrium experiments for the Tungu- dissolved in melt), c 975 °C and *5.7 wt% H2O, and
rahua 2006 andesite, showing a mineral stability fields as d 975 °C and 5.2 wt% H2O. Note the drastic increase of
a function of temperature and H2O content at 200 MPa crystal content with cooling or dehydration; gl for glass
and oxidizing conditions (fO2 = NNO + 0.8 log unit). (L for silicate liquid), ol for olivine, cpx for clinopyrox-
SEM images of the experimental charges are shown in ene, opx for orthopyroxene, pl for plagioclase, amph for
b for 1000 °C and H2O saturation (*6.1 wt% H2O amphibole, mag for magnetite, and ilm for ilmenite

is the appropriate choice of the volatile species et al. (2010) speculate that many magmas had
(H2O, CO2, sulphur, etc.) and contents to be original CO2 contents significantly higher than
added to the starting material. These dissolved those recorded by any melt inclusions.
volatiles can impact crystallization (sequence, It is clear that an approach involving the
mineral stability fields, and phase compositions; combination of petrological study of the natural
Scaillet and Pichavant 2003; Riker et al. 2015). products and phase-equilibrium experiments can
Where volatile measurements on melt inclusions help to retrieve the storage conditions of magmas
are available, these can be used, although it is of a wide range of compositions. This is a pre-
possible that the melt inclusions may no longer requisite step for the interpretation of unrest
be representative of the initial volatile content signals and construction of eruptive scenarios.
(e.g. due to possible leakage or recrystallization).
Consequently, volatile species and contents
become experimental parameters that have to be 3 Magma Ascent
varied within a range that is first inferred from
the study of the glass inclusions (when avail- During ascent in the volcanic conduit the silicate
able), and/or based on previous work carried out melt around the phenocrysts degasses by
on similar bulk rock compositions. It should be exsolving its dissolved volatiles as gas bubbles.
remembered that entrapment of melt inclusions This may lower the liquidus triggering the crys-
requires crystallisation to occur and magmas may tallization of microlites (i.e. crystals smaller than
undergo substantial volatile loss (degassing) about 50–100 µm). The residual melt is trans-
prior to any crystallisation. For instance, Blundy formed both chemically, by degassing and
106 C. Martel et al.

differentiation as microlites crystallize and processes, because diffusion in the melt, that con-
physically, by an increase in melt viscosity and a trols crystal growth, occurs on timescales ranging
change from a single liquid phase to a from hours in mafic melts, to days in silicic melts.
three-phase suspension (i.e. liquid, gas bubble, Microlite number density, volume proportion, size
and microcrystals). Both types of transformations and shape have all been used as markers of the
have drastic effects on the bulk magma flow undercooling (liquidus temperature minus magma
conditions (rheology) that control ascent rate and temperature) that drives crystallization (e.g. Ham-
the ductile versus brittle behaviour of the magma. mer et al. 1999); the higher the undercooling, the
The rate of magma decompression/ascent is more numerous, smaller, and irregularly-shaped the
the key parameter that controls the kinetics of crystals. With decompression (and dehydration of
degassing and crystallization, and ultimately, the the melt), liquidus temperature increases (as does
eruptive style. In silicic to intermediate systems, undercooling), so that it becomes possible to infer
slow ascent rates typical of effusive eruptions the depth of crystallization in the conduit by relating
such as lava flow or dome growth, i.e. cm/s to the textural characteristics of the microlites to
mm/s (Gardner and Rutherford 2000) yield undercooling and pressure (Fig. 3). This approach
timescales long enough for extensive degassing has been used by Melnik et al. (2011) to constrain
and crystallization. In contrast, the high ascent both magma flow and reservoir shape for the 1980–
rates prevalent during paroxysmal Strombolian or 86 dome-forming eruptions of Mount St. Helens
Plinian eruptions (i.e. of the order of m/s; Gard- (USA). Such modelling requires an accurate
nerand and Rutherford 2000) are able to generate determination of the dependence of undercooling
physico-chemical disequilibria of both degassing on pressure, which can be achieved through
and crystallization processes, driving gas over- decompression experiments (e.g. Riker et al. 2015).
pressures that may be released explosively.

3.2 Dynamic Experiments

3.1 Textures of Natural Pyroclasts
Dynamic experiments, such as decompression,
Decompression-induced degassing of the magma deformation or shock-wave experiments, are
creates gas bubbles, the number density of which valuable tools to investigate degassing, crystal-
has been demonstrated to correlate with the lization, strain, mixing, or fragmentation of a
decompression rate simulated by experiments magma. They provide information on the physics,
(Mourtada-Bonnefoi and Laporte 2004; Mangan chemistry, and kinetics of syn-eruptive magmatic
et al. 2004). The growing bubbles can rapidly processes, which can be used in turn to decode
coalesce and form gas escape channels. In this case, natural pyroclast formation in order to better
the bubble number densities in the identify geophysical and geochemical precursory
erupted/quenched pyroclasts may no longer be signals of an eruption. Shock-wave experiments
representative of the initial ascent rates under which dedicated to the fragmentation process are covered
nucleation was triggered. The timescale of out- in Chapter “From unrest to eruption: Conditions
gassing by magma foam collapse varies from a for phreatic versus magmatic activity”; the dis-
couple of hours to about 1000 h for magmas having cussion below concentrates on decompression and
bulk viscosities of *104 to 105.5 Pa.s, respectively deformation experiments.
(Martel and Iacono-Marziano 2015). This restricts Decompression experiments performed at
the use of the degassing process to simulations of elevated pressure (HP) and temperature
rapid magma ascent rates such as those during (HT) have proved useful in accurately simulating
Plinian events. To investigate longer transit times in magma ascent in volcanic conduits (e.g. Hammer
the conduit, one requires information from mag- and Rutherford 2002). In particular, experimental
matic processes with timescales longer than decompression rates can cover the most of the
degassing. Microlite crystallization is one of those perceived range of magma ascent rates during
Experimental Simulations of Magma Storage and Ascent 107

Temperature (°C)
900 1000 1100 1200

ΔTeff = 200°C
50 Plagioclase microlites

ΔTeff = 125°C
Pressure (MPa)

100
Dendritic ΔTeff = 200°C

150 Skeletal ΔTeff = 125°C

200
Tabular ΔTeff = 50°C
875°C
10 µm

Fig. 3 Relationships between effective undercooling represented by scarce large tabular crystals. With decreas-
(DTeff; in green) and microlite textural characteristics in ing pressure and increasing DTeff, microlites become more
a H2O-saturated rhyolitic melt (modified after Mollard numerous, but smaller in size, and they show more
et al. 2012). Plagioclase microlites crystallized at complex shapes ranging from skeletal (hollow) to
150 MPa after an isothermal quasi-instantaneous decom- dendritic
pression from 200 MPa, i.e. DTeff = 50 °C, are

volcanic eruptions. For instance, Plinian ascent (Martel and Iacono-Marziano 2015). Deforma-
rates of the order of m/s can be simulated exper- tion experiments performed in vessels equipped
imentally by decompression durations from sec- with torsion or coaxial deformation modules have
onds to hours whereas the slow ascent rates shown that the lifetime of such magmatic foams is
recorded for dome eruptions can be reproduced drastically reduced when a differential stress field
by decompression durations of several days or prevails, because it enhances bubble coalescence
weeks. More generally, decompression experi- (e.g. Okumura et al. 2009). The recent imple-
ments can simulate different natural eruptive mentation of HP-HT devices that allow magma
scenarios depending on the applied decompres- deformation at pressure coupled with in situ
sion rate, final pressure, and dwell time at final measurements of permeability, represents a con-
pressure. In basaltic H2O- and CO2-bearing siderable step forward for investigating the
magmas, experimental decompression in the explosive-effusive transition of volcanic erup-
duration range of <1–10 h has provided infor- tions in the laboratory under realistic conditions
mation on degassing processes leading to either (Kushnir et al. 2017).
regular or paroxysmal Strombolian eruptions Figure 4 illustrates how timescales of degas-
(Chapter 15 “Magma degassing: the diffusive sing and crystallization during decompression can
fractionation model and beyond”). In silicic be used to decipher eruption style. Magmas from
melts, decompression pathways and durations both, Plinian and dome-forming eruptions (dome,
from ten seconds to forty days have been inves- block-and-ash flows, surges), degas during ascent.
tigated experimentally to evaluate the lifetime of However, gases in dome-forming magmas escape
rhyolitic foams as a function of bulk viscosity from the melt (leading to dense pyroclasts)
108 C. Martel et al.

Fig. 4 Deciphering eruption Plinian eruption Dome-forming eruptions

explosivity from

Eruption
decompression-induced
Fallout Surge Block-and-ash flow
(Reboubt, 1990) (Mt Pelée, 1902) (Merapi, 2006)
degassing and crystallization
experiments (see text)
[Photograph of Redoubt is
from R. Clucas (USGS), Mt.
Pelée is from Lacroix (1904),
and Merapi comes from the

(decompression-induced degassing and crystallization)

website: gunungmerapi. Pumice
weebly.com] No microlites Shallow-level Deep-conduit
microlite microlite
Volcanic conduit + Dome
nucleation nucleation and
and gas continuous
10 µm overpressure 20 µm growth

Ascent < 1 day Ascent 1-3 days Ascent > 3 days

Foam collapse
(gas loss)

Bubble nucleation, growth,

and coalescence
Reservoir

100 µm

whereas Plinian foams have no time to collapse This application of decompression experi-
through gas escape (leading to pumiceous pyro- ments highlights the possibility of forecasting the
clasts) which suggests Plinian ascent durations are style of an eruption provided the magma ascent
limited to a couple of hours (Martel and rate towards the surface can be determined by
Iacono-Marziano 2015). Furthermore, in contrast some remote means (seismology, gravity, geo-
to Plinian magmas, dome-forming magmas have desy, gas discharge).
time to crystallize during ascent. At Mt. Pelée, the
moderately-explosive block-and-ash flows in
1929–1932 may have degassed and crystallized 4 Future Directions
continuously during an ascent lasting more than
3–6 days, so that little gas overpressure remains at The combination of equilibrium and dynamic
dome level. In contrast, the devastating surges in experiments can simulate many of the conditions
1902 may have resulted from rapid ascent relevant to magma eruption, because realistic
(i.e. <3 days) that did not allow crystallization in pressures, temperatures and rates of decompres-
the conduit, followed by extensive microlite sion or shear are now accessible in the laboratory.
crystallization at dome level (due to large effective The one parameter that remains impossible to
undercooling). The exsolving gas and high over- simulate is an extended timescale. Experiments
pressurization resulting from this extensive crys- last typically a maximum of weeks, or occasion-
tallization may have triggered the violent surges ally months. In practical terms, this leads to small
(Martel 2012). crystal sizes compared with nature, occasional
Experimental Simulations of Magma Storage and Ascent 109

difficulties in establishing equilibrium, or very Hammer JE, Cashman KV, Hoblitt RP, Newman S (1999)
short diffusion profiles for controlled ‘disequi- Degassing and microlite crystallization during
pre-climactic events of the 1991 eruption of Mt.
librium’ experiments. However, as new analytical Pinatubo Philippines. Bull Volcanol 60:355–380
techniques are developed we can start to make Hammer JE, Rutherford MJ (2002) An experimental
nanoscale measurements that allow us to measure study of the kinetics of decompression-induced crys-
profiles developed on laboratory timescale, pro- tallization in silicic melts. J Geophys Res 107
(ECV8):1–23
viding access to ever faster natural processes Holland T, Blundy JD (1994) Non-ideal interactions in
(Saunders et al. 2014; Lloyd et al. 2014). calcic amphiboles and their bearing on amphibole-
Development of in situ observation or mea- plagioclase thermometry. Contrib Miner Petrol 116(4):
surement techniques represents a major step 433–447
Kilgour GN, Blundy JD, Cashman KV, Mader HM
forward in the understanding of the magmatic (2013) Small volume andesite magmas and melt-mush
processes linked to volcanic eruptions. It is interactions at Ruapehu, New Zealand: evidence from
becoming possible to make in situ observations melt inclusions. Contrib Minerology Petrol 166
using cameras coupled to HP-HT vessels equip- (2):371–392
Kushnir ARL, Martel C, Champallier R, Arbaret L (2017)
ped with transparent windows (e.g. Gondé et al. In situ confirmation of permeability development in
2011) or 4D in situ X-ray tomography (e.g. shearing bubble-bearing melts and implications for
Pistone et al. 2015). HP-HT vessels coupled with volcanic outgassing. Earth Planet Sci lett 458:315–326
in situ analytical techniques are already capable Lacroix A (1904) La Montagne Pelée et ses éruptions.
Masson Paris, p 662
of measuring the volatile species either dissolved Lindsley DH, Andersen DJ (1983) A two-pyroxene
in the melt under pressure by in situ spectroscopy thermometer. J Geophys Res 88(S02):001. doi:10.
techniques (Raman or Infrared) or exsolved as 1029/JB088iS02p0A887
Lloyd AS, Ruprecht P, Hauri EH, Rose W, Gonner-
vapour. Laboratory simulation of magma degas-
man HM, Plank T (2014) NanoSIMS results from
sing and crystallization with these new in situ olivine-hosted melt embayments: magma ascent rate
approaches will allow the identification of during explosive basaltic eruptions. J Volcanol Geoth
potential geophysical and geochemical signals Res 283:1–18
Mangan MT, Sisson TW, Hankins WB (2004) Decom-
that may be used as unrest precursors.
pression experiments identify kinetic controls on
explosive silicic eruptions. Geophys Res Lett 31:
L08605. doi:10.1029/2004GL019509
References Martel C (2012) Eruption dynamics inferred from micro-
lite crystallization experiments: application to Plinian
and dome-forming eruptions of Mt. Pelée (Martinique,
Blundy JD, Cashman KV (2008) Petrologic reconstruc- Lesser Antilles). J Petrol 53:699–725
tion of magmatic system variables and processes. Martel C, Iacono-Marziano G (2015) Bubble coalescence,
Miner Inclusions Volcanic Proc 69:179–239 outgassing, and foam collapsing in decompressed
Blundy JD, Cashman KV, Rust AC, Witham F (2010) A rhyolitic melts. Earth Planet Sci Lett 412:173–185
case for CO2-rich arc magmas. Earth Planet Sci Lett Melnik OE, Blundy JD, Rust AC, Muir DD (2011)
290(3–4):289–301 Subvolcanic plumbing systems imaged through crystal
Costa F, Morgan D (2010) Time constraints from size distributions. Geology 39(4):403–406
chemical equilibration in magmatic crystals. In Dos- Mollard E, Martel C, Bourdier J-L (2012) Decompression-
seto A, Turner SP, Van Orman JA (eds) Timescales of induced crystallization in hydrated silica-rich melts:
magmatic processes: from core to atmosphere. Empirical models of experimental plagioclase nucle-
Wiley-Blackwell, pp 125–159 ation and growth kinetics. J Petrol 53:1743–1766
Druitt TH, Costa F, Deloule E, Dungan M, Scaillet B Mourtada-Bonnefoi CC, Laporte D (2004) Kinetics of
(2012) Decadal to monthly timescales of magma bubble nucleation in a rhyolitic melt: an experimental
transfer and reservoir growth at a caldera volcano. study of the effect of ascent rate. Earth Planet Sci Lett
Nature 482:77–80 218:521–537
Ghiorso MS, Evans BW (2008) Thermodynamics of Okumura S, Nakamura M, Takeuchi S, Tsuchiyama A,
rhombohedral oxide solid solutions and a revision of Nakano T, Uesugi K (2009) Magma deformation may
the Fe-Ti Two-oxide geothermometer and oxygen- induce non-explosive volcanism via degassing
barometer. Am J Sci 308:957–1039 through bubble networks. Earth Planet Sci Lett
Gondé C, Martel C, Pichavant M, Bureau H (2011) In situ 281:267–274
bubble vesiculation in silicic magmas. American Pichavant M, Costa F, Burgisser A, Scaillet B, Martel C,
Mineralogy 96:111–124 Poussineau S (2007) Equilibration scales in silicic to
110 C. Martel et al.

intermediate magmas—implications for experimental Saunders KE, Blundy JD, Dohmen RG, Cashman KV
studies. J Petrol 48:1955–1972 (2012) Linking petrology and seismology at an active
Pistone M, Arzilli F, Dobson KJ, Cordonnier B, Reusser E, volcano. Science 336(6084):1023–1027
Ulmer P, Marone F, Whittington AG, Mancini L, Saunders K, Buse B, Kilburn MR, Kearns S, Blundy JD
Fife JL, Blundy JD (2015) Gas-driven filter pressing in (2014) Nanoscale characterisation of crystal zoning.
magmas: insights into in-situ melt segregation from Chem Geol 364:20–32
crystal mushes. Geology 43:699–702 Scaillet B, Pichavant M (2003) Experimental constraints on
Putirka KD (2008) Thermometers and barometers for volatile abundances in arc magmas and their implica-
volcanic systems. Rev Mineral Geochim 69:61–120 tions for degassing processes. In: Oppenheimer C,
Ridolfi F, Renzulli A (2012) Calcic amphiboles in Pyle D, Barclay J (eds) Volcanic degassing. Geological
calc-alkaline and alkaline magmas: thermobarometric Society, London, Special Publications 213, 23–52
and chemometric empirical equations valid up to Streck MJ (2008) Mineral textures and zoning as evidence
1130 °C and 2.2 GPa. Contrib Mineral Petrol for open system processes. Rev Mineral Geochem
163:877–895 69:595–622
Riker JM, Blundy JD, Rust AC, Botcharnikov RE, Tamic N, Behrens H, Holtz F (2001) The solubility of
Humphreys MCS (2015) Experimental phase equilib- H2O and CO2 in rhyolitic melts in equilibrium with a
ria of a Mount St. Helens rhyodacite: a framework for mixed CO2-H2O fluid phase. Chem Geol 174:
interpreting crystallization paths in degassing silicic 333–347
magmas. ContribMineral Petrol 170(6):535–560. Waters LE, Lange RA (2015) An updated calibration of
doi:10.1007/s00410-015-1160-5 the plagioclase-liquid hygrometer-thermometer appli-
Rutherford MJ, Gardner JE (2000) Rates of magma cable to basalts through rhyolites. Am Miner
ascent. In: Sigurdsson H (ed) Encyclopedia of volca- 100:2172–2184
noes. Academic San Diego, California, pp 207–217

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Magma Chamber Rejuvenation:
Insights from Numerical Models

C.P. Montagna, P. Papale, A. Longo and M. Bagagli

Abstract
Most volcanic systems on Earth are characterized by chemically different
magmas that can be found in the erupted products throughout their history.
The reasons are multiple, including variations in the mantle source and/or
crustal assimilation, as well as shallower processes such as fractional
crystallization or mixing and mingling. Magma chamber rejuvenation
indicates the processes that happen whenever a magma intrudes from the
mantle to shallower depths and encounters an already established storage
zone (i.e. a magma chamber or reservoir). Magmas rising from depth are
typically characterized by higher temperatures, larger volatile contents and
more primitive, mantle-like compositions than those residing in the
shallow crust. The interaction with magmas that have already resided at
shallower depths for a while (years to thousands of years) varies the
physical and chemical properties of both the involved magmatic
end-members. Typically, volatile-rich magmas coming from depth are
lighter than degassed shallow magma; therefore, a gravitational instability
sets in as the two come into contact, which generates convection and thus
intense mingling and mixing among the two. These dynamic interactions
cause variations in the physical and chemical properties of the magmas
themselves, as well as in the stress conditons both inside the reservoir and
in the host rock. The volcanic system as a whole enters an unrest scenario,
that can evolve to eruption or not depending on the specific conditions.
Numerical simulations of the dynamics within magmatic systems can shed
light on the features of magma chamber rejuvenation, providing the time

C.P. Montagna (&)
P. Papale
A. Longo

M. Bagagli
Istituto Nazionale di Geofisica e Vulcanologia, via
U. della Faggiola 32, Pisa, Italy
e-mail: [email protected]
Present Address:
M. Bagagli
D-ERDW, ETH-Zuerich, Zürich, Switzerland

Advs in Volcanology (2019) 111–122

DOI 10.1007/11157_2017_21
© The Author(s) 2017
Published Online: 17 August 2017
112 C.P. Montagna et al.

scales of mixing processes and possibly of the evolution towards eruption.

Coupling with models for the visco-elastic response of the host rock
allows the identification of the onset of recharge processes from the
analysis of geophysical signals observed at the surface.

Keywords
Magma chamber
Magma dynamics
Magma mixing

having a higher liquid density: they are charac-

1 Extended English/Spanish
terized by a much larger gas content. The light
abstract
magma tends to rise inside the denser reservoir; a
gravitational instability sets in as the two mag-
Most volcanic systems on Earth are characterized
matic mixtures come into contact, and generates
by chemically different magmas that can be found
convection inside the reservoir. As a conse-
in the erupted products throughout their history,
quence, intense mingling and mixing are gener-
either in synchronous eruptive episodes, or in
ated among the two end-members. These dynamic
different epochs of volcanic activity. This chem-
interactions cause variations in the physical and
ical heterogeneity can have multiple reasons, and
chemical properties of the magmas themselves,
it originates from deep in the mantle, due to
that loose their identity as initial end-members and
variations in the source and/or crustal assimilation
become a more homogeneous mixture. The vol-
during ascent, to shallower crustal regions, where
canic system as a whole enters an unrest scenario,
processes such as fractional crystallization or
that can evolve to eruption or not depending on
mixing and mingling take place. Magma chamber
the specific conditions. Numerical simulations of
rejuvenation comprises some of the aforemen-
a magmatic system representing magma injection
tioned processes at shallow level. Whenever a
into a shallow reservoir show that mixing is very
magma intrudes from the mantle to shallower
intense at the time of contact, and can be efficient
depths and encounters an already established
on time scales of hours to day in homogeneizing
storage zone (a magma chamber or reservoir), the
the system. Depending on the geometry of the
magma already emplaced gets rejuvenated by the
volcano feeding system, and even more on the
incoming more primitive magma. Magmas rising
volatile content of the incoming and resident
from the deep regions of a volcano feeding system
magmas, the process can be suppressed or
are typically characterized by higher tempera-
enhanced. Sills favour mixing, while more verti-
tures, larger volatile contents and more primitive,
cally elongated, dike-like reservoirs slow the
mantle-like compositions. On the other hand,
dynamical interactions. As the presence of a
magmas that have resided at shallower depths for
gaseous phase is the engine of the gravitational
a while (years to thousands of years) have evolved
instability that triggers the dynamics, a higher
by fractional crystallization, thus they have
volatile content, which translates into a higher gas
changed their composition towards a more felsic
content, in the deep regions of the feeding system
one, and have lost most of their gaseous phase,
strongly accelerates the rejuvenation process. As
that can escape towards the surface. The interac-
mixing patterns are found almost ubiquitously in
tion between primitive and evolved magmas
products from volcanoes around the world, com-
varies the physical and chemical properties of
parison of the observed features to the model
both the end-members involved. Typically,
predictions can provide insights on the features of
volatile-rich magmas coming from depth are
magma chamber rejuvenation, including the time
lighter than degassed shallow magma, albeit
Magma Chamber Rejuvenation: Insights from Numerical Models 113

scales over which mixing processes are efficient more fluidal, especially if injection episodes are
and possibly the timings for the evolution towards frequent (Voight et al. 2010), giving rise to mixing
an eruption or not. Coupling to models that and mingling phenomena (Montagna et al. 2015).
describe the visco-elastic response of the host The interaction among the deep and shallow
rock to stress variations within the magmatic components changes the physical and chemical
system provides hints as to how to identify properties of both the involved magmatic
recharge processes at depth from the analysis of end-members, triggering an unrest phase that can
geophysical signals observed at the surface. evolve to eruption or not depending on the specific
Characteristic features of ground deformation conditions. Evidence of chamber rejuvenation
associated to convection and mixing is the both in igneous and in intrusive rocks, manifested
appearance of oscillation of exremely long period, mostly by mingling and mixing patterns, is almost
on the order of hours (Ultra-Long-Period, ULP), ubiquitous at volcanic systems worldwide, and it
that can be detected by instruments such as con- is often invoked as eruption trigger.
tinuous tiltmeters and dilatometers. Their records Magma movement at depth implies mass
can identify the onset of the interaction among re-distribution, pressure changes, and pressure
different magmas, thus provide time scales for transients which translate into variations in the
unrest duration and evolution. gravity field, shape and slope of the volcano flanks,
and seismic signals registered at the surface.
Understanding the complex relationships between
2 Introduction quantities measured by volcano monitoring net-
works and shallow magma processes is a crucial step
Magmas evolve in many ways during their resi- for the comprehension of volcanic processes and in
dence time within the crust, determining whether evaluating more realistic hazard forecast. The ability
they are going to be erupted or not. Magma to detect the onset of magma recharge at depth is
chamber rejuvenation takes place whenever a fundamental as it can provide hints to unrest dura-
magma intruding from the mantle to shallower tion and evolution, and possibly eruption timings.
depths encounters an already established storage In this work we describe a forward-modeling
zone (i.e. a magma chamber or reservoir). It can approach to describe magma chamber dynamics,
involve many different processes such as reheat- specifically for what concerns rejuvenation epi-
ing and melting of the residing magmas, fractional sodes, and link it to the geophysical observables
crystallization due to changes in the pressure and that are expected as a consequence. This provides
temperature conditions, mingling and mixing a framework for the consistent interpretation of
among the different components; typically, it geological and geophysical records of unrest
takes place at shallow crustal depths. Magmas periods at active volcanoes. This methodology
rising from depth are often characterized by allows for identification of rejuvenation episodes
higher temperatures, larger volatile contents and in ground deformation records, and possibly
more primitive, mantle-like compositions with discrimination between those episodes that lead
respect to those that have been residing at shal- to eruption or not.
lower levels for a while (months, years to thou-
sands of years). This general scenario can have a
variety of declinations, depending on the specific 3 Numerical Simulations of Magma
setting and physico-chemical characteristics of Chamber Rejuvenation
the magmatic mixtures involved. The shallow
magma can be highly crystalline, a mush, that can 3.1 Magmatic System
be rejuvenated by the heat from the incoming
component (Bachmann and Bergantz 2003, 2008; We refer as an archetypal case to the Phlegraean
Girard and Stix 2009; Bain et al. 2013; Till et al. Fields magmatic system, where seismic imaging
2015); or, at the other end, it can still be hot and and attenuation tomographies have identified a
114 C.P. Montagna et al.

huge (probably around 10 km wide) magma show that deep magmas are typically rich in gas,
reservoir at a depth of around 8 km (Zollo et al. especially CO2 (Mangiacapra et al. 2008), while
2008; De Siena et al. 2010), while a variety of shallow magmas are unusually crystal-poor,
geophysical and geochemical evidence suggests down to less than 3 wt% (Arienzo et al. 2009). To
that smaller (probably less than 1 km3), shallower study the magmatic dynamics occurring as a
batches of magma have been forming throughout consequence of a recharge event, we simplify the
the caldera history at virtually any depth smaller magmatic system retaining its most peculiar fea-
than 9 km (Arienzo et al. 2010; Di Renzo et al. tures. We model the injection of CO2-rich
2011). These shallow magma bodies have been shoshonitic magma coming from a deep reservoir
identified as actively involved in past eruptions, into a shallower, much smaller chamber, con-
which at least in some cases shortly followed the taining more evolved and partially degassed
arrival of volatile-rich, less differentiated magmas phonolitic magma (see Table 1 for composi-
from the deep feeding system (Arienzo et al. tions). The two chambers are connected by a
2009; Fourmentraux et al. 2012). Chemical dyke. This idealized layout captures several
compositions of erupted magmas range from first-order characteristics of prototype magmatic
shoshonitic to trachytic to phonolitic; geochemi- systems, including a composite structure, vertical
cal analyses on melt inclusions suggest a variety extension, and heterogeneous composition, and it
of processes contributing to this variability, such approximates systems composed by long-lived,
as recharge from depth, intra-chamber mixing, interconnected multiple reservoirs believed to
syn-eruptive mingling (Arienzo et al. 2010; exist at many active volcanoes (Elders et al.
Fourmentraux et al. 2012). The same analyses 2011).

Table 1 Composition of the phonolite and shoshonite magma types employed in the simulations
Composition SiO2 TiO2 Al2O3 Fe2O3 FeO MnO MgO CaO Na2O K2O
(wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%)
Phonolite 53.5 0.6 19.8 1.6 3.2 0.1 1.8 6.8 4.7 7.9
Shoshonite 52.5 0.9 17.6 1.9 5.7 0.1 3.6 7.9 3.4 4.3

Fig. 1 Initial conditions for

the numerical simulations of
the magmatic system. On the
left, the whole domain is
shown, indicating the two
magmatic end-members. On
the right, the upper portion of
the domain shows the three
different geometries explored
Magma Chamber Rejuvenation: Insights from Numerical Models 115

Figure 1 shows the system domain for the turn depends on pressure, thus on the depth at
numerical simulations. We assume one of the which the interface is placed, which is different
horizontal dimensions of the magmatic system to for each geometry of the shallow chamber
be much larger than the other, so that our domain (Fig. 1). Temperature differences between inter-
is two-dimensional. The deep chamber is ellipti- acting magmas are often negligible (Sparks et al.
cal, 1 km thick and 8 km wide; its top is at 8 km 1977), particularly at Phlegraean Fields (Man-
depth. The geometry of the shallow chamber has giacapra et al. 2008; Arienzo et al. 2010), thus
been varied as shown in Fig. 1, keeping its sur- the system is assumed isothermal. As a result,
face area fixed. In the elliptical cases, the there is no need to speculate on the thermal status
semi-axes measure 400 and 800 m, respectively, of the surrounding rock, thus reducing model
while the circular chamber has a radius of 283 m. uncertainties. Moreover, heat transfer effects are
The initial conditions of the system are also expected to play a minor role on the short sim-
shown in Fig. 1. The shallow chamber hosts a ulated time scales (hours; Di Renzo et al. 2011).
differentiated, volatile-poor phonolitic magma.
Its volatile content has been varied from 0.3 wt%
CO2 and 2.5 wt% H2O to 0.1 wt% CO2 and 1 wt 3.2 Magma Dynamics
% H2O. In the feeding dyke and deep reservoir is
a less evolved, basaltic shoshonite, containing 1 Interaction among the two magmas develops as a
wt% CO2 and 2 wt% H2O. Such a low water consequence of the initial gravitational instability
content in the phonolitic end-member derives at the interface. We solve numerically the
from melt inclusion data from Phlegrean Fields two-dimensional space-time evolution of the
(Arienzo et al. 2010). Typically, more evolved system, consisting of a mixture of two different
magmas are expected to have a relatively larger magmatic components, each of them including a
water content (Signorelli et al. 2001; Cannatelli liquid (silicate melt and dissolved volatiles) and a
et al. 2007; Pappalardo et al. 2007; Mollo et al. gaseous (exsolved volatiles) fractions. The equa-
2015), resulting possibly in smaller density tions of motion for the mixture express conser-
contrasts at the interface among the two magmas vation of mass for each component k = 1, 2, and
thus less efficient mixing dynamics. momentum for the whole mixture (Longo et al.
Volatiles partitioning between gaseous and 2012a):
liquid phases is computed following Papale et al.
(2006) as a function of composition and pressure.
Pressure at time 0 consists of a depth-dependent

magmastatic contribution superimposed to the In the above, t is time; q is mixture density,

host rock confining pressure. The interface is mass fraction of component k, u is fluid
between the two magmas, at the inlet of the velocity, Dk is the k-th coefficient of mass dif-
shallow chamber, is gravitationally unstable, the fusion, p is pressure, l is viscosity and g is
lower magma being less dense due to its higher gravity acceleration.
gas content. The dynamics is solely driven by The magmatic mixture is considered ideal. Its
buoyancy, without any external forcing. density is evaluated as weighted sum of the
The density contrast at the interface varies for components’ densities; for each component,
each simulated scenario, as it depends on both density is calculated using a non-ideal equation
volatiles content and their partitioning between of state for the liquid phase, real gas properties
liquid and gaseous phases; volatiles exsolution in and ideal mixture laws for multiphase fluids.
116 C.P. Montagna et al.

Mixture viscosity is computed through standard composition in time in the shallow chamber for
rules of mixing for one phase mixtures and with a the five different simulation scenarios.
semi-empirical relation in order to account for The initial inverse density contrast at the
the effect of non-deformable gas bubbles. Liquid contact interface between the two magmas gives
viscosity is modeled as in Giordano et al. (2008), rise to convective mass transfer from the deeper
and it depends on liquid composition and dis- parts of the system to shallower depths and vice
solved water content. The assumption of New- versa. The unstable density contrast is solely due
tonian rheology is justified by the very low strain to the different volatile content of the two mix-
rates and the crystal-free nature of the magmas. tures: the shoshonitic melt has an higher density
The generalized Fick’s law is used to describe than the phonolitic. The role played by volatiles
mass diffusion. Volatile partitioning between is crucial, and it is exsolved gases that ultimately
gaseous and liquid phases is evaluated at every determine the buoyant dynamics.
point in the space-time domain as function of A Rayleigh-Taylor instability develops, which
mixture composition and pressure as in Papale acts to bring the system to gravitational equilib-
et al. (2006). All the physical properties of the rium by overturning it. The instability develops
two magmas are evaluated at every point in the starting from the perturbed interface, with a first
space-time domain depending on the local con- plume of light material that rises into the cham-
ditions of pressure, velocity and mass fractions, ber. Depending on the initial density contrast as
which are the unknowns in Eqs. (1) and (2). The well as on the geometry of the shallow chamber,
equations are solved numerically using GALES, the initial plume starts developing at different
a finite element C++ code specifically designed times. The dynamics is strongly enhanced by
for volcanic fluid dynamics (Longo et al. 2012a). higher density contrasts; geometry also plays an
The evolution in space and time of the system important role when density contrasts are similar,
is complex and presents a number of interesting with horizontally elongated, sill-like chambers
features. Figure 2 summarizes the results regard- favouring convection with respect to more
ing magma dynamics, showing the evolution of dyke-like setups (see also Fig. 2).

Fig. 2 Snapshots of variation of composition with time in the shallower parts of the system for the different
simulations. Columns correspond to different simulations; rows correspond to different times
Magma Chamber Rejuvenation: Insights from Numerical Models 117

Plumes of light magma coming from depth two initial end-members, and not the pure
keep entering the shallow reservoir as discrete shoshonitic composition.
filaments, following irregular trajectories and As the dynamics proceeds, faster for higher
showing typical convective patterns. The lighter density contrasts and sill-like setups, the gas-rich
material tends to rise into the chamber, thereby mixture tends to accumulate at the top of the
decreasing more and more its density as vola- chamber, thereby originating a stable density
tiles exsolve in lower-pressure environments; on stratification that has indeed been testified at
the other hand, the denser magmatic mixture various magmatic systems (Arienzo et al. 2009).
initially residing in the chamber sinks into the The stratification is more prominent in vertically
feeder dyke, increasing its density by the elongated, dyke-like reservoirs (Fig. 2). The
reverse process of volatile dissolution at higher density profile along the vertical direction, eval-
pressures. The plumes thus progressively uated averaging along horizontal planes (Fig. 3),
increase their buoyancy, enhancing their illustrates that a quasi-stable profile is reached
expansion and acceleration. During the rise, after some hours of simulated time.
vortexes form at the head of the plumes and As time proceeds, convection slows down due
subsequent plumes interact among themselves, to smaller buoyancy of the incoming already
further favouring mixing. The dynamics creates mixed component, and the instability proceeds in
complicated patterns that maximize the interac- time asymptotically: the more the two
tion among the two different magmatic mixtures end-members have mingled, the less intense is
(Petrelli et al. 2011). Mingling is evident for all convection.
simulated conditions both within the chamber The evolution of pressure in the system is
itself and even more in the feeding dyke highly heterogeneous in space and time. Alter-
(Fig. 2), and it is strongly intensified by the nating phases dominated by buoyancy and sink-
chaotic patterns that form as a consequence of ing at chamber inlet result in pressure fluctuations
deep magma injection. with periods of hundreds of seconds and ampli-
Independently from system geometry or den- tudes decreasing with time (Fig. 4). Typically
sity contrast at the interface, mingling is very pressure variations are smaller than 1 MPa; under
efficient in the feeding dyke, more than inside the these conditions, it is unlikely that rejuvenation
upper chamber. Figure 2 shows that since the can trigger eruption, as the stresses needed to
very beginning of the simulations, the magma create a pathway to the surface in the host rock are
entering the chamber is already a mixture of the typically larger than that (Gudmundsson 2006).

Fig. 3 Total mixture density averaged over horizontal black line represents the quasi-equilibrium density profile
planes as function of depth for simulation 1, at different at the end of the simulation
times. The inset shows the upper 5 km of the domain; the
118 C.P. Montagna et al.

Fig. 4 Pressure variations as a function of time at a point pressure at current time and at time zero, while the bottom
on the boundary of the upper chamber, for simulation #1. diagram shows the same quantity after subtraction of a
The upper diagram shows the difference between the local detrending function (red curve in the upper diagram)

3.3 Ground Deformation ing, over all sources, the Green’s functions
associated with individual sources.
Determining the time–space-dependent ground Continuity of pressure and stress is taken as
displacement requires modeling the magma– the boundary condition along the non moving
rocks boundary conditions and the mechanical magma–rock interface. Physical properties of
response of rocks, the latter depending on rocks are homogeneous averages that describe the
heterogeneous rock properties, presence and volcanic edifices within the range of considered
distribution of faults, interfaces, fluids, and vol- depths (<10 km, vP = 3000 m/s; vP/vS = 1/√3,
cano topography (e.g., O’Brien and Bean 2004). q = 2500 kg/m3).
A first-order analysis performed here assumes Propagation of pressure disturbances in the
magma–rock one-way coupling and adopts the host rock medium reveals that the computed
Green’s functions formulation for a homoge- pressure oscillations, originated by the ingression
neous, infinite medium (Aki and Richards 2002). of buoyant magma in the magma chamber,
We consider as point sources the fluid translate into Ultra Long Period ground displace-
dynamics computational grid nodes located at the ment dynamics with amplitudes of millimeter to
reservoir walls. As source time functions, we use micrometer order (Fig. 5; Longo et al. 2012b).
the respective temporal evolutions of magmatic ULP ground movements like those predicted by
forces computed from pressures and stresses the present modeling could not be detected by
provided at those nodes by the numerical simu- classical broadband seismometers (although more
lations of magma convection and mixing recent seismometers extend their working range
dynamics. Ground displacement at a series of up to 100–200 s periods), while they are visible in
virtual receivers is finally obtained by integrat- the records from other instruments, especially
Magma Chamber Rejuvenation: Insights from Numerical Models 119

Fig. 5 Synthetic seismic

signal (blue detrended as to
represent an instrumental
record, black filtered
[0.001,0.01] Hz) and
corresponding frequency
spectrum for the vertical
component at all synthetic
stations. Example from
simulation #1

borehole dilatometers characterized by high duration: indications of mixing in erupted prod-

signal-to-noise ratio (Sacks et al. 1971). ucts suggest that recharge events can happen
within a very short time frame from eruption,
otherwise the evidence would be wiped out by the
4 Discussion and Conclusions efficient mixing process (Montagna et al. 2015).
In terms of geophysical observables, convec-
The arrival of fresh magma into an already tive mingling dynamics in magmatic reservoirs is
emplaced reservoir and the consequent internal associated with ultra-long-period seismic signals,
dynamics have often been invoked as possible characterized by frequencies in the range
eruption triggers, especially at Phlegraean Fields 10−2 − 10−4 Hz (Longo et al. 2012b).
(Arienzo et al. 2010); on the other hand, foot- The results obtained by our modeling specif-
prints of magma chamber rejuvenation are often ically refer to the Phlegraean system. They can
found in intrusive granites as well, testifying that be extended to many other volcanoes where
it does not necessarily lead to eruption. Magma evidence of rejuvenation has been observed in
chamber rejuvenation can thus be regarded as a similar magmatic settings, characterized by rel-
prototypical volcanic unrest process, that can atively primitive magmas that are not very dif-
lead to eruption or not depending on the specific ferent in composition and temperature, such as
system conditions (e.g. host rock compliance, e.g. Mount Etna (Viccaro et al. 2006) or Strom-
volume and volatile content of injected magma). boli (La Felice et al. 2011). For more evolved
The time scales for rejuvenation processes to magmas, reservoirs can be dominated by
be effective in magmatic reservoirs are relatively crystal-rich regions (Marsh 1981; Koyaguchi and
short, on the order of hours. This is consistent Kaneko 1999; Bachmann and Bergantz 2004;
with what has been observed from the analyses of Hildreth 2004; Huber et al. 2009; Cooper and
erupted products (Fourmentraux et al. 2012) as Kent 2014), and they are often at a lower tem-
well as from experiments on diffusive fractiona- perature. The approach described above must be
tion (Perugini et al. 2010), and opens a com- applied with caution in such cases, as the
pletely new perspective in terms of unrest dynamics of the incoming primitive magma is
120 C.P. Montagna et al.

more likely to be described as flow through a Its composition is characterized by higher silica
porous medium (mush) than as fluid mingling content.
and mixing. Nonetheless, there is some evidence Convection Exchange of mass and energy by
for crystal-poor silicic magma reservoirs to be means of cell patterns.
reactivated as well (Bachmann et al. 2002;
Green’s functions Source to receiver transfer
Deering et al. 2011; Huber et al. 2012; Sliwinski
function; in this context through the volcanic
et al. 2015; Wolff et al. 2015).
rock medium.
Given the short time scales over which the
dynamical processes described here can be effec-
tive and lead to eruption, it would be beneficial to
be able to routinely detect the signals described Index
above for eruption forecasting and mitigation
actions. This is especially true for long-dormant Volcanic unrest
volcanoes such as Phlegraean Fields, one of the Magma chamber dynamics
highest-risk volcanic areas in the world given the Magma mingling: magma mixing
large population living within the caldera borders Eruption precursors
(Arienzo et al. 2010), for which there is still no Volcanic unrest duration
widely accepted means of discriminating the Magma evolution
precursors of an impending eruption (Druitt et al. Ground deformation
2012). Ultra-Long-Period seismicity
Volcano seismicity
Acknowledgements This work has received funds from
the European Union’s Seventh Programme for research,
technological development and demonstration under grant
agreements No. 282769 VUELCO and No. 308665
MED-SUV. The manuscript has largely benefited from References
reviews by Fabio Arzilli and Olivier Bachmann.
Aki K, Richards PG (2002) Quantitative seismology.
University Science, Sausalito
Arienzo I, Civetta L, Heumann A, Woerner G, Orsi G
Glossary (2009) Isotopic evidence for open system processes
within the Campanian Ignimbrite (Campi Flegrei—
Magma chamber A storage volume for magma Italy) magma chamber. Bull Volcanol 71(3):285–300
in the crust. Typically, during their rise from Arienzo I, Moretti R, Civetta L, Orsi G, Papale P (2010)
The feeding system of Agnano—Monte Spina erup-
the mantle magmas accumlate in regions tion (Campi Flegrei, Italy): Dragging the past into
where there are geological discontinuities. present activity and future scenarios. Chem Geol 270
Rejuvenation The process by which magmas (1–4):135–147
Bachmann O, Bergantz GW (2003) Rejuvenation of the
coming from depth modify chemical and Fish Canyon magma body: a window into the
physical properties of more evolved magmas evolution of large-volume silicic magma systems.
already emplaced at shallower levels. Geology 31(9):789–792
Bachmann O, Bergantz GW (2004) On the origin of
Primitive magma A magmatic melt that has crystal-poor rhyolites: Extracted from batholithic
composition similar to that of the mantle crystal mushes. J Petrol 45:1565–1582
(smaller silica content). Bachmann O, Bergantz GW (2008) The Magma Reser-
voirs That Feed Supereruptions. Elements 4(1):17–21
Evolved magma A magmatic melt that has Bachmann O, Dungan MA, Lipman PW (2002) The Fish
undergone processes within the crust that Canyon magma body, San Juan volcanic field, Color-
modified its chemical properties, such as frac- ado: Rejuvenation and eruption of an upper-crustal
tional crystallization and crustal assimilation. batholith. J Petrol 43:1469–1503
Magma Chamber Rejuvenation: Insights from Numerical Models 121

Bain AA, Jellinek AM, Wiebe RA (2013) Quantitative Huber C, Bachmann O, Dufek J (2012) Crystal-poor
field constraints on the dynamics of silicic magma versus crystal-rich ignimbrites: a competition between
chamber rejuvenation and overturn. Contrib Min Petr stirring and reactivation. Geology 40:115–118
165(6):1275–1294 Koyaguchi T, Kaneko K (1999) A two-stage thermal
Cannatelli C, Lima A, Bodnar RJ, De Vivo B, Webster JD, evolution model of magmas in continental crust.
Fedele L (2007) Geochemistry of melt inclusions from J Petrol 40:241–254
the Fondo Riccio and Minopoli 1 eruptions at Campi La Felice S, Landi P (2011) The 2009 paroxysmal
Flegrei (Italy). Chem Geol 237:418–432 explosions at Stromboli (Italy): magma mixing and
Cooper KM, Kent AJR (2014) Rapid remobilization of eruption dynamics. Bull Volc 73(9):1147–1154
magmatic crystals kept in cold storage. Nature Longo A, Barsanti M, Cassioli A, Papale P (2012a) A
506:480–483 finite element Galerkin/least-squares method for com-
Deering CD, Bachmann O, Vogel TA (2011) The putation of multicomponent compressible in com-
Ammonia Tanks Tuff: erupting a melt-rich rhyolite pressible flows. Comp Fluids 67:57–71
cap and its remobilized crystal cumulate. Earth Planet Longo A, Papale P, Vassalli M, Saccorotti G, Mon-
Sc Lett 310:518–525 tagna CP, Cassioli A, Giudice S, Boschi E (2012b)
De Siena L, Del Pezzo E, Bianco F (2010) Seismic Magma convection and mixing dynamics as a source
attenuation imaging of Campi Flegrei: evidence of gas of Ultra-Long-Period oscillations. Bull Volcanol
reservoirs, hydrothermal basins, and feeding systems. 74:873–880
J Geophysi Res 115(B9):1–18 Mangiacapra A, Moretti R, Rutherford MJ, Civetta L,
Di Renzo V, Arienzo I, Civetta L, D’Antonio M, Orsi G, Papale P (2008) The deep magmatic system of
Tonarini S, Di Vito MA, Orsi G (2011) The magmatic the Campi Flegrei caldera (Italy). Geophys Res Lett 35
feeding system of the Campi Flegrei caldera: archi- Marsh BD (1981) On the crystallinity, probability of
tecture and temporal evolution. Chem Geol 281(3– occurrence, and rheology of lava and magma. Contrib
4):227–241 Min Petr 78:85–98
Druitt TH, Costa F, Deloule E, Dungan M, Scaillet B Mollo S, Masotta M, Forni F, Bachmann O, De Astis G,
(2012) Decadal to monthly timescales of magma Moore G, Scarlato P (2015) A K-feldspar-liquid
transfer and reservoir growth at a caldera volcano. hygrometer specific to alkaline differentiated magmas.
Nature 482(7383):77–80 Chem Geol 392:1–8
Elders WA, Friðleifsson GÓ, Zierenberg RA, Pope EC, Montagna CP, Papale P, Longo A (2015) Timescales of
Mortensen AK, Guðmundsson Á, Lowenstern JB, mingling in shallow magmatic reservoir. In: Caric-
Marks NE, Owens L, Bird DK, Reed M (2011) Origin chi L, Blundy J D (eds) Chemical, physical and
of a rhyolite that intruded a geothermal well while temporal evolution of volcanic systems. Gelogical
drilling at the Krafla volcano Iceland. Geology 39 Society, London, Special Publications 422
(3):231–234 O’Brien GS, Bean CJ (2004) A 3D discrete elastic lattice
Fourmentraux C, Metrich N, Bertagnini A, Rosi M (2012) method for seismic wave propagation in heteroge-
Crystal fractionation, magma step ascent, and neous media with topography. Geophys Res Lett 31,
syn-eruptive mingling: the Averno 2 eruption (Phle- L14608
graean Fields, Italy). Contrib Min Petr 163(6):1121– Papale P, Moretti R, Barbato D (2006) The compositional
1137 dependence of the multicomponent volatile saturation
Giordano D, Russell JK, Dingwell DB (2008) Viscosity surface in silicate melts. Chem Geol 229:78–95
of magmatic liquids: a model. Earth Planet Sci Lett Pappalardo L, Ottolini L, Mastrolorenzo G (2007) The
271:123–143 Campanian Ignimbrite (southern Italy) geochemical
Girard G, Stix J (2009) Magma recharge and crystal mush zoning: insight on the generation of a super-eruption
rejuvenation associated with early post-collapse Upper from catastrophic differentiation and fast withdrawal.
Basin Member rhyolites, Yellowstone caldera. Wyom- Contrib Min Petr 156:1–26
ing. J Petrol 50(11):2095–2125 Perugini D, Poli G, Petrelli M, Campos CP, Dingwell DB
Gudmundsson A (2006) How local stresses control (2010) Time-scales of recent Phlegrean Fields erup-
magma-chamber ruptures, dyke injections, and erup- tions inferred from the application of a diffusive
tions in composite volcanoes. Earth Sci Rev 79:1–31 fractionation model of trace elements. Bull Volcanol
Hildreth WS (2004) Volcanological perspectives on Long 72(4):431–447
Valley, Mammoth Mountain, and Mono Craters: Petrelli M, Perugini D, Poli G (2011) Transition to chaos
several contiguous but discrete systems. J Vol Geoth and implications for time-scales of magma hybridiza-
Res 136:169–198 tion during mixing processes in magma chambers.
Huber C, Bachmann O, Manga M (2009) Homogeniza- Lithos 125(1–2):211–220
tion processes in silicic magma chambers by stirring Sacks IS, Selwyn S, Evertson DW (1971) Sacks-Evertson
and mushification (latent heat buffering). Earth Planet strainmeter, its installation in Japan and some prelim-
Sci Lett 283:38–47 inary results concerning strain steps. P Jpn Acad 47
(9):707–712
122 C.P. Montagna et al.

Signorelli S, Vaggelli G, Carroll C, Romano MR (2001) event at Mount Etna (Italy): Effects on the eruptive
Volatile element zonation in Campanian Ignimbrite dynamics. J Vol Geo Res 149(1–2):139–159
magmas (Phlegrean Fields, Italy): evidence from the Voight B, Widiwijayanti C, Mattioli G S, Elsworth D,
study of glass inclusions and matrix glasses. Contrib Hidayat D, Strutt M (2010) Magma-sponge hypothesis
Min Petr 140:543–553 and stratovolcanoes: Case for a compressible reservoir
Sliwinski JT, Bachmann O, Ellis BS, Dávila-Harris P, and quasi-steady deep influx at Soufriere Hills Vol-
Nelson BK, Dufek J (2015) Eruption of shallow cano, Montserrat. Geophys Res Lett 37(19): L00E05
crystal cumulates during caldera-forming events on Wolff JA, Ellis BS, Ramos FC, Starkel WA, Boroughs S,
Tenerife. Canary Islands, J Petrol 56(11):2173–2194 Olin PH, Bachmann O (2015) Remelting of cumulates
Sparks R, Sigurdsson H, Wilson L (1977) Magma as a process for producing chemical zoning in silicic
mixing: a mechanism for triggering acid explosive tuffs: A comparison of cool, wet and hot, dry rhyolitic
eruptions. Nature 267:315–318 magma systems. Lithos 236–237:275–286
Till CB, Vazquez JA, Boyce JW (2015) Months between Zollo A, Maercklin N, Vassallo M, Dello Iacono D,
rejuvenation and volcanic eruption at Yellowstone Virieux J, Gasparini P (2008) Seismic reflections
caldera. Wyoming. Geology 43(8):695–698 reveal a massive melt layer feeding Campi Flegrei
Viccaro M, Ferlito C, Cortesogno L, Cristofolini R, caldera. Geophys Res Lett 35(12):L12306
Gaggero L (2006) Magma mixing during the 2001

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Magma Mixing: History
and Dynamics of an Eruption Trigger

Daniele Morgavi, Ilenia Arienzo, Chiara Montagna,

Diego Perugini and Donald B. Dingwell

Abstract
The most violent and catastrophic volcanic eruptions on Earth have been
triggered by the refilling of a felsic volcanic magma chamber by a hotter
more mafic magma. Examples include Vesuvius 79 AD, Krakatau 1883,
Pinatubo 1991, and Eyjafjallajökull 2010. Since the first hypothesis,
plenty of evidence of magma mixing processes, in all tectonic environ-
ments, has accumulated in the literature allowing this natural process to be
defined as fundamental petrological processes playing a role in triggering
volcanic eruptions, and in the generation of the compositional variability
of igneous rocks. Combined with petrographic, mineral chemistry and
geochemical investigations, isotopic analyses on volcanic rocks have
revealed compositional variations at different length scales pointing to a
complex interplay of fractional crystallization, mixing/mingling and
crustal contamination during the evolution of several magmatic feeding
systems. But to fully understand the dynamics of mixing and mingling

D. Morgavi
D. Perugini
Department of Physics and Geology,
University of Perugia, Piazza Università,
06100 Perugia, Italy
D. Morgavi (&)
D.B. Dingwell
Department Earth and Environmental Sciences,
Ludwig-Maximilian-University (LMU),
Theresienstrasse 41/III, 80333 Munich, Germany
e-mail: [email protected]
I. Arienzo
Istituto Nazionale Di Geofisica E Vulcanologia,
Osservatorio Vesuviano, Via Diocleziano, 328,
8124 Naples, Italy
C. Montagna
Istituto Nazionale Di Geofisica E Vulcanologia,
Via Della Faggiola, 32, Pisa 56126, Italy

Advs in Volcanology (2019) 123–137

DOI 10.1007/11157_2017_30
© The Author(s) 2017
Published Online: 05 December 2017
124 D. Morgavi et al.

processes, that are impossible to observe directly, at a realistically large

scale, it is necessary to resort to numerical simulations of the complex
interaction dynamics between chemically different magmas.

Keywords
Magma mixing
Mingling
Isotope
Modelling

processes and igneous differentiations are con-

1 Magma Mixing: A Brief Historical
ceived. The conceptual model of fractional
Overview
crystallization was firmly established as the most
fundamental petrological process for generating
One of the first investigations on magma mixing
the diversity of igneous rocks and remained so
recorded in the literature is the work of the
for many decades. Although Bowen did not
chemist Bunsen (1851), a scholar at the Univer-
explicitly deny the possibility of magma mixing,
sity of Heidelberg who published research on the
he reinterpreted field evidence of magma mixing
chemical variation of igneous rocks from the
rather as immiscibility of liquids (e.g., Bowen
western region of Iceland. Through chemical
1928). In 1944, Wilcox published a work on the
analyses Bunsen highlighted that the linear cor-
Gardner River complex (Yellowstone, USA;
relation between pairs of chemical elements in
Wilcox 1944) which is now considered a mile-
binary plots in those Icelandic rocks was the
stone for evidence of magma mixing, even if at
consequence of “simple” binary mixing between
that time it received strong comments from
two magmas with different chemical composi-
Fenner and remained one of the few papers on
tion. Bunsen published this data and, for the first
the topic. Only in the 1970’s geoscientists started
time, magma mixing was taken into account to
to deeply investigate magma mixing, recorded as
explain the chemical variation of a suite of
a plethora of unequivocal evidence in both plu-
igneous rocks. This idea triggered a strong crit-
tonic and volcanic rocks, as a major petrogenetic
ical reaction from the geological community; the
process (e.g., Eichelberger 1978, 1980; Blundy
strongest opposition coming from Wolfgang
and Sparks 1992; Wiebe 1994; Wilcox 1999).
Sartorius Freiherr von Waltershausen an expert
Since the first hypotheses about the origin of
on the Iceland and Etna volcanic areas at that
mixed igneous rocks (e.g., Bunsen 1851), plenty
time. He mostly argued against the method used
of evidence of magma mixing processes, in all
by Bunsen of averaging rock analyses to calcu-
tectonic environments, throughout geological
late the starting end-members that eventually
time, has accumulated in the literature allowing
took part in the mixing process. Sartorius criti-
this natural process to be defined as a funda-
cized not only the arbitrary choice of the
mental petrological process playing a key role in
end-members but also disliked the idea of Bun-
the generation of the compositional variability of
sen of an extensive layering of felsic/mafic rocks
igneous rocks and as a major process for plane-
and magmas beneath Iceland.
tary differentiation (e.g., Eichelberger 1978,
Since the beginning of the 20th century, the
1980; Blundy and Sparks 1992; Wiebe 1994; De
experimental and thermodynamic work of
Campos et al. 2004; Perugini and Poli 2012;
Norman L. Bowen (e.g., Bowen 1928) has had a
Morgavi et al. 2016).
profound influence upon the way petrological
Magma Mixing: History and Dynamics of an Eruption Trigger 125

2 Magma Mixing: Field Evidence The most common evidence for magma mix-
ing in igneous rocks is the occurrence of textural
It is common practice in the petrological com- heterogeneity; the processes responsible for this
munity to split magma mixing into two separate have been discussed extensively in many works
physico/chemical processes: (i) mechanical in the last decades (e.g., Eichelberger 1975;
mixing (also referred to as “magma mingling”), Anderson 1976; Bacon 1986; Didier and Bar-
by which two or more batches of magma mingle barin 1991; Wada 1995; De Rosa et al. 1996;
without chemical exchanges between them, and Ventura 1998; Smith 2000; Snyder 2000; De
(ii) a chemical mixing (also referred to as Rosa et al. 2002; Perugini et al. 2002, 2007;
“magma mixing”) triggered by chemical Perugini and Poli 2005, 2012; Pritchard et al.
exchanges between the interacting magmas in 2013; Morgavi et al. 2016).
which elements move from one magma to the In order to provide possible classification of
other according to compositional gradients con- magma mixing structures, the evidence of
tinuously generated by the mechanical dispersion mechanical mixing in igneous rocks can be
of the two magmas (e.g., Flinders and Clemens roughly divided into three different groups:
1996). Physically, “magma mingling” is mainly (i) flow structures, (ii) magmatic enclaves and
controlled by the viscosity contrast between the (iii) physico-chemical disequilibria in melts and
two magmas; decreasing of the viscosity contrast crystals (e.g., Walker and Skelhorn 1966; Didier
results in progressively more efficient mingling and Barbarin 1991; Hibbard 1995; Flinders and
dynamics (e.g., Sparks and Marshall 1986; Clemens 1996; Wilcox 1944, 1999; Perugini
Grasset and Albarede 1994; Bateman 1995; Poli et al. 2002, 2003; Streck 2008; Perugini and Poli
et al. 1996; Perugini and Poli 2005). Chemically, 2012; Morgavi et al. 2016). Flow structures can
“magma mixing” is driven by the mobility of be readily recognized in field outcrops as they
chemical elements in the melt fractions of the show alternating light and dark coloured bands
two magmas (e.g., Lesher 1990; Baker 1990). constituted by magmas with different composi-
Linear variations in inter-elemental plots for a set tions. Figure 1a, b shows some examples of fluid
of rock samples have long been considered as the structures occurring in volcanic rocks from
sole evidence for the occurrence of magma Grizzly Lake outcrop in Yellowstone National
mixing (e.g., Fourcade and Allegre 1981). Park (USA) (Pritchard et al. 2013; Morgavi et al.
The adoption of the above conceptual models 2016) and from Soufrière Hills volcano (Island
led to the common practice of applying the term of Montserrat, UK) (Plail et al. 2014). In partic-
magma mingling to indicate the process acting to ular, Fig. 1a shows flow bands of rhyolitic
physically disperse (no chemical exchanges are magma (white) intruding in a basaltic/hybrid
involved) two or more magmas, whereas the term magma (red to dark grey) whereas Fig. 1b shows
magma mixing indicates that the mingling pro- an alternation of flow bands of rhyolitic (white)
cess is also accompanied by chemical exchanges. and hybrid magma (dark grey) across which
Although such a conceptual approach may allow basaltic enclaves (light grey) occur. The latter are
us to simplify the complexity of the magma surrounded by flow bands of hybrid filaments
mixing process and make it more tractable from (blue).
the petrological point of view, unfortunately such Magmatic enclaves are probably the structural
terminology is not consistently used in the liter- evidence that, according to common thinking,
ature and this causes some misunderstanding. mostly characterize magma mixing processes.
Although it is not always easy to clearly dis- The term magmatic enclave is used to identify a
criminate between the two processes, mingling is discrete portion of a magma occurring within a
quite a rare process in nature as physical dis- host magma with a different composition (e.g.,
persion and chemical exchanges must occur in Wilcox 1944; Walker and Skelhorn 1966; Bacon
tandem during magma mixing processes (e.g., 1986, Didier and Barbarin 1991). Generally,
Wilcox 1999; Perugini and Poli 2012). enclaves display quite sharp contacts with the
126 D. Morgavi et al.

(a) Hybrid
(b)

Rhyolite

Basaltic enclaves
Basaltic enclaves

(c) Basaltic enclaves (d)

Andesitic host
Andesitic host

Basaltic enclaves

Basaltic enclaves

Fig. 1 Detailed images of the mixing features present at right. Figure 1b shows the stretching and folding of
Grizzly Lake (Yellowstone) (a, b) and Soufrière Hills hybrid magma (dark grey) into a rhyolitic portion (white)
volcano (Montserrat) (c, d). Figure 1a shows the rhyolitic with the presence of several basaltic enclaves. Figure 1c
magma (white) has apparently intruded into the basaltic shows a basaltic enclave surrounded by andesitic magma
magma and the hybrid portions are present at the contact (Soufrière Hills, from the 2010 eruption). Figure 1d
between the two end-members. Two large basaltic exhibits at the centre a basaltic enclave in and andesitic
enclaves are visible at the bottom left and at the centre host from Soufrière Hills volcano

host rock, although it is not rare to observe that to reconstruct the time evolution of thermal and
some enclaves display engulfment and disruption compositional exchanges between the two mag-
of their boundaries due to infiltration of the host mas during mixing. Recent studies highlighted
magma. Some examples of enclaves found in the the importance of detailed investigations of
volcanic rocks from Soufrière Hills are shown in crystal compositional variability not only to
Fig. 1c, d. reconstruct the fluid-dynamic regime governing
Disequilibrium textures in minerals the evolution of the igneous body, but also to
(Fig. 2a–c) can be viewed as recorders of the understand the length-scale of the compositional
thermal and compositional disequilibria operat- variability induced by the mixing process, the
ing in the magmatic system during the develop- latter being considered as a proxy to estimate the
ment of magma mixing processes. As the zoning residence time of magmas in sub-volcanic
pattern can be well preserved in minerals from reservoirs prior to eruption (e.g., Costa and
both the plutonic and volcanic rocks, crystal Chakraborty 2004; Martin et al. 2008; Cham-
populations from both environments can be used berlain et al. 2014; Perugini et al. 2003).
Magma Mixing: History and Dynamics of an Eruption Trigger 127

(a) (b)

Yellowstone Mixing Montserrat Mixing

Andesite
Hybrid

Rhyolite
Basalt

Basalt

500 µm 500 µm

(c)

Sanidine

Phlegrean fields Mixing 100 µm

Fig. 2 Backscattered electron (BSE) image of a section andesitic rock. c Crystal from the 4.67 ± 0.09 cal ka
of disequilibrium textures in rock and minerals from Agnano Monte-Spina eruption (Phlegrean fields) occurred
Yellowstone, Montserrat and Phlegrean Fields. a Mixed from a vent in the Agnano-San Vito area, has a darker
rock from Grizzly Lake Complex (Yellowstone) showing (i.e., Ba-poorer) resorbed core and an inner rim with
the interaction between the basaltic portion, the hybrid “swirly” zonation textures that indicate crystallization and
portion and the rhyolitic portion. b Mixed rock section dissolution. The outer rim is characterized by small-scale
from the 2010 Soufrière Hill eruption (Montserrat) wavy oscillatory zoning that results from high-frequency
showing disequilibrium texture in the basaltic and the growth and resorption events

Laumonier et al. 2014) can provide additional

3 Numerical and Experimental tools for a better understanding of the complexity
Studies: New Ideas of the mixing process, the evolution of which is
for Deciphering the Complexity governed by a continuous exchanges. One of the
of Magma Mixing most striking results arising from these studies is
that, during mixing, chemical elements experi-
Studies focused on numerical and experimental ence a diffusive fractionation process due to the
investigation of magma mixing dynamics (e.g., development in time of chaotic mixing dynamics
Perugini et al. 2003, 2008, 2015; De Campos (Perugini et al. 2006, 2008). This process is
et al. 2004, 2008, 2011; Petrelli et al. 2011; considered the source of strong deviations in
Montagna et al. 2015; Morgavi et al. 2015; many chemical elements from the linear
128 D. Morgavi et al.

variations in inter-elemental plots that would 4 Geochemical Evidence of Magma

otherwise be expected, based on a conceptual Mixing/Mingling: An Example
model classically adopted in the geochemical from the Campi Flegrei Volcanic
modelling of magma mixing processes (e.g., Area
Fourcade and Allegre 1981; Perugini and Poli
2012 and references therein). Recent studies on In some volcanic areas chemically and isotopi-
the mineralogical and geochemical features of cally distinct magmas have been erupted, and
mixed rocks (e.g., Hibbard 1981, 1995; Wallace their composition identified by analyzing the
and Bergantz 2002; Costa and Chakraborty chemical composition of the erupted products
2004; Perugini and Poli 2005; Slaby et al. 2010), (e.g., Pantelleria (Italy), Gedemsa and Fanta ‘Ale
as well as those focused on quantitative analyses (Main Ethiopian Rift), Gorely Eruptive Center
of morphologies related to textural heterogeneity (Kamchatka); Civetta et al. 1997; Giordano et al.
(e.g., Wada 1995; De Rosa et al. 2002; Perugini 2014; Seligman et al. 2014). However, in other
and Poli 2005; Perugini et al. 2002, 2003) have volcanic complexes the majority of the erupted
highlighted the dominant role played by chaotic products are chemically rather homogeneous,
mixing dynamics in producing the substantial displaying a dominant composition (e.g., trachy-
complexity of geochemical variations and textu- basalt at Mt. Etna; trachyte at Campi Flegrei,
ral patterns found in the resultant rocks (e.g., Italy; basalt at Réunion Island, Indian Ocean;
Flinders and Clemens 1996; De Campos et al. andesite at Popocatepetl, Mexico). Despite the
2011; Morgavi et al. 2013a, b, c, 2016). Despite roughly homogenous composition of products
significant attention in the past, however, few from these volcanic areas, their isotopic features
works have focused on the understanding of the suggest that complex open system processes
relationship between the morphology of the occurred and superposed the main fractional
mixing patterns and the geochemical variability crystallization trend. In fact, isotopic analyses
of the system using experimental devices (e.g., (e.g., Sr, Nd, Pd, B) have been proven to be an
De Rosa et al. 2002). important tool for discriminating between
Based upon the combination of field obser- closed-system fractional crystallization and
vations and the outcome from numerical simu- open-system magma mixing/mingling or crustal
lations, a new experimental apparatus has been contamination (e.g., James 1982; Knesel et al.
developed to perform mixing experiments using 1999; Turner and Foden 2001). Combined with
high viscosity silicate melts at high temperature petrographic, mineral chemistry and chemical
(De Campos et al. 2011; Morgavi et al. 2013a, c, investigations, isotopic analyses on volcanic
2015). This device has been used to study the rocks have revealed compositional variations at
mixing process between natural melts, enabling different length-scales (bulk rock, minerals, single
the investigation of the influence of chaotic crystals) pointing to a complex interplay of frac-
dynamics on the geochemical evolution of the tional crystallization, mixing/mingling and crus-
system of mixing magmas (Morgavi et al. 2013a, tal contamination during the evolution of several
b, c, 2015). magmatic feeding systems (e.g., Di Renzo et al.
Preliminary results indicate that the time 2011 and references therein; Melluso et al. 2012;
evolution of compositional exchanges between Corsaro et al. 2013 and references therein; Di
magmas from the experiments can be effectively Muro et al. 2014 and references therein; Brown
modelled, leading to the prospect that the record et al. 2014). Furthermore, together with conven-
of magma mixing processes may serve as tional isotopic analyses, current technologies
chronometers to estimate the time interval permit high precision, in situ determination of Sr
between mixing and eruption (Perugini et al. isotopic ratios of portions of phenocryst and
2010; Perugini and Poli 2012; Morgavi et al. glasses. In fact, microsampling by MicroMill™,
2013a, b, c, 2015; Perugini et al. 2015). coupled with isotopic measurement by Thermal
Magma Mixing: History and Dynamics of an Eruption Trigger 129

Ionization Mass Spectrometry (TIMS), allows for the evolution of its feeding system up to eruption
the performance of high precision determination (e.g., Civetta et al. 1997; D’Antonio et al. 1999,
of Sr isotopic composition of single crystals or 2013; de Vita et al. 1999; Pappalardo et al. 2002;
portions of them. This information, unobtainable Fedele et al. 2008, 2009; Tonarini et al. 2004,
from bulk samples, has been used successfully to 2009; Arienzo et al. 2010, 2011; Perugini et al.
gather information on the time- and length-scales 2010; Di Vito et al. 2011; Melluso et al. 2012;
of the pre-eruptive magmatic processes, for Arienzo et al. 2015).
identifying mantle sources and/or magmatic In particular, detailed investigations of the
end-members and for tracking the time evolution geochemical and isotopic (Sr, Nd, Pb, and B)
of magma differentiation (e.g., Davidson et al. features of the younger than 15 ka Campi Flegrei
1990; Davidson and Tepley 1997; Davidson et al. volcanic products gave understanding to how
1998; Knesel et al. 1999; Font et al. 2008; Kin- many variable magmatic components, rising from
man et al. 2009; Francalanci et al. 2012; Braschi large depth and/or stagnating in middle crustal
et al. 2012; Jolis et al. 2013; Arienzo et al. 2015). reservoir(s), recharged the shallowest reservoir(s)
Among the active volcanic areas worldwide and interacted with magma batches left from
the volcanic hazard posed by the Campi Flegrei previous eruptions (Di Renzo et al. 2011). One
caldera is extremely high, due to its explosive identified magmatic component, geochemically
character. Both the high volcanic hazard and the similar to magma from the Neapolitan ca.
intense urbanization result in an extreme volcanic 0.70750, 143Nd/144Nd ratio of ca. 0.51247,
risk in this area, leading to a considerable interest 206Pb/204Pb of ca. 19.04 and d11B of ca.
in understanding which processes might con- −7.8‰), has been the most prevalent component
tribute to triggering of eruptions and controlling? over the past 15 ka. A second magmatic compo-
eruptive dynamics. The Campi Flegrei caldera is nent, having geochemical features similar to the
a nested and resurgent structure in the Campania Minopoli 2 magma (D’Antonio et al. 1999; Di
Region, South Italy (Orsi et al. 1996), possibly Renzo et al. 2011), first erupted 10 ka ago, is
formed after two large caldera forming eruptions: shoshonitic in composition. It is the most enriched
the Campanian Ignimbrite eruption (39 ka, in radiogenic Sr (87Sr/86Sr of ca. 0.70850) and
Fedele et al. 2008) and the Neapolitan Yellow unradiogenic Nd and Pb (143Nd/144Nd ratio of
Tuff (15 ka, Deino et al. 2004). Its magmatic ca. 0.51238, 206Pb/204Pb of ca. 18.90), and it is
system is still active as testified by the occurrence characterised by the lowest d11B value of ca.
of the last eruption in 1538 AD, as well as the −7.4‰. The third component is trachytic in
present widespread fumaroles and hot springs composition and is characterized by lower
activity, and the persistent state of unrest (Del 206Pb/204Pb (ca. 19.08), 87Sr/86Sr (ca. 0.70720)
Gaudio et al. 2010; Chiodini et al. 2003, 2012, and d11B (−9.8‰) and higher 143Nd/144Nd (ca.
2015; Moretti et al. 2013). For compositionally 0.51250), with respect to the Neapolitan Yellow
homogenous magmas such as those extruded at Tuff component (Tonarini et al. 2009; Di Renzo
the Campi Flegrei caldera (trachytes and et al. 2011; Arienzo et al. 2015). This third com-
phonolites being by far the most abundant rocks), position is known as the Astroni 6 component due
major oxide and trace element variations cannot to the fact that it best recognized in the Astroni 6
be used to unequivocally establish which magma erupted products (Di Renzo et al. 2011). During
evolution processes operated. Thus, together the past 5 ka of activity, this new component has
with petrographic, mineral chemistry and chem- been suggested to have mixed in variable pro-
ical data, isotopic investigations on volcanic portions with the Neapolitan Yellow Tuff and
rocks spanning the history of the volcano have Minopoli 2 magmatic components, which domi-
been performed in recent decades in order to nated the Campi Flegrei volcanic activity mostly
define the role of variable magmatic processes in in the time span from 15 to 5 ka (Fig. 3).
130 D. Morgavi et al.

Fig. 3 Modelling of the Sr-Nd isotopic features of some Tuff and the Minopoli 2 eruptions, respectively. Symbols
of the Campi Flegrei volcanics of the past 5 ka, by inside the plot represent volcanic products belonging to
assuming mixing among the Astroni 6 (Ast-6)-, Neapoli- the listed eruptions. The vertical error bar is the uncer-
tan Yellow Tuff (NYT)- and Minopoli 2 (Min 2)-like tainty in 143Nd/144Nd determination at the 2r level of
magmatic components. The green, yellow and black confidence; that for 87Sr/86Sr is included in the symbols.
boxes represent the range of Sr and Nd isotopes of the Modified after Arienzo et al. (2016)
products erupted during the Astroni 6, Neapolitan Yellow

Based on isotope investigations and melt volcanic hazards and mitigating the related risks
inclusions studies, Arienzo et al. (2016) sug- and (ii) the basic geochemical and petrologic
gested that the Astroni 6 component, although knowledge for the numerical simulations.
undergoing differentiation during uprising, had a
deep origin (larger than 8 km depth). Indeed,
this magma rose not only inside and along the 5 Numerical Simulation of Magma
margins of the caldera, but also at the intersection Mingling and Mixing
between SE-NW and NE-SW regional fault
systems mixing with the NYT-like magma To understand mixing and mingling processes at
component at shallower depth, and possibly a realistically large scale, it is necessary to resort
entrapping crystals accumulated during older to numerical simulations of the complex inter-
eruptions. action dynamics between chemically different
This detailed study of the Campi Flegrei vol- magmas. Referring to the archetypal case of the
canic system highlights that Sr isotopic micro- Campi Flegrei magmatic system as described
analysis and, in general, more conventional above, the interaction of a shoshonite and a more
isotopic analyses, coupled with petrographic, evolved phonolite has been investigated in detail
mineral chemistry and geochemical data can pro- to provide constraints on the time and length
vide a better knowledge of the mixing/mingling scales of the mixing dynamics. The simulated
processes and of the mixing end-members. In turn, system consists of a very large, deep (8 km)
they provide (i) information for evaluating the reservoir connected by a dike to a shallower,
Magma Mixing: History and Dynamics of an Eruption Trigger 131

Fig. 4 Initial setup for the

numerical simulations of
magma chamber
replenishment at Campi
Flegrei caldera. The two
interacting magmas are
characterized by different
compositions and volatile
contents, thus densities and
gas volume fractions

smaller chamber (Fig. 4). The chemical interac- that the two magmas mingle very efficiently
tions between the two magmas cannot be from the beginning of their interaction. As time
resolved on the simulated large scale, as the progresses, convection slows down due to
computational costs required would be too high. smaller buoyancy of the incoming mixed com-
The shoshonitic magma, being richer in volatiles ponent, and the instability proceeds in time
than the resident phonolitic melt, rises into the asymptotically: the more the two end-members
phonolitic chamber by buoyancy, generating have mingled, the less intense the convection.
convection and mixing within the reservoir and A time-dependent mixing efficiency gC can be
in the feeding dike. Both magmatic components defined as:
include a liquid (silicate melt) and a gaseous
phase, that cannot decouple from the host melt. jmR ðtÞ  mR ð0Þj
gC ¼ : ð1Þ
Space-time varying volatile exsolution is com- mR ð0Þ
puted as a function of local composition, pressure
and temperature following Papale et al. (2006). In the equation above, mR(t) is the mass of the
More details on magma chamber dynamics can resident phonolitic magma at time t, thus mR(0) is
be found in Chap. 8 ‘Magma chamber rejuve- the initial phonolite mass. The mixing efficiency
nation: insights from numerical models’ of this gC represents the relative variation of the mass of
book. the initially resident magma in a certain region of
The main results show that the chaotic pat- the domain. Figure 6 shows the time evolution of
terns observed in the products and in recently mixing efficiency in the shallow chamber, for
developed experimental setups (Morgavi et al. different simulated setups in terms of chamber
2013a, b, c, 2015) are reproducible (Fig. 5), and geometry and volatile content. It clearly shows
132 D. Morgavi et al.

Fig. 5 Magma mixing in natural samples, experimental experimental setup, from Morgavi et al. (2013b); right:
setups 684 and numerical modelling. Left: Lesvos simulations of magma chamber replenishment
(Greece) lava flow, from Perugini and Poli (2012); centre:

Fig. 6 Variation of
convection efficiency with
time in the shallow chamber
for different simulated
scenarios, characterized by
varying geometry and total
volatile content in the shallow
chamber

that mingling is very effective on relatively short

time scales, on the order of hours, in agreement 6 Magma Mixing Time Scale
with mixing timescales derived from the geo- and Eruption Trigger
chemical modelling of magma mixing experi-
ments (Perugini et al. 2010). When buoyancy The most violent and catastrophic volcanic
drives fast dynamics, more than 40% of the eruptions on Earth have been triggered by the
original end-member magmas have mingled in refilling of a felsic volcanic magma chamber by a
the feeder dyke within 4 h from the arrival of the hottest and more mafic magma (Kent et al. 2010;
gas-rich magma from depth. The asymptotic Murphy 1998). Examples include Vesuvius 79
behaviour seems to be reached earlier for less AD (Cioni et al. 1995), Krakatau 1883 (Self
efficient setups: after 4–5 h from the onset of the 1992), Pinatubo 1991 (Kress 1997), the Campa-
instability, the system seems to have reached a nian Ignimbrite (Arienzo et al. 2009) and
quasi-steady state. In these cases, a much smaller Eyjafjallajökull 2010 (Sigmundsson et al. 2010).
part of the two magmas has mingled. Injection of the more mafic magma into the felsic
Magma Mixing: History and Dynamics of an Eruption Trigger 133

magma triggers convection dynamics and wide- high velocities of ascending magmas may imply
spread mixing (Sparks et al. 1977). Vesiculation little warning time in volcanic crises. These
induced by convection increases magma pressure findings can be a starting point towards a uni-
and may fracture the volcanic edifice triggering fying model explaining chemical exchanges in
an explosive eruption. The injection and mixing magmatic systems and supplying information on
process is accompanied by geophysical signals, the use of chemical element mobility as
such as earthquakes, gravity changes and ultra- geochronometers for volcanic eruptions. This
long-period ground oscillations, that can now be may provide unparalleled clues for building an
accurately detected (Williams and Rymer 2002; inventory of past and recent volcanic eruption
Longo et al. 2012; Bagagli et al. 2017). The timescales and could be decisive for hazard
knowledge of the time elapsing between the assessment in active volcanic areas.
beginning of mixing (and associated geophysical
signals) and eruption is thus of greatest impor- Acknowledgements This research was funded by the
European Union’s Seventh Programme for research
tance in forecasting the onset of a volcanic technological development and demonstration under grant
eruption. agreement No 282759—VUELCO and by the ERC
Recent studies highlighted that in order to Consolidator Grant 612776—CHRONOS.
preserve magma mixing structures (i.e., fila-
ments, swirls, bandings) in the rocks, the time
elapsed between the beginning of mixing and the References
subsequent eruption must be very short; on the
order of hours or days (Perugini et al. 2010). Anderson AT (1976) Magma mixing: petrological process
and volcanological tool. J Volcanol Geotherm Res
Preserved structures would indicate mixing 1:3–33
was the last process recorded by the magmatic Arienzo I, D’Antonio M, Di Renzo V, Tonarini S,
system and its study can unravel unprecedented Minolfi G, Orsi G, Carandente A, Belviso P, Civetta L
information on pre-eruptive behaviour of (2015) Isotopic microanalysis sheds light on the
magmatic endmembers feeding volcanic eruptions:
volcanoes. the Astroni 6 case study (Campi Flegrei, Italy) J
The compositional heterogeneity produced by Volcanol Geoth Res 304:24–37
magma mixing, and subsequently frozen in time Arienzo I, Mazzeo FC, Moretti R, Cavallo A, D’Antonio
in the volcanic products, can hence be viewed as M (2016) Open-system magma evolution and fluid
transfer at Campi Flegrei caldera (Southern Italy)
a broken clock at a crime scene; it can potentially during the past 5 ka as revealed by geochemical and
be used to determine the time of the incident. isotopic data: the archetype of Nisida eruption. Chem
Following this idea and combining numerical Geol 427:109–124
simulations with magma mixing experiments Bacon CR (1986) Magmatic inclusions in silicic and
intermediate volcanic rocks. J Geophys Res 91:6091–
using natural compositions and statistical analy- 6112
ses it was shown that for three volcanic eruptions Bagagli M, Montagna CP, Papale P, Longo A (2017)
from the Campi Flegrei volcanic system Signature of magmatic processes in strainmeter
(Astroni, Averno and Agnano Monte Spina) the records at Campi Flegrei (Italy). Geophys Res Lett
44(2):718–725
mixing-to-eruption timescale are of the order of a Baker DR (1990) Chemical Interdiffusion of Dacite and
few minutes (Perugini et al. 2015). These time- Rhyolite—Anhydrous Measurements at 1 Atm and 10
scales indicate that very little time elapsed from Kbar, Application of Transition-State Theory, and
the moment mixing started until eruption. These Diffusion in Zoned Magma Chambers. Contrib Min-
eral Petr 104:407–423
results are in agreement with recent numerical Bateman R (1995) The interplay between crystallization,
simulations of magma mixing (Montagna et al. replenishment and hybridisation in large felsic magma
2015) that highlight mixing timescales of a few chambers. Earth Sci Rev 39:91–106
hours to attain complete hybridization of mag- Blundy J, Sparks SRJ (1992) Petrogenesis of mafic
enclaves in granitoids of the Adamello massif, Italy.
mas for the Campi Flegrei magmatic systems. J Petrol 33:1039–1104
These results have implications for civil pro- Bowen N (1928) The evolution of the igneous rocks.
tection planning of future volcanic crisis as the Princeton University Press, Princeton, N.J
134 D. Morgavi et al.

Braschi E, Francalanci L, Vougioukalakis GE (2012) D’Antonio M, Tonarini S, Arienzo I, Civetta L, Dallai L,

Inverse differentiation pathway by multiple mafic Moretti R, Orsi G, Andria M, Trecalli A (2013)
magma refilling in the last magmatic activity of Nisyros Mantle and crustal processes in the magmatism of the
Volcano, Greece. Bull Volcanol 74:1083–1100 Campania region: inferences from mineralogy, geo-
Brown RJ, Civetta L, Arienzo I, D’Antonio M, Moretti R, chemistry, and Sr–Nd–O isotopes of young hybrid
Orsi G, Tomlinson EL, Albert PG, Menzies MA volcanics of the Ischia island (South Italy). Contrib
(2014) Geochemical and isotopic insights into the Mineral Petr 165:1173–1194. https://doi.org/10.1007/
assembly, evolution and disruption of a magmatic s00410-013-0853-x
plumbing system before and after a cataclysmic Davidson JP, McMillan NJ, Moorbath S, Worner G
caldera Dcollapse eruption at Ischia volcano (Italy). (1990) The Nevados de Payachata volcanic region
Contrib Mineral Petrol 168, 1035. doi:https://doi.org/ (18S, 69 W, N. Chile) II. Evidence for widespread
10.1007/s00410-014-1035-1 crustal involvement in Andean magmatism. Contrib
Bunsen W (1851) Uber die Processe der vulkanischen Mineral Petr 105:412–432
Gesteinsbildungen Islands. Ann Phys Chem (Dritte Davidson JP, Tepley FJ III (1997) Recharge in volcanic
Reihe) 83:197–272 systems; evidence from isotopic profiles of phe-
Chamberlain KJ, Morgan DJ, Wilson CJN (2014) Time- nocrysts. Science 275:826–829
scales of mixing and mobilisation in the Bishop Tuff Davidson JP, Tepley FJ III, Knesel KM (1998) Crystal
magma body: perspectives from diffusion chronome- isotopic stratigraphy; a method for constraining
try. Contrib Mineral Petr 167:1034. https://doi.org/10. magma differentiation pathways. Eos 79:185–189
1007/s00410-014-1034-2 De Campos CP, Dingwell DB, Fehr KT (2004) Decou-
Chiodini G, Caliro S, De Martino P, Avino R, Gherardi F pled convection cells from mixing experiments with
(2012) Early signals of new volcanic unrest at Campi alkaline melts from Phlegrean Fields. Chem Geol
Flegrei caldera? Insights from geochemical data and 213:227–251
physical simulations. Geology 40:943–946 De Campos CP, Dingwell DB, Perugini D, Civetta L,
Chiodini G, Todesco M, Caliro S, Del Gaudio C, Fehr TK (2008) Heterogeneities in magma chambers:
Macedonio G, Russo M (2003) Magma degassing as insight from the behaviour of major and minor
a trigger of bradyseismic events; the case of Phlegrean elements during mixing experiments with natural
Fields (Italy). Geophys Res Lett 30(8):1434. https:// alkaline melts. Chem Geol 256:131–145
doi.org/10.1029/2002GL01679 De Campos CP, Perugini D, Ertel-Ingrisch W, Ding-
Chiodini G, Vandemeulebrouck J, Caliro S, D’Auria L, well DB, Poli G (2011) Enhancement of Magma
De Martino P, Mangiacapra A, Petrillo Z (2015) mixing efficiency by Chaotic Dynamics: an experi-
Evidence of thermal-driven processes triggering the mental study. Contrib Mineral Petr 161:863–881
2005–2014 unrest at Campi Flegrei caldera. Earth Deino AL, Orsi G, de Vita S, Piochi M (2004) The age of
Planet Sci Lett 414:58–67 the Neapolitan Yellow Tuff caldera-forming eruption
Cioni R, Civetta L, Marianelli P, Metrich N, Santacroce R, (Campi Flegrei caldera—Italy) assessed by 40Ar/39Ar
Sbrana A (1995) Compositional layering and dating method. J Volcanol Geotherm Res 133:
syn-eruptive mixing of a periodically refilled shallow 157–170
magma chamber: the AD 79 Plinian eruption of De Rosa R, Donato P, Ventura G (2002) Fractal analysis
Vesuvius. J Petrol 36:739–776 of min- gled/mixed magmas: an example from the
Civetta L, Orsi G, Pappalardo L, Fisher RV, Heiken G, Upper Pollara eruption (Salina Island, Southern
Ort M (1997) Geochemical zoning, mingling, eruptive Tyrrhenian Sea, Italy). Lithos 65:299–311
dynamics and depositional processes—the Campanian De Rosa R, Mazzuoli R, Ventura G (1996) Relationships
Ignimbrite, Campi Flegrei Caldera, Italy. J Volcanol between deformation and mixing processes in lava
Geotherm Res 75:183–219 flows: a case study from Salina (Aeolian Islands,
Corsaro RA, Di Renzo V, Di Stefano S, Miraglia L, Tyrrhenian Sea). Bull Volcanol 58:286–297
Civetta L (2013) Relationship between magmatic de Vita S, Orsi G, Civetta L, Carandente A, D’Antonio M,
processes in the plumbing system of Mt. Etna and Deino A, di Cesare T, Di Vito MA, Fisher RV,
the dynamics of the eastern flank: inferences from the Isaia R, Marotta E, Necco A, Ort M, Pappalardo L,
petrologic study of the products erupted from 1995 to Piochi M, Southon J (1999) The Agnano-Monte Spina
2005. J Volcanol Geotherm Res 251:75–89 eruption (4100 years BP) in the restless Campi Flegrei
Costa F, Chakraborty S (2004) Decadal time gaps caldera (Italy). J Volcanol Geotherm Res 91:269–301
between mafic intrusion and silicic eruption obtained Del Gaudio C, Aquino I, Ricciardi GP, Ricco C, Scan-
by chemical zoning patterns in olivine. Earth Planet done R (2010) Unrest episodes at Campi Flegrei: a
Sci Lett 227:517–530 reconstruction of vertical ground movements during
D’Antonio M, Civetta L, Orsi G, Pappalardo L, Piochi M, 1905–2009. J Volcanol Geotherm Res 195:48–56
Carandente A, de Vita S, Di Vito MA, Isaia R (1999) Di Muro A, Metrich N, Vergani D, Rosi M, Armienti P,
The present state of the magmatic system of the Campi Fourgeroux T, Deloule E, Arienzo I, Civetta L (2014)
Flegrei caldera based on a reconstruction of its The shallow plumbing system of Piton de la Fournaise
behavior in the past 12 ka. J Volcanol Geotherm volcano (La Reunion island, Indian Ocean) revealed
Res 91:247–268 by the major 2007 caldera forming eruption. J Petrol
Magma Mixing: History and Dynamics of an Eruption Trigger 135

55(7):1287–1315. https://doi.org/10.1093/petrology/ Jolis EM, Freda C, Troll VR, Deegan FM, Blythe LS,
egu025 McLeod CL, Davidson JP (2013) Experimental sim-
Di Renzo V, Arienzo I, Civetta L, D’Antonio M, ulation of magma-carbonate interaction beneath Mt.
Tonarini S, Di Vito MA, Orsi G (2011) The magmatic Vesuvius, Italy. Contrib Mineral Petr 166:1335–1353
feeding system of the Campi Flegrei caldera: archi- Kent AJR, Darr C, Koleszar AM, Salisbury MJ,
tecture and temporal evolution. Chem Geol 281: Cooper KM (2010) Preferential eruption of andesitic
227–241 magmas through recharge filtering. Nat Geosci 3:
Di Vito MA, Arienzo I, Braia G, Civetta L, D’Antonio M, 631–636
Orsi G (2011) The Averno 2 fissure eruption: a recent Kinman WS, Neal CR, Davidson JP, Font L (2009) The
small size explosive event at the Campi Flegrei dynamics of Kerguelen Plateau magma evolution: new
caldera. Bull Volcanol 73:295–320. https://doi.org/ insights from major element, trace element and Sr
10.1007/s00445-010-0417-0 isotopic microanalysis of plagioclase hosted in Elan
Didier J, Barbarin B (1991) Enclaves and Granite Bank basalts. Chem Geol 264:247–265
Petrology, vol 13. Elsevier, Amsterdam Knesel KM, Davidson JP, Duffield WA (1999) Evolution
Eichelberger JC (1975) Origin of andesite and dacite: of silicic magma through assimilation and subsequent
evidence of mixing at Glass Mountain in California recharge: evidence from Sr isotopics in sanidine
and other circum-Pacific volcanoes. Geol Soc Amer phenocrysts Taylor Creek Rhyolites. J Petrol
Bull 86:1381–1391 40:773–786
Eichelberger JC (1978) Andesitic volcanism and crustal Kress V (1997) Magma mixing as a source for Pinatubo
evolution. Nature 275:21–27 sulphur. Nature 389:591–593
Eichelberger JC (1980) Vesiculation of mafic magma Lesher CE (1990) Decoupling of chemical and isotopic
during replenishment of silicic magma reservoirs. exchange during magma mixing. Nature 344:235–237
Nature 288:446–450 Longo A, Papale P, Vassalli M, Saccorotti G, Mon-
Fedele L, Scarpati C, Lanphere M, Melluso L, Morra V, tagna CP, Cassioli A, Giudice S, Boschi E (2012)
Perrotta A, Ricci G (2008) The Breccia Museo Magma convection and mixing dynamics as a source
formation, Campi Flegrei, southern Italy: geochronol- of Ultra-Long-Period oscillations. Bull Volcanol 74
ogy, chemostratigraphy and relationship with the (4):873–880. doi:https://doi.org/10.1007/s00445-011-
Campanian Ignimbrite eruption. Bull Volcanol 0570-0, https://doi.org/10.1007/s00445-011
70:1189–1219 Martin VM, Morgan DJ, Jerram DA, Caddick MJ,
Fedele L, Zanetti A, Morra V, Lustrino M, Melluso L, Prior DJ, Davidson JP (2008) Bang! Month-scale
Vannucci R (2009) Clinopyroxene/liquid trace ele- eruption triggering at Santorini Volcano. Science
ment partitioning in natural trachyte-trachyphonolite 321:1178
systems: insights from Campi Flegrei (southern Italy). Melluso L, De’ Gennaro R, Fedele L, Franciosi L,
Contrib Mineral Petr 158:337–356. https://doi.org/10. Morra V (2012) Evidence of crystallization in resid-
1007/s00410-009-0386-5 ual, Cl-F-rich, agpaitic, trachyphonolitic magmas and
Flinders J, Clemens JD (1996) Non-linear dynamics, primitive Mg-rich basalt-trachyphonolite interaction in
chaos, complexity and enclaves in granitoid magmas. the lava domes of the Phlegrean Fields (Italy). Geol
Trans R Soc Edinburgh: Earth Sci 87:225–232 Mag 149(3):532–550
Font L, Davidson JP, Pearson DG, Nowell GM, Jer- Montagna CP, Papale P, Longo A (2015) Timescales of
ram DA, Ottley CJ (2008) Sr and Pb isotopic mingling in shallow magmatic reservoir. In: Caricchi L,
micro-analysis of plagioclase crystals from Skye Blundy JD (eds) Chemical, Physical and Temporal
lavas: an insight into open-system processes in a Evolution of Volcanic Systems. Gelogical Society,
flood basalt province. J Petrol 49(8):1449–1471 London, Special Publications 422, pp 131–140
Fourcade S, Allegre CJ (1981) Trace element behaviour Morgavi D, Perugini D, De Campos CP, Ertl-Ingrisch W,
in granite genesis: a case study the calc-alkaline Dingwell DB (2013a) Time evolution of chemical
plutonic association from the Querigut Complex exchanges during mixing of rhyolitic and basaltic
(Pyrenees France). Contrib Mineral Petr 76:177–195 melts. Contrib Mineral Petr 166(2):615–638
Francalanci L, Avanzinelli R, Nardini I, Tiepolo M, Morgavi D, Perugini D, De Campos CP, Ertl-Ingrisch W,
Davidson JP, Vannucci R (2012) Crystal recycling in Dingwell DB (2013b) Morphochemistry of patterns
the steady-state system of the active Stromboli produced by mixing of rhyolitic and basaltic melts.
volcano: a 2.5-ka story inferred from in situ J Volcanol Geotherm Res 253:87–96
Sr-isotopic and trace element data. Contrib Mineral Morgavi D, Perugini D, De Campos CP, Ertl-Ingrisch W,
Petr 163:109–131 Lavallee Y, Morgan L, Dingwell DB (2013c) Inter-
Grasset O, Albarede F (1994) Hybridisation of mingling actions Between Rhyolitic and Basaltic Melts Unrav-
magmas with different densities. Earth Planet Sci Lett eled by Chaotic Magma Mixing Experiments. Chem
121:327–332 Geol 346:119–212
Hibbard MJ (1981) The magma mixing origin of mantled Morgavi D, Petrelli M, Vetere FP, Gonzalez-Gartia D,
feldspar. Contrib Mineral Petrol 76:158–170 Perugini D (2015) High temperature apparatus for
Hibbard MJ (1995) Petrography to petrogenesis. MacMil- chaotic mixing of natural silicate melts. Rev Sci
lan Publishing Company, London Instrum 86(10):105108
136 D. Morgavi et al.

Morgavi D, Arzilli F, Pritchard C, Perugini D, Mancini L, Plail M, Barclay J, Humphreys MCS, Edmonds M,
Larson P, Dingwell BD (2016) The Grizzly Lake Herd RA, Christopher TE (2014) Characterization of
complex (Yellowstone Volcano, USA): mixing mafic enclaves in the erupterd products of Soufriere
between basalt and rhyolite unraveled by microanal- Hills Volcano, Montserrat, 2009–2010. In Wadge G,
ysis and X-ray microtomography. Lithos 260:457–474 Robertson REA, Voight B, (eds) The eruption of
Murphy MD (1998) The role of Magma Mixing in Soufriere Hills Volcano, Montserrat from 2000 to
triggering the current eruption at the Soufriere Hills 2010. Geological Society, London, Memoirs, 39, 343–
Volcano, Montserrat, West Indies. Geophys Res Lett 360
25:3433–3436 Poli G, Tommasini S, Halliday AN (1996) Trace elements
Laumonier M, Scaillet B, Pichavant M, Champallier R, and isotopic exchange during acid-basic magma
Andujar J, Arbaret L (2014) On the conditions of interaction processes. Trans Royal Soc Edinburgh:
magma mixing and its bearing on andesite production Earth Sci 87:225–232
in the crust. Nature communications, volume 5 Pritchard CJ, Larson PB, Spell TL, Tarbert KD (2013a)
Papale P, Moretti R, Barbato D (2006) The compositional Eruption-triggered mixing of extra-caldera basalt and
dependence of the saturation surface of H2O + CO2 rhyolite complexes along the East Gallatin-Washburn
fluids in silicate melts. Chem Geol 229(1–3):78–95 fault zone, Yellowstone National Park, WY, USA.
Pappalardo L, Piochi M, D’Antonio M, Civetta L, Lithos 175–176:163–177. doi:https://doi.org/10.1016/
Petrini R (2002) Evidence for multistage magmatic j.lithos.2013.04.022
evolution during the past 60 kyr at Campi Flegrei Self S (1992) Krakatau revisited: the course of events and
(Italy) deduced from Sr, Nd and Pb isotopic data. interpretation of the 1883 eruption. Geoj Lib 28:109–
J Petrol 43(8):1415–1434 121
Perugini D, De Campos CP, Dingwell DB, Petrelli M, Sigmundsson F, Hreinsdottir S, Hooper A, Arnadottir T,
Poli G (2008) Trace element mobility during magma Pedersen R, Roberts JM, Oskarsson N, Auriac A,
mixing: preliminary experimental results. Chem Geol Decriem J, Einarsson P, Geirsson H, Hensch M,
256:146–157 Ofeigsson BG, Sturkell E, Sveinbjornsson H,
Perugini D, de Campos CP, Petrelli M, Dingwell DB Feigl LK (2010) Intrusion triggering of the 2010
(2015) Concentration variance decay during magma Eyjafjallajokull explosive eruption. Nature 468:426–
mixing: a volcanic chronometer. Sci Rep 5:14225 430
Perugini D, Petrelli M, Poli G (2006) Diffusive fraction- Slaby E, Gotze J, Worner G, Simon K, Wrzalik R,
ation of trace elements by chaotic mixing of magmas. Smigielski M (2010) K-feldspar phenocrysts in micro-
Earth Planet Sci Lett 243:669–680 granular magmatic enclaves: a cathodoluminescence
Perugini D, Petrelli M, Poli G, De Campos C, Ding- and geochemical study of crystal growth as a marker
well DB (2010) Time-scales of recent Phlegrean fields of magma mingling dynamics. Lithos 105:85–97
eruptions inferred from the application of a ‘Diffusive Smith JV (2000) Structures on interfaces of mingled
Fractionation’ model of trace elements. Bull Volcanol magmas, Stewart Island, New Zealand. J Struct Geol
72:431–447 22:123–133
Perugini D, Poli G (2005) Viscous fingering during Sparks SRJ, Marshall LA (1986) Thermal and mechanical
replenishment of felsic magma chambers by continu- constraints on mixing between mafic and silicic
ous inputs of mafic magmas: field evidence and magmas. J Volcanol Geotherm Res 29:99–124
fluid-mechanics experiments. Geology 33:5–8 SparksSRJ Sigurdsson H, Wilson L (1977) Magma
Perugini D, Poli G (2012) The mixing of magmas in mixing: a mechanism for triggering acid explosive
plutonic and volcanic environments: analogies and eruptions. Nature 267:315–318
differences. Lithos. https://doi.org/10.1016/j.lithos. Tonarini S, D’Antonio M, Di Vito MA, Orsi G, Caran-
2012.02.002 dente A (2009) Geochemical and isotopical (B, Sr,
Perugini D, Poli G, Gatta G (2002) Analysis and Nd) evidence for mixing and mingling processes in the
simulation of Magma mixing processes in 3D. Lithos magmatic system feeding the Astroni volcano (4.1–
65:313–330 3.8 ka) within the Campi Flegrei caldera (South Italy).
Perugini D, Poli G, Mazzuoli R (2003) Chaotic advection, Lithos 107:135–151
fractals and diffusion during mixing of Magmas: Tonarini S, Leeman WP, Civetta L, D’Antonio M,
evidence from Lava Flows. J Volcanol Geotherm Res Ferrara G, Necco A (2004) B/Nb and 511B systematics
124:255–279 in the Phlegrean Volcanic District (PVD). J Volcanol
Perugini D, Valentini L, Poli G (2007) Insights into Geotherm Res 113:123–139
Magma chamber processes from the analysis of size Ventura G (1998) Kinematic significance of
distribution of Enclaves in Lava Flows: a case study mingling-rolling structures in lava flows: a case study
from Vulcano Island (Southern Italy). J Volcanol from Porri volcano (Salina Southern Tyrrhenian Sea).
Geotherm Res 166:193–203 Bull Volcanol 59:394–403
Petrelli M, Perugini D, Poli G (2011) Transition to Chaos Wada K (1995) Fractal structure of heterogeneous ejecta
and implications for Time-scales of Magma from the Meakan volcano, eastern Hokkaido, Japan:
Hybridization during mixing processes in Magma implications for mixing mechanism in a volcanic
chambers. Lithos 125:211–220 conduit. J Volcanol Geotherm Res 66:69–79
Magma Mixing: History and Dynamics of an Eruption Trigger 137

Walker GPL, Skelhorn RR (1966) Some associations of Wilcox RE (1944) Rhyolite-basalt complex of Gardiner
acid and basic igneous rocks. Earth Sci Rev 2:9–109 River, Yellowstone Park, Wyoming. Geol Soc Am
Wallace G, Bergantz G (2002) Wavelet-based correlation Bull 55:047–1080
(WBC) of crystal populations and magma mixing. Wilcox RE (1999) The idea of Magma Mixing: history of
Earth Planet Sci Lett 202:133–145 a struggle for Acceptance. J Geol 107(4):421–432
Wiebe RA (1994) Silicic magma chambers as traps for Williams-Jones G, Rymer H (2002) Detecting volcanic
basaltic magmas: the Cadillac mountain intrusive eruption precursos: a new method using gravity and
complex, Mount Desert island, Maine. J Geol deformation measurements. J Volcanol Geotherm Res
102:423–427 113:379–389

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Gases as Precursory Signals:
Experimental Simulations, New
Concepts and Models of Magma
Degassing

M. Pichavant, N. Le Gall and B. Scaillet

Abstract experimental support for disequilibrium degas-

Volatile release during magma ascent in vol- sing during ascent of CO2-bearing magmas.
canic conduits (magma degassing) forms the The existence of volatile concentration gradi-
basis for using volcanic gases as precursory ents around nucleated gas bubbles suggests
signals. Recent high temperature high pressure that degassing is controlled by the respective
experimental simulations have yielded results mobilities (diffusivities) of volatiles within the
that challenge key assumptions related to melt. The recently formulated diffusive frac-
magma degassing and are important for the tionation model reproduces the main charac-
interpretation of glass inclusion and gas data teristics, especially the volatile concentrations,
and for using volcanic gas as precursory of experimental glasses. The model also shows
signals. The experimental data show that, for that the gas phase is more H2O-rich than
ascent rates expected in natural systems, pure expected at equilibrium because CO2 transfer
H2O basaltic melts will evolve mostly close to toward the gas phase is hampered by its
equilibrium when decompressed from 200 to retention within the melt. However, only
25 MPa. In the same way, degassing of H2O–S integrated gas compositions are calculated.
species evolves at near equilibrium, although Similarly, only bulk experimental fluid com-
this conclusion is limited by the number of S positions are determined in recent experiments.
solubility data available for basaltic melts. Thus, constraints on the local gas phase are
However, degassing of CO2 is anomalous in all becoming necessary for the application to
studies, whether performed on basaltic or volcanoes. This stresses the need for the direct
rhyolitic melts. CO2 stays concentrated in the analysis of gas bubbles nucleated in decom-
melt at levels far exceeding solubilities. The pression experiments. Pre-eruptive changes in
anomalous behaviour of CO2, when associated volcanic CO2/SO2 and H2O/CO2 gas ratios are
with near equilibrium H2O losses, yields interpreted to reflect different pressures of
post-decompression glasses with CO2 concen- gas-melt segregation in the conduit, an
trations systematically higher than equilibrium approach that assume gas-melt equilibrium.
degassing curves. Therefore, there is strong However, if disequilibrium magma degassing
is accepted, the use of volatile saturation codes
is no longer possible and caution must be
exercised with the application of local equilib-
M. Pichavant (&)
N. Le Gall
B. Scaillet rium to volcanic gases. Future developments in
CNRS, Orléans, France
the interpretation of gas data require progress
e-mail: [email protected]

Advs in Volcanology (2019) 139–154

DOI 10.1007/11157_2018_35
© The Author(s) 2018
Published Online: 09 June 2018
140 M. Pichavant et al.

from both sides, experimental and volcanolog- degassing during ascent of CO2-bearing magmas.
ical. One priority is to reduce the gap in scales The existence of volatile concentration gradients
between experiments and gas measurements. around nucleated gas bubbles suggests that
degassing is controlled by the respective mobili-
ties (diffusivities) of volatiles within the melt. The
Keywords contrasted diffusivities of dissolved volatile spe-


Magma ascent Decompression experiments cies (in particular H2O and CO2) selectively limit


Disequilibrium degassing Volatile diffusion their transfer toward the gas phase for timescales
Gas compositions typical for magma ascent. The diffusive fraction-
ation model recently formulated reproduces the
main characteristics, especially the volatile con-
Extended Abstract centrations, of experimental glasses. It provides a
framework to interpret the new experimental
Volcanic gases are one of the main tools to observations and the systematic deviations from
monitor changes in the activity of volcanoes and equilibrium observed in CO2-bearing systems,
forecast their eruption. Magma ascent toward the although coupling between volatile diffusion and
surface is associated with the exsolution of vola- vesiculation requires a more elaborate treatment.
tiles initially dissolved in the melt, a process The model also shows that the gas phase is more
designated as “magma degassing”. Classically, H2O-rich than expected at equilibrium because
the interpretation of volcanic gases relies on the CO2 transfer toward the gas phase is hampered by
assumption that degassing takes place at equilib- its retention within the melt. However, only inte-
rium. However, several observations (CO2 con- grated gas compositions are calculated. In the
tents of basaltic seafloor glasses, H2O and CO2 same way, only bulk experimental fluid compo-
concentrations in glass inclusions, explosive sitions are determined in recent experiments.
basaltic volcanism) do not fit easily in such a Since the gas phase is essential for the application
model. Recently, decompression, ascent and to volcanoes, constraints on the local gas phase
degassing of magmas in volcanic conduits have are becoming necessary. Compositions of gas
been simulated by high temperature high pressure bubbles in decompression experiments must be
experiments. Results from these simulations linked not only with pressure but also with volatile
stress the need to critically reconsider the whole concentrations of local melts, specific degassing
mechanism of degassing in basaltic but also rhy- textures and mechanisms. As a way in this
olitic magmas. The new experimental data show direction, local gas-melt equilibrium assumes that
that, at the decompression rates tested, pure H2O chemical equilibrium persists at the local scale,
basaltic melts will evolve mostly close to equi- despite evidence for disequilibrium at larger
librium when decompressed from 200 to 25 MPa. scales. However, there are alternative ways to
In the same way, degassing of H2O–S species constrain the composition of nucleated gas bub-
evolves at near equilibrium, although this con- bles, thus stressing the need for their direct anal-
clusion is limited by the number of S solubility ysis in decompression experiments. Pre-eruptive
data available for basaltic melts. Degassing of changes in CO2/SO2 and, in some cases,
CO2 is anomalous in all studies, whether per- H2O/CO2 gas ratios observed at several basaltic
formed on basaltic or rhyolitic melts. CO2 stays volcanoes are generally interpreted to reflect dif-
concentrated in the melt at levels far exceeding ferent pressures of gas-melt segregation in the
solubilities. The anomalous behaviour of CO2, conduit, an approach that assume gas-melt equi-
when associated with near equilibrium H2O los- librium. However, if disequilibrium magma
ses, yields post-decompression glasses with CO2 degassing is accepted, volatile saturation codes
concentrations systematically higher than equi- can no longer be directly used. Caution also must
librium degassing curves. Therefore, there is be exercised with the application of local gas-melt
strong experimental support for disequilibrium equilibrium to volcanic gases which are probably
Gases as Precursory Signals: Experimental Simulations … 141

closer to integrated rather than to local composi- such a model. They include (1) the existence of
tions. Future developments in the interpretation of basaltic seafloor glasses often supersaturated in
gas data require progress from both sides, exper- CO2 (e.g., Aubaud et al. 2004), (2) the occurrence
imental and volcanological. One priority is to of glass inclusions with H2O and CO2 concen-
reduce the gap in scales between experiments and trations inconsistent with closed system equilib-
gas measurements to refine interpretations of gas rium degassing (e.g., Metrich et al. 2010) and
compositions as unrest signals. (3) the occurrence of explosive basaltic volcanism
(e.g., Head and Wilson 2003) which implies
sudden rather than gradual release of volatiles.
Recently, decompression and ascent of basaltic
magmas in volcanic conduits has been simulated
by high temperature high pressure petrological
1 Magma Degassing and Volcanic experiments. These simulations stress the need to
Gases as Precursory Signals critically reconsider the whole mechanism of
degassing in basaltic but also more silicic mag-
Volcanic gases are one of the main tools used to mas. In particular, the assumption of equilibrium
monitor changes in the activity of volcanoes and degassing is now becoming increasingly chal-
forecast their eruption. This approach is rooted in lenged. This has major implications for the
the strong pressure dependence of the solubility interpretation of glass inclusion and gas data and,
of volatiles (mainly H2O, CO2, SO2, H2S, Cl) in more generally, for the use of volcanic gas as
silicate melts. Accordingly, magma ascent toward precursory signals. In this Chapter, first, the recent
the surface is associated with the exsolution of experimental simulations are reviewed. We show
volatiles initially dissolved in the melt, a process that they all demonstrate an anomalous behaviour
designated as “magma degassing”. The different for CO2 which tends to stay dissolved within the
volatiles have contrasted solubilities in silicate melt at concentrations too high for equilibrium.
melts and, therefore, are expected to react differ- Second, the diffusive fractionation model which
ently to decompression. This forms the basis for has been proposed to account for the new exper-
using volcanic gas ratios to infer magma ascent imental observations is described and critically
and depth of gas segregation in volcanic conduits. discussed. Finally, the implications of disequi-
For example, the sudden increase of gas CO2/SO2 librium degassing for experimental fluid compo-
ratio has been used as an indication for deep sitions and the interpretation of volcanic gas data
magma recharge at Stromboli (Aiuppa et al. as precursory signals are explored.
2010). At Soufriere Hills volcano (Montserrat), a
correlation has been noted between gas HCl/SO2
and the level of shallow activity as marked by the 2 Experimental Simulations
rate of lava extrusion and dome growth
(Christopher et al. 2010; Edmonds et al. 2010). 2.1 Basaltic Systems
Classically, the interpretation of volcanic gases
relies on the assumption that degassing takes place Following early work on systems with only pure
at equilibrium. In the case of basaltic magmas, this CO2 (Lensky et al. 2006), decompression
assumption is supported by the high temperatures, experiments on hydrous basaltic melts have been
low viscosities and high volatile diffusivities carried out recently by Pichavant et al. (2013) at
(Sparks et al. 1994). Vesiculation (i.e., the com- 1150–1180 °C, for initial pressures of 200–
bined processes of bubble nucleation, growth and 250 MPa, final pressures of 100, 50 and 25 MPa
coalescence) is thought to be relatively easy in and for decompression rates between *1.5
basaltic melts and degassing of basaltic magma is down to 0.25 m/s. Melts from Stromboli,
classically viewed as an equilibrium process. pre-synthesized to incorporate dissolved H2O
However, several observations do not fit easily in (2.7–3.8 wt%) and CO2 (600–1300 ppm), were
142 M. Pichavant et al.

used as starting materials. The experiments were texturally by X-ray microtomography. Volatile
of continuous decompression type, and both equilibrium was reached or approached during
constant (one ramp) and variable (two ramps) decompression in all series #1 melts with just
decompression rates were imposed. Final melt water. In contrast, disequilibrium degassing
H2O concentrations were homogeneous and occurred systematically in series #2 and #3 melts
always close to equilibrium solubility values. In which retained elevated CO2 concentrations
contrast, the rate of vesiculation was found to (Fig. 1a). In similar experiments performed on
control the final melt CO2 concentration. High the same three glass series but at a slower
vesicularity charges had glass CO2 concentra- decompression rate of 1.5 m/s, Le Gall and
tions that follow theoretical equilibrium degas- Pichavant (2016b) found that series #1 (CO2-
sing paths whereas glasses from low vesicularity free) melts followed equilibrium degassing until
charges showed marked deviations from equi- 100 MPa final pressure (Pfin). But at both 60 and
librium, with CO2 concentrations up to one order 50 MPa Pfin, a slight H2O-supersaturation was
of magnitude higher than equilibrium solubilities recognized, associated with a second bubble
(Fig. 1a). The experimental results were inter- nucleation event that occurred at 25 MPa. In
preted in light of the slower diffusivity of CO2 comparison, in series #2 and #3 (CO2-bearing)
relative to H2O in basaltic melts. melts, disequilibrium degassing was systematic,
Yoshimura (2015) decompressed a natural glasses retaining high non-equilibrium CO2
evolved basaltic melt containing dissolved H2O concentrations (Fig. 1a).
and CO2 at 1200 °C and between 1000 and The behavior of H2O–, CO2– and S-bearing
500 MPa. The short decompression duration of basaltic melts during decompression was inves-
10 min over this pressure interval simulates a tigated by Le Gall et al. (2015a). Stromboli melts
very fast ascent rate (*32 m/s for a rock density with 2.72 ± 0.02 wt% H2O, 1291 ± 85 ppm
of 2650 kg/m3). A vesiculated glass was pro- CO2 and 1535 ± 369 ppm S were synthesized at
duced and Fourier Transform Infrared Spec- 1200 °C and 200 MPa and then decompressed to
troscopy (FTIR) profiles revealed large CO2 final pressures (Pfin) ranging from 150 to
concentration gradients in the melt adjacent to 25 MPa, followed by rapid quenching. The
gas bubbles. In contrast, the melt H2O content continuous decompressions were conducted at
was almost constant throughout the sample. The rates of 1.5 and 3 m/s. During decompression, S
glass volatile concentration data cover a near (and H2O) were lost slightly more from the melt
vertical trend in the H2O–CO2 diagram (Fig. 1a). than expected from equilibrium degassing mod-
Le Gall and Pichavant (2016a) extended the els, whilst significant CO2 was retained at ele-
decompression experiments performed by vated concentrations in the melt (Fig. 1a). It was
Pichavant et al. (2013), using essentially the found that the degassing trend recorded by
same procedures and materials. Three starting Stromboli glass inclusions could be closely
volatile compositions were investigated: series reproduced by the experiments (Fig. 1b; Le Gall
#1 (4.91 wt% H2O, no CO2), series #2 et al. 2015a). For andesitic melts, Fiege et al.
(2.41 ± 0.04 wt% H2O, 973 ± 63 ppm CO2) (2014) observed that the fluid/melt partition
and series #3 (0.98 ± 0.16 wt% H2O, coefficient for sulfur increases with the decom-
872 ± 45 ppm CO2). The volatile-bearing glas- pression rate. However, the influence of decom-
ses were synthesized at 1200 °C and 200 MPa, pression rate on S degassing was marked only for
then continuously decompressed at a fast oxidizing conditions, corresponding to sulfate as
decompression rate of 3 m/s in the pressure the only S species, thus making necessary to
range 150–25 MPa and then rapidly quenched. consider the different behaviour of S2− and S6+
Post-decompression glasses were characterized during degassing.
Gases as Precursory Signals: Experimental Simulations … 143

4000
(a) Basaltic experiments

3000
500 MPa
(2)

CO2 (ppm)
2000
Closed-system
(4) degassing

1000 200
(3)
100 (1)
50
25
0
0 1 2 3 4
4000 H2O (wt%)

(b) Stromboli Pumice-MI

Pumice-Emb.
3000 Scoria-MI
500 MPa
CO2 (ppm)

2000

1000 200
Closed-system
100 degassing
50
25
0
0 1 2 3 4
H2O (wt%)

200
(c) Rhyolitic Cinit
experiments
150
CO2 (ppm)

50 100
MPa MPa
100

50
Closed-system
degassing
Cnew
0
2.0 2.5 3.0 3.5 4.0
H2O (wt%)

Fig. 1 H2O–CO2 glass concentration diagrams for et al. 2013; Yoshimura 2015; Le Gall and Pichavant
a basaltic decompression experiments, b Stromboli glass 2016a, b; Le Gall et al. 2015a). In b, the glass inclusion
inclusions and c rhyolitic decompression experiments. In data are from Metrich et al. (2010). MI: glass inclusions,
a, b and c, light grey curves are isobars labelled with Emb: embayments. In c, the glass data field and the bold
pressure in MPa. In a, fields for post-decompression theoretical equilibrium closed-system degassing curve are
glasses are distinguished with (1) referring to Pichavant redrawn from Yoshimura (2015). Cinit and Cnew are the
et al. (2013), (2) to Yoshimura (2015), (3) to Le Gall and composition of the pre-decompression melt and of the
Pichavant (2016a, b) and (4) to Le Gall et al. (2015a). estimated post-decompression melt at the gas-melt inter-
Black curves are closed-system equilibrium degassing face, respectively (Yoshimura 2015; see also Figs. 5
trajectories redrawn from the original figures (Pichavant and 6)
144 M. Pichavant et al.

2.2 Rhyolitic Systems degassing of S species evolves at near equilibrium

(Le Gall et al. 2015a) although the reference equi-
Yoshimura (2015) decompressed a natural rhy- librium model (Burgisser et al. 2015) is somewhat
olitic melt containing dissolved H2O and CO2 uncertain due to the limited number of S solubility
from 100 to 50 MPa at 800 °C. The duration of data for calibration. Degassing of CO2 is anoma-
the decompression was 5000 s corresponding to lous in all studies, whether performed on basaltic or
a decompression rate of 0.38 m/s (for a density rhyolitic melts (Pichavant et al. 2013; Yoshimura
of 2650 kg/m3). FTIR analysis of the vesiculated 2015; Le Gall and Pichavant 2016a, b; Le Gall et al.
glass sample showed CO2 concentration gradi- 2015a). CO2 stays concentrated in the melt at
ents in the melt away from gas bubbles. In con- concentrations far exceeding solubilities (Fig. 1).
trast, H2O was found to be distributed Except in the very fast basalt decompression
homogeneously within the sample although H2O experiment of Yoshimura (2015), the anomalous
concentrations decreased significantly relative to behaviour of CO2 is associated with significant
the pre-decompression melt showing it had ree- H2O losses which results in post-decompression
quilibrated. On the H2O–CO2 diagram, the glass glasses plotting systematically left of theoretical
volatile concentrations define a near vertical equilibrium degassing trajectories in Fig. 1. We
array located left of the theoretical equilibrium conclude that recent experimental studies strongly
degassing curve (Fig. 1c). support the possibility of disequilibrium degassing,
i.e., that ascending melts can keep volatile con-
centrations (particularly CO2) significantly differ-
2.3 Summary of Experimental ent from those expected from equilibrium
Evidence modelling. The questions thus arise of (1) the
mechanisms responsible for this disequilibrium
In all experiments above, melt vesiculation is the behaviour and (2) of the consequences of disequi-
result of decompression, in most cases single-step librium degassing for the composition of the gas
(constant decompression rate) and, more rarely, phase.
multi-step (variable decompression rates,
Pichavant et al. 2013). Vesiculation leads to the
generation of a gas (or fluid) phase. Volatiles par- 3 Modelling Disequilibrium
tition between melt and gas, and volatile concen- Degassing
trations in post-decompression glasses evolve from
those initially dissolved in pre-decompression 3.1 The Diffusive Fractionation
glasses. The evaluation of equilibrium vs. dise- Model
quilibrium degassing is performed by comparing
volatile concentrations of post-decompression Both Pichavant et al. (2013) and Yoshimura
glasses with theoretical closed-system equilibrium (2015) observed decoupling between the beha-
degassing trajectories (Fig. 1). For pure H2O melts, viour of H2O and CO2 during experimental
this equilibrium trajectory is calculated using the decompression and degassing. In both studies,
experimental solubility data of Lesne et al. (2011). CO2 concentration gradients were found in
For CO2– and S-bearing systems, gas-melt equi- post-decompression glasses, either around gas
librium thermodynamic models (Newman and bubbles or near the gas-melt interface. In con-
Lowenstern 2002; Papale et al. 2006; Burgisser trast, no such diffusion profiles were identified
et al. 2015) are used. Results show that pure H2O for H2O, despite concentrations being lower (in
basaltic systems evolve close to equilibrium when most cases) than in pre-decompression glasses
decompressed from 200 to 25 MPa with ascent (Pichavant et al. 2013; Yoshimura 2015; Le Gall
rates of 1.5 and 3 m/s, although small levels of H2O et al. 2015a; Le Gall and Pichavant 2016a, b).
supersaturation are observed below 100 MPa (Le Pichavant et al. (2013) suggested that two
Gall and Pichavant 2016a, b). In the same way, characteristic distances, the gas interface distance
Gases as Precursory Signals: Experimental Simulations … 145

(either the distance between two bubbles in the Yoshimura (2015) emphasized the relative
melt or the distance to the gas-melt interface) and simplicity of his model. For example, bubble
the volatile diffusion distance (a function of growth was not considered as in other
respective diffusivities of volatiles in the melt) more elaborated theoretical treatments (e.g.,
control the degassing process. Yoshimura (2015) Gonnermann and Manga 2005). Rather than
quantitatively formulated a diffusive fractiona- continuously varying boundary (gas-melt) inter-
tion model to describe the ascent and degassing face volatile concentrations and bubble-bubble
of volatile-bearing magmas. The reader is refer- distances as in a natural ascending magma, the
red to this work for details about the calculations. calculations were performed step-by-step (i.e., at
The model is based on a diffusivity of CO2 being different pressures) along the decompression
one log unit lower than for H2O (e.g., Zhang and ramp, with fixed boundary concentrations and
Ni 2010). Decompression trajectories computed bubble-bubble distance (Yoshimura 2015). It is
from the model are shown on Fig. 2 for different also important to note that the volatile concen-
ascent rates, from 0.1, 1, 10, 100 to ∞ m/s. trations on Fig. 2 correspond to averages com-
Although very high ascent rates (e.g., Peslier puted by integrating the concentrations in the
et al. 2015) are necessary for degassing trajec- melt along the diffusion profiles (distance inte-
tories to shift significantly left to the equilibrium grated compositions).
reference curve, the modelling results qualita- Gas phase compositions were calculated by
tively reproduce the main characteristics of mass balance using the initial volatile concen-
experimental post-decompression glasses, i.e., trations and the average volatile concentrations
the elevated CO2 glass concentrations, the sig- left in the melt after decompression and degas-
nificant H2O losses and the melt concentration sing (Yoshimura 2015). Results are shown on
trends in H2O–CO2 diagrams (Fig. 1). Fig. 3 and they correspond to compositions

Fig. 2 H2O and CO2 melt volatile concentrations com- The heavy curve labelled “eq” is the equilibrium
puted with the diffusive fractionation model for different closed-system degassing trajectory as calculated by
decompression/ascent rates (v from 0.1 to ∞, in m/s). Yoshimura (2015). Figure redrawn from Yoshimura
Isobars (light curves) are labelled with pressure in MPa. (2015). See text for details
146 M. Pichavant et al.

Fig. 3 Variations of the gas H2O/CO2 molar ratio The heavy curve labelled “eq” is the equilibrium
(integrated compositions) with pressure computed from closed-system degassing trajectory. Figure redrawn from
the diffusive fractionation model using different Yoshimura 2015. See text for details
decompression/ascent rates (v from 0.1 to 1000, in m/s).

integrated along decompression (pressure inte- vesicularity term should embody the textural
grated compositions). These compositions are variations associated with magma ascent. In the
more H2O-rich (higher H2O/CO2 ratios) than model of Yoshimura (2015), the vesicularity was
gases generated under equilibrium degassing. computed from the amount of volatiles exsolved
upon decompression, using an equation of state
for H2O–CO2 gas mixtures to calculate the den-
3.2 Coupling Between Diffusion sity of the gas and assuming a constant density for
and Vesiculation the melt. In so doing, it is apparent that only a
vesicularity corresponding to equilibrium degas-
Coupling between volatile diffusion and vesicu- sing is considered. Thus, for a given initial
lation is a necessity in diffusive degassing models bubble-bubble distance, the distance between
because vesiculation defines the density of bub- bubbles in the decompression column depends
bles, their sizes and the distances between them only on the equilibrium vesicularity. Degassing
(e.g., Pichavant et al. 2013; Le Gall et al. 2016a, trajectories (Fig. 2) and integrated gas composi-
b). This issue was addressed by Yoshimura tions (Fig. 3) were calculated on this basis.
(2015), although in a relatively simplified man- For comparison, experimental vesicularities,
ner. The distance between bubbles was defined as bubble diameters and bubble number densities
being a function of only two variables, the dis- are shown on Fig. 4 for three series of basaltic
tance between bubbles at the bottom of the melts decompressed from 200 to 25 MPa final
decompression column (arbitrary value) and the pressure (Pfin) at 3 m/s (Le Gall and Pichavant
vesicularity. Vesicularity must change along with 2016a). Systematic variations within the three
decompression and degassing. So, the glass series are observed depending on Pfin. In
Gases as Precursory Signals: Experimental Simulations … 147

most cases, the vesicularity data plot intermediate (with pure H2O) are in general much lower than
between the two equilibrium vesicularity curves, the theoretical vesicularities calculated for pure
which were computed in a similar way than H2O gas. The data also show large changes in
Yoshimura (2015) but only for two end-member bubble sizes and bubble number densities that do
cases corresponding to pure H2O and pure CO2 not directly correlate with vesicularity. Le Gall
gas. The vesicularity data for the series #1 melts and Pichavant (2016a) emphasized that

Fig. 4 Textural data for 100

post-decompression a
experimental glasses plotted #1
as a function of final pressure 80
#2
(Pfin) and comparison with
#3
data for natural basaltic
V (vol.%)
60 Strombolian pumices Equi. H2O
pumices (Stromboli, Masaya).
Vesicularities (a), bubble Equi. CO2
diameters (b) and bubble 40
number densities (c) for three
series of basaltic melts
decompressed from 200 to 20
150, 100, 50 and to 25 MPa
Pfin at 3 m/s (Le Gall and 0
Pichavant 2016a). 0 50 100 150 200
Pre-decompression melt
concentrations, series #1:
4.91 wt% H2O (no CO2), 250
series #2: 2.41 ± 0.04 wt% S+D38#1 b
H2O, 973 ± 63 ppm CO2 and Unfragmented part
series #3: 0.98 ± 0.16 wt% 200
H2O, 872 ± 45 ppm CO2.
Note that, for charge 150 Strombolian pumices
D (µm)

S + S38#1 which was

partially fragmented, the
vesicularity (a) and BND 100 S+D38#1
(c) data concern the Fragmented part
unfragmented part. The
bubble diameter data (b) are 50
for both the unfragmented
(black symbol) and the 0
fragmented (minimum and 0 50 100 150 200
maximum values) parts.
Figure redrawn from Le Gall
and Pichavant (2016a) 1E+5
Masaya basaltic Plinian clasts

1E+4 c
BNDmelt (mm-3)

1E+3 Strombolian pumices

1E+2

1E+1

1E+0

1E-1
0 50 100 150 200
Pfin (MPa)
148 M. Pichavant et al.

degassing textures result from several processes (calculated assuming equilibrium degassing)
including bubble nucleation, growth, coales- gases (Fig. 3) is the consequence of CO2 degas-
cence, plus buoyancy-driven bubble migration. sing being hampered by its retention within the
We conclude that, although the diffusive frac- melt. Therefore, disequilibrium is evidenced in
tionation model of Yoshimura (2015) provides a the compositions of the pressure integrated fluids.
basis for coupling volatile diffusion calculations
and vesiculation processes, more work is needed
to incorporate the complex textural changes 4.2 Composition of Gas Bubbles
associated with ascent of volatile-bearing melts.
The experiments and the diffusive fractionation
model show that melt and gas both evolve under
4 Implications for Gas Phase disequilibrium during magma ascent and degas-
Compositions sing. For the melt, this conclusion is based either
on volatile concentration measurements in glass
4.1 Available Data and Models at some distance of the gas/melt interface
(Pichavant et al. 2013; Yoshimura 2015; Le Gall
Despite the limitations noted above, the diffusive and Pichavant 2015a, 2016a, b) or on average
fractionation model provides a framework to concentrations calculated by integration along
interpret the experimental observations and the diffusion profiles (Yoshimura 2015). For the gas,
systematic deviations from equilibrium degas- constraints are available only on integrated
sing observed in CO2-bearing systems. However, compositions (Le Gall et al. 2015a; Yoshimura
it should be emphasized that the model uses 2015). Since the gas phase is essential for the
analytical data (glass volatile concentrations) and application to volcanoes, and given the inter-
physicochemical properties (volatile diffusivities) pretations proposed for the melt phase, con-
related only to the melt phase. The question straints on the gas phase composition at smaller
arises of the consequences of disequilibrium scales are becoming necessary. This requires
degassing for the gas phase composition. It is linking compositions of gas bubbles in decom-
worth remembering here that the precursory pression experiments not only with pressure but
signals come from gas data. also with volatile concentrations of local melts as
In the decompression experiments summa- well as with degassing textures and mechanisms.
rized above, the gas phase has not been chemi- As a way toward this direction, local gas-melt
cally analysed although some mass balance equilibrium can be assumed. This implies that
calculations were performed to estimate the chemical equilibrium persists locally between
composition of the gas phase in the H2O–, CO2– gas and melt, despite evidence for disequilibrium
and S-bearing experiments of Le Gall et al. at larger scales. Therefore, the volatile composi-
(2015a). However, it is emphasized that, with this tions of melt and gas at the interface are defined
method, only bulk experimental gas compositions by equilibrium partitioning of volatiles between
are provided (charge and pressure integrated these two phases (e.g., Dixon and Stolper 1995).
compositions). No information is available on the To illustrate this concept, a schematic represen-
composition of individual bubbles generated tation of the gas-melt interface for a H2O– and
during decompression. The gas calculations per- CO2-bearing melt decompressed isothermally
formed by Yoshimura (2015) also use a similar from an initial (Pinit) to a final (Pfin) pressure is
mass balance approach, i.e., pressure integrated shown on Fig. 5a. Initial volatile concentrations
fluid compositions are given. However, the local (Cinit), together with the Pinit and Pfin isobars and
gas at the gas-melt interface has an equilibrium the equilibrium degassing trajectory are shown
composition (local gas-melt equilibrium). The on the H2O–CO2 diagram of Fig. 5b. If local
differences between the disequilibrium (calcu- gas-melt equilibrium is assumed, the interface
lated with the model) and the equilibrium melt H2O and CO2 concentrations at Pfin (Cnew)
Gases as Precursory Signals: Experimental Simulations … 149

(a) (b)
Ci Qjoju Cinit
Cinit Nfmu!
gas bubble
H2O Qgjo dpnqptjujpot!

CO2 melt
melt

Cinit
Cnew=Csol Hbt!cvccmf!
Cnew=Csol
CO2 DP30I3P
Cnew=Csol Frvjmjcsjvn
efhbttjoh
distance from bubble H2O melt
interface

Fig. 5 Schematic illustration of local gas-melt equilib- H2O and grey for CO2. b H2O–CO2 diagram illustrating
rium. a detail of the gas-melt interface region in a the evolution during decompression and degassing. The
concentration (Ci) versus distance diagram where Ci black bold curve is the equilibrium degassing trajectory.
refers to the volatile concentration in the melt. The gas The two black lines are isobars labelled with initial (Pinit)
bubble is on the left. The two curves are melt volatile and final (Pfin) pressures along the decompression path.
concentration profiles for H2O (black) and CO2 (grey) The dashed curve is the CO2/H2O isopleth passing
respectively, generated as a result of diffusion in the melt through Csol and it defines the composition of the gas
during decompression from Pinit to Pfin. Cinit give volatile bubble in local equilibrium with the interface melt.
concentrations of the pre-decompression melt, Cnew The shaded domain gives the range of possible melt
gas/melt interface volatile concentrations at Pfin and Csol compositions generated upon decompression from Pinit
volatile solubilities at Pfin. Black lettering is used for down to Pfin

are equal to their solubilities (Csol) at P = Pfin An alternative way to constrain the composi-
(intersection of the equilibrium degassing curve tion of gas bubbles is illustrated on Fig. 6. It
with the Pfin isobar, Fig. 5b). Note that diffusive starts from the observation that bubble nucleation
fractionation generates H2O and CO2 concen- is, from a kinetic point of view, an instantaneous
tration gradients within the melt (Fig. 5a), the process (e.g., Mourtada-Bonnefoi and Laporte
range of possible melt compositions during 2002, 2004). Nucleation of a gas bubble draws
decompression being represented by the dark volatiles from the local melt and the possibility
grey domain in Fig. 5b. The interface melt is the that the initial CO2/H2O of the gas bubble is the
only melt at equilibrium with the local gas at Pfin same as the local melt should be considered.
which has a CO2/H2O corresponding to the fluid According to this hypothesis, represented
isopleth on Fig. 5b (e.g., Dixon and Stolper schematically on Fig. 6, the local melt next to the
1995). For a low pressure (e.g., 25 MPa), the nucleated bubble (Cnew) is volatile-depleted but
local gas (e.g., a gas bubble nucleated at Pfin) is its CO2/H2O (r, Fig. 6a) is the same than the
relatively H2O-rich. In comparison, the pressure initial melt (Cinit). Melt and gas bubble compo-
integrated gas assuming bulk equilibrium sitions are thus both located on a mixing line
degassing from Pinit to Pfin would be necessarily between Cinit and Cnew which passes through the
less H2O-rich since most of the CO2 must have origin of the H2O–CO2 diagram (Fig. 6b). The
been outgassed from the melt. This gas is less net result is the nucleation of individual gas
H2O-rich than the pressure integrated gas pro- bubbles more CO2-rich than expected from local
duced by disequilibrium degassing from Pinit to gas-melt equilibrium (Fig. 5b).
Pfin (Fig. 3). Thus, individual bubbles nucleated The previous discussion emphasizes the com-
at Pfin can have CO2/H2O different from the positional variability of nucleated gas bubbles
composition of integrated gases generated con- and the need for their direct analysis in decom-
tinuously during decompression. pression experiments. Comparison between the
150 M. Pichavant et al.

(a) (b)
Ci Qjoju
volatile-depleted melt
H2O Nfmu! Cinit
dpnqptjujpot
r Cinit
CO2 Qgjo

CO2 melt
gas bubble

Cinit
Cnew
melt
Hbt!cvccmf!
DP30I3P

Cnew Csol
r
Frvjmjcsjvn
Cnew efhbttjoh
distance from bubble H2O melt
interface

Fig. 6 Hypothetical model for the composition of a gas pre-decompression melt, Cnew gas/melt interface volatile
bubble nucleated during decompression of H2O–, CO2- concentrations left after bubble nucleation at Pfin. Black
bearing basaltic melts. a detail of the gas-melt interface lettering is used for H2O and grey for CO2. b H2O–CO2
region in a concentration (Ci) versus distance diagram diagram illustrating the evolution during decompression
where Ci refers to the volatile concentration in the melt. and degassing. The black bold curve gives the schematic
The gas bubble is on the left. The two horizontal lines are location of the theoretical equilibrium degassing trajectory.
melt volatile concentrations for H2O (black) and CO2 The two black lines are isobars labelled with initial (Pinit)
(grey) drawn as straight lines because bubble nucleation is and final (Pfin) pressures along the decompression path.
an instantaneous event. The narrow zone near the interface Melts produced as a result of decompression and bubble
is the melt region depleted in volatiles drawn to form the nucleation plot on the straight line joining Cinit and Cnew.
bubble. Cnew are volatile concentrations in the depleted This line passes through the origin of the diagram because
melt region. Notice that the H2O/CO2 ratio (r) is identical in both Cinit and Cnew have the same CO2/H2O ratio. Note that
both the depleted and non-depleted melt regions because the location of Cnew along this line is arbitrary. The CO2/
during nucleation volatiles are drawn from the local melt H2O ratio of the gas bubble (dashed line) is also the same as
and the initial CO2/H2O of the gas bubble is the same as the Cinit and Cnew. It is higher than the gas bubble controlled by
local melt. Cinit give volatile concentrations of the local gas-melt equilibrium (Fig. 5)

CO2/H2O of nucleated bubbles and results of

gas-melt volatile partitioning models (e.g., Dixon 5 Discussion and Perspectives
and Stolper 1995; Papale et al. 2006) would for Gas Monitoring
provide a crucial test of the local gas-melt equi-
librium model (Fig. 5). If the nucleated bubbles 5.1 Degassing Processes
prove to be CO2-rich, then alternative models of
control of gas composition would be supported Experimental simulations show that, for ascent
(Fig. 6). One important aspect is that, in CO2- rates expected in natural systems, equilibrium
bearing systems, bubble nucleation during degassing occurs in pure H2O melts. In contrast,
decompression is continuous, occurring over a results for CO2-bearing melts conclusively
large pressure range (Le Gall and Pichavant demonstrate that degassing generates melt vola-
2016a, b). This is because CO2-bearing melts are tile concentrations out of equilibrium. The
volatile-supersaturated (Fig. 1) and, so, the driv- experimental database supporting this conclusion
ing force for nucleation of new bubbles is always has recently expanded. It now includes basaltic
present. Thus, decompression of CO2-bearing and rhyolitic melts, and S-bearing as well as
melts continuously leads to the nucleation of new S-free systems. Several of those experimental
bubbles, which increases the relevance of the decompression studies have been scaled to nat-
hypothetical mechanism illustrated in Fig. 6. ural systems so that results are realistic and
Gases as Precursory Signals: Experimental Simulations … 151

applicable. The decompression experiments on Disequilibrium degassing, as documented in

Stromboli basalt cover ascent rates ranging from the experiments, is the consequence of the
0.25 to 3 m/s (Pichavant et al. 2013; Le Gall and anomalous behaviour of CO2. CO2-super-
Pichavant 2016a, b; Le Gall et al. 2015a), well in saturated melts are systematically generated dur-
the range of current estimates for basaltic mag- ing decompression. The interpretation suggested
mas (e.g., Rutherford 2008; Peslier et al. 2015). by Pichavant et al. (2013) and quantitatively
An ascent rate approximately 10 times faster was formulated by Yoshimura (2015) is that, because
used by Yoshimura (2015). Therefore, in the case of its restricted diffusive mobility within the melt,
of basaltic magmas, equilibrium degassing CO2 has limited access to the gas phase for
should be viewed more as a reference situation timescales typical of magma ascent. However,
rather than as a general mechanism. This is quite our knowledge of volatile diffusivities in silicate
a change in paradigm which has major implica- melts is still very fragmentary. There are very few
tions for how gas signals are interpreted. diffusivity data for H2O and CO2 on the same
One remaining issue concerns the role of melt. S is another volatile which reputedly has a
crystals on bubble nucleation, heterogeneous slow diffusivity in silicate melts. Yet, the beha-
rather than homogeneous. All decompression viour of S during degassing differs from that of
studies considered in this paper were performed CO2, although we are still short of S solubility
on very crystal-poor, if not totally crystal-free, data for basaltic melts (e.g., Lesne et al. 2015).
melts and bubble nucleation appears to be mostly Acquisition of fundamental data (especially
homogeneous. Crystals present in experimental volatile diffusivity and solubility data for basaltic
basaltic products include Fe–Ti oxides (Le Gall melts) is needed for the elaboration of more
and Pichavant 2015b) and rare Fe sulphides (Le detailed interpretations of the decompression
Gall et al. 2015a). Le Gall and Pichavant (2015b) experiments. Future magma ascent models should
have documented heterogeneous nucleation of also incorporate the textural complexities asso-
bubbles on Fe–Ti oxide crystals (and also on Fe ciated with the vesiculation process.
sulphides, Le Gall et al. 2015a). Recently, Shea
(2017) has stressed the importance of magnetite
as a key mineral phase promoting heterogeneous 5.2 Gases as Unrest Signals
bubble nucleation in natural magmas. However,
Fe oxide phenocrysts and sulphides are uncom- Pre-eruptive changes in gas ratios have been
monly present in amounts exceeding a few observed at several basaltic volcanoes such as
vol.% in natural magmas. This led Le Gall Stromboli (Burton et al. 2007; Aiuppa et al.
and Pichavant (2015b) to conclude that 2010), Etna (Aiuppa et al. 2007) and Villarrica
heterogeneous bubble nucleation is not an (Aiuppa et al. 2017) among others. Transition
important mechanism in basaltic melts if driven from passive degassing to more explosive
by Fe oxides. Yet, heterogeneous nucleation on paroxysmal eruption regimes is marked by tem-
silicate phases is still an open question. For poral increases of the CO2/SO2 gas ratio in the
example, olivine, clinopyroxene and plagioclase volcanic plume. In some cases, the CO2/SO2
are typical phenocrysts and microlites in Strom- variations are correlated with a decrease of the
boli basalts (e.g., Pichavant et al. 2011). On the H2O/CO2 gas ratio (e.g., Aiuppa et al. 2017).
basis of limited textural evidence, Pichavant et al. These variations in volcanic gas ratios have been
(2013) ruled out the possibility of heterogeneous generally interpreted to reflect different pressures
nucleation of gas bubbles on clinopyroxene and of gas-melt segregation in the conduit, high CO2/
olivine crystals. However, additional investiga- SO2 (and low H2O/CO2) indicating deep condi-
tions seem warranted to guarantee full applica- tions and low CO2/SO2 (and high H2O/CO2)
bility of the decompression experiments above. shallow conditions (e.g., Edmonds 2008; Burton
152 M. Pichavant et al.

et al. 2007; Allard 2010; Aiuppa et al. 2017). In within the melt) than the same gases calculated
this approach, the pressure-dependent evolution assuming equilibrium with the melt (Fig. 3). This
of the gas phase exsolved upon magma ascent demonstrates the possibility of changing the gas
and decompression is calculated by using volatile ratios depending on the degassing mechanism
saturation codes (Newman and Lowenstern (equilibrium vs. disequilibrium). It is worth
2002; Moretti and Papale 2004; Papale et al. emphasizing that disequilibrium degassing asso-
2006; Burgisser et al. 2015). This implicitly ciated with CO2 retention produces integrated
assumes chemical equilibrium between gas and fluids that are less, not more, CO2-rich (Fig. 3).
melt, an assumption which, as shown above, is The CO2-rich gases observed on basaltic vol-
now largely questioned. If disequilibrium magma canoes have been generally attributed to
degassing is accepted, then the consequences for deep-seated processes such as fluxing of CO2 or
the interpretation of gas signals need to be arrival of CO2-rich magmas (e.g., Aiuppa et al.
examined. 2010, 2017; Allard 2010). In contrast, the
Firstly, one might argue that gas-melt equilib- degassing mechanism of Fig. 6 (although it needs
rium can persist at local scale, despite disequi- validation from direct analysis of gas bubbles in
librium at larger scales. Thus, volatile saturation decompression experiments) allows CO2-rich gas
codes could still be used and applied to local gas bubbles to be generated at low pressures. It also
and melt compositions, for example to model the provides an example of how gas ratios can be
composition of unconnected bubbles nucleated changed at constant pressure depending on the
within the melt. In contrast, volcanic gases nec- degassing mechanism. The initially CO2-rich
essarily require, to be sampled, that the magma is bubbles (Fig. 6) will probably shift rapidly with
permeable and, so, that the gas phase is connected. time toward lower CO2/H2O because of prefer-
It is quite possible that the gases sampled are ential diffusion of H2O from the melt. However,
mixtures of different components, either inte- nucleation is a continuous process in CO2-bearing
grated from several discrete degassing events basaltic melts (Le Gall and Pichavant 2016a, b)
along ascent or issued from different parts of the and reequilibration of previously nucleated bub-
plumbing system. Therefore, volcanic gases are bles by diffusion will be accompanied by the
probably more representative of integrated com- nucleation of new CO2-rich bubbles.
positions as discussed above than to compositions We conclude that future developments in the
of local gases. We have shown previously that interpretation of gas data require progress from
individual bubbles with compositions defined by both sides, experimental and volcanological.
local gas-melt equilibrium at a given pressure Some crucial experimental information at small
(Fig. 5) can have CO2/H2O different from inte- scale is still missing such as the composition of
grated gases generated continuously during individual gas bubbles nucleated in the decom-
decompression (Fig. 3). We conclude to the lim- pression experiments and the influence of crys-
ited applicability of local gas melt equilibrium to tals on bubble nucleation. In parallel, at larger
interpret volcanic gas ratios. scales, the representativity and the significance of
Secondly, disequilibrium gas-melt degassing the gas phase sampled on active basaltic volca-
due to CO2 retention within the melt implies that noes needs to be better demonstrated, for exam-
CO2/SO2 and H2O/CO2 gas ratios can no longer be ple by combining gas measurements with
directly related to pressures of gas-melt segrega- detailed textural studies of eruption products. It is
tion. Calculations using the diffusive fractionation expected that future work will narrow the gap in
model (Fig. 3) show that the pressure integrated scales between experiments and gas measure-
gases have a higher H2O/CO2 (and also presum- ments to refine interpretations of gas composi-
ably a lower CO2/SO2 because CO2 is retained tions as unrest signals.
Gases as Precursory Signals: Experimental Simulations … 153

Acknowledgements This paper has benefited from dis- at an andesite arc volcano. Geochem Geophys Geosyst
cussions with P. Allard, C. Martel, N. Metrich, 11:Q04005. https://doi.org/10.1029/2009GC002781
A. Bertagnini, R. Moretti, P. Papale and M. Pompilio, Fiege A, Behrens H, Holtz F, Adams F (2014) Kinetic vs.
reviews by R. Brooker and F. Wadsworth and from thermodynamic control of degassing of H2O–
editorial comments by B. Scheu. Discussion with S ± Cl-bearing andesitic melts. Geochim Cosmochim
S. Yoshimura was helpful. The VUELCO consortium Acta 125:241–264
provided a scientifically demanding and interdisciplinary Gonnermann HM, Manga M (2005) Non-equilibrium
forum for the elaboration of ideas developed in this study. magma degassing: results from modelling of the ca.
The Ph.D. thesis of NLG was supported by the VUELCO 1340 AD eruption of Mono craters, California. Earth
project. Planet Sci Lett 238:1–16
Head JW III, Wilson L (2003) Deep submarine pyroclas-
tic eruptions: theory and predicted landforms and
deposits. J Volc Geotherm Res 121:155–193
References Le Gall N, Pichavant M (2015b) Heterogeneous bubble
nucleation on Fe–Ti oxides in H2O- and H2O-CO2-
Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, bearing basaltic melts. (In preparation)
Liuzzo M, Papale P, Shinohara H, Valenza M (2007) Le Gall N, Pichavant M (2016a) Homogeneous bubble
Forecasting Etna eruptions by real-time observation of nucleation in H2O- and H2O-CO2-bearing basaltic
volcanic gas composition. Geology 35:1115–1118 melts: results of high temperature decompression
Aiuppa A, Bertagnini A, Metrich N, Moretti R, Di experiments. J Volcanol Geotherm Res 327:604–621
Muro A (2010) A model of degassing for Stromboli Le Gall N, Pichavant M (2016b) Effect of ascent rate on
volcano. Earth Planet Sci Lett 295:195–204 homogeneous bubble nucleation in the system
Aiuppa A, Bitetto M, Francofonte V, Velasquez G, basalt-H2O-CO2: Implications for Stromboli volcano.
Bucarey Parra C, Giudice G, Liuzzo M, Moretti R, Am Mineral 101:1967–1985
Moussallam Y, Peters N, Tamburello G, Valder- Le Gall N, Pichavant M, Di Carlo I, Scaillet B (2015a)
rama OA, Curtis A (2017) A CO2-gas precursor to the Sulfur partitioning between melt and fluid during
March 2015 Villarrica volcano eruption. Geochem degassing of ascending C-O-H-S-bearing basaltic
Geophys Geosyst 18:2120–2132 magma: an experimental study. In preparation
Allard P (2010) A CO2-rich gas trigger of explosive Lensky NG, Niebo RW, Holloway JR, Lyakhovsky V,
paroxysms at Stromboli basaltic volcano, Italy. J Vol- Navon O (2006) Bubble nucleation as a trigger for
canol Geotherm Res 189:363–374 xenolith entrapment in mantle melts. Earth Planet Sci
Aubaud C, Pineau F, Jambon A, Javoy M (2004) Kinetic Lett 245:278–288
disequilibrium of C, He, Ar and carbon isotopes Lesne P, Scaillet B, Pichavant M, Iacono-Marziano G,
during degassing of mid-ocean ridge basalts. Earth Bény J-M (2011) The H2O solubility of alkali basaltic
Planet Sci Lett 222:391–406 melts: an experimental study. Contrib Mineral Petrol
Burgisser A, Alletti M, Scaillet B (2015) Simulating the 162:133–151
behavior of volatiles belonging to the C-O–H–S Lesne P, Scaillet B, Pichavant M (2015) The solubility of
system in silicate melts under magmatic conditions sulphur in hydrous basaltic melts. Chem Geol
with the software D-Compress. Comput Geosci 418:104–116
79:1–14 Metrich N, Bertagnini A, Di Muro A (2010) Conditions of
Burton M, Allard P, La Spina A, Murè F (2007) magma storage, degassing and ascent at Stromboli:
Magmatic gas composition reveals the source depth new insights into the volcanic plumbing system with
of slug-driven strombolian explosive activity. Science inferences on the eruptive dynamics. J Petrol
317:227–230 51:603–626
Christopher T, Edmonds M, Humphreys MCS, Herd RA Moretti R, Papale P (2004) On the oxidation state and
(2010) Volcanic gas emissions from Soufrière Hills volatile behaviour in multicomponent gas–melt equi-
Volcano, Montserrat 1995–2009, with implications for libria. Chem Geol 213:265–280
mafic magma supply and degassing. Geophys Res Lett Mourtada-Bonnefoi CC, Laporte D (2002) Homogeneous
37: L00E04. https://doi.org/10.1029/2009gl041325 bubble nucleation in rhyolitic magmas: an experimen-
Dixon JE, Stolper EM (1995) An experimental study of tal study of the effect of H2O and CO2. J Geophys Res
water and carbon dioxide solubilities in mid-ocean 107(B4):2066. https://doi.org/10.1029/2001jb000290
ridge basaltic liquids. Part II. Applications to degas- Mourtada-Bonnefoi CC, Laporte D (2004) Kinetics of
sing. J Petrol 36:1633–1646 bubble nucleation in a rhyolitic melt: an experimental
Edmonds M (2008) New geochemical insights into vol- study of the effect of ascent rate. Earth Planet Sci Lett
canic degassing. Philos Trans R Soc A 366:4559–4579 218:521–537
Edmonds M, Aiuppa A, Humphreys M, Moretti R, Newman S, Lowenstern JB (2002) VolatileCalc: a silicate
Giudice G, Martin RS, Herd RA, Christopher T (2010) melt–H2O–CO2 solution model written in Visual
Excess volatiles supplied by mingling of mafic magma Basic for Excel. Comput Geosci 28:597–604
154 M. Pichavant et al.

Papale P, Moretti R, Barbato D (2006) The compositional Shea T (2017) Bubble nucleation in magmas: A domi-
dependence of the saturation surface of H2O + CO2 nantly heterogeneous process? J Volcanol Geotherm
fluids in silicate melts. Chem Geol 229:78–95 Res 343:155–170
Peslier AH, Bizimis M, Matney M (2015) Water Sparks RSJ, Barclay J, Jaupart C, Mader HM, Phillips JC
disequilibrium in olivines from Hawaiian peridotites: (1994) Physical aspects of magmatic degassing I.
recent metasomatism, H diffusion and magma ascent Experimental and theoretical constraints on vesicula-
rates. Geochim Cosmochim Acta 154:98–117 tion. In: Carroll MR, Holloway JR (eds) Volatiles in
Pichavant M, Pompilio M, D’Oriano C, Di Carlo I (2011) Magmas. Mineralogical Society of America Reviews
The deep feeding system of Stromboli, Italy: insights in Mineralogy, vol 30, pp 413–445
from a primitive golden pumice. Eur J Miner Yoshimura S (2015) Diffusive fractionation of H2O and
23:499–517 CO2 during magma degassing. Chem Geol 411:172–181
Pichavant M, Di Carlo I, Rotolo SG, Scaillet M, Burgis- Zhang Y, Ni H (2010) Diffusion of H, C, and O
ser A, Le Gall N, Martel C (2013) Generation of CO2- components in silicate melts. In: Zhang Y, Cher-
rich melts during basalt magma ascent and degassing. niak DJ (eds) Diffusion in minerals and melts.
Contrib Mineral Petrol 166:545–561 Mineralogical Society of America Reviews in Miner-
Rutherford MJ (2008) Magma ascent rates. In: alogy vol 72, pp 171–225
Putirka KD, Tepley F (eds) Minerals, inclusions and
volcanic processes. Mineralogical Society of America
Reviews in Mineralogy vol 69, pp 241–271

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter are
the Creative Commons Attribution 4.0 International included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Crystals, Bubbles and Melt:
Critical Conduit Processes
Revealed by Numerical Models

M. E. Thomas, J. W. Neuberg and A. S. D. Collinson

Abstract volatile content. Therefore, understanding the

Understanding how magma moves within a processes controlling the volatile content within
conduit is an important question that is still the conduit system and the outgassing of these
poorly understood. In particular, estimation of volatiles is crucial to understanding and pre-
the magma ascent rate is key for interpreting dicting potential unrest or eruption scenarios.
monitoring signals and therefore, predicting
volcanic activity. This relies on understanding
how strongly different magmatic processes Keywords
occurring within the conduit control the ascent

Numerical modelling Conduit processes
rate. These processes are controlled by changes

Low frequency earthquakes Magma flow
in magmatic parameters such as the water Magma ascent rate
content or temperature and understanding/
linking changes of such parameters to moni-
toring data is an essential step in the use of
these data as a predictive tool. The results
presented here are from a suite of conduit flow
1 Introduction
models based on Soufrière Hills Volcano,
Montserrat, that assesses the influence of
A volcanic conduit provides the pathway for
individual model parameters. By systematically
transport of magma and magmatic fluids within a
changing these parameters, the results indicate
volcano. It is possible to detect both this move-
that changes in conduit diameter and excess
ment and the occurrence of conduit processes
pressure in the magma chamber are amongst
through geophysical monitoring techniques as
the dominant controlling variables. However,
discussed in Chap. 2 of this book. However, the
the single most important parameter controlling
extent to which changes in magma flow proper-
variations in the magma ascent rate is the
ties affect the data recorded on volcanoes is not
well understood. Is it possible that a small change
in magma temperature or water content could
M. E. Thomas (&)
J. W. Neuberg alter the processes or flow within the conduit
A. S. D. Collinson
enough to be recorded by geophysical monitor-
School of Earth and Environment, Institute
of Geophysics and Tectonics, University of Leeds, ing instruments or simple visual observation?
Leeds LS12 9JT, UK What effect does the size of gas bubbles within
e-mail: [email protected]

Advs in Volcanology (2019) 155–169

DOI 10.1007/11157_2018_36
© The Author(s) 2018
Published Online: 30 August 2018
156 M. E. Thomas et al.

the magma have on the overall flow dynamics, Soufrière Hills Volcano, Montserrat—a long
and how big do these changes need to be to alter lived andesitic dome forming eruption (Sparks
the eruption style? These types of question are et al. 2000; Wadge et al. 2014) and excellent
addressed within this chapter in an attempt to natural laboratory. While these initial models are
identify the crucial parameters that cause changes based on extrusive eruptions, the results of
in observed volcanic behaviour. changing the model parameters have the poten-
We use conduit flow models to analyse the tial to alter the eruption style to either more
key input parameters that control magma flow violent or gentle forms and it is noted that the
properties, such as the magma water content, underlying principles discussed here are appli-
crystal content and conduit geometry, to assess cable to other volcanic systems, including those
their relative importance to the overall magma exhibiting signs of unrest, yet to develop into a
flow dynamics. A list of all input parameters is full blown eruption. If we can develop the find-
presented in Table 1 along with the range of ings presented in this chapter into the creation of
values studied. We focus on evolved silicic threshold levels for recorded geophysical data, it
magmatic systems because of the wealth of rel- may become possible to begin to predict when a
evant monitoring information and previous volcano will evolve from a state of unrest to
numerical modelling attempts relating to eruption.

Table 1 Parameters used in the reference model and range of parameter variations
Symbol/abbreviation Variable “Reference” model value Range of
modelled values
– The melt composition Rhyolitic (>71% SiO2) See Table 2
(Barclay et al. 1998)
bni Bubble number density 1010 m−3 (Cluzel et al. 2008) 107–1011 m−3
DTBL Thickness of thermal boundary layer 0.3 m (Collier and Neuberg 0.3–0.5 m
over which Tdiff is lost 2006)
C Bubble surface tension 0.06 N m−1 (Lyakhovsky 0.05–0.25 N m−1
et al. 1996)
vc Magma chamber crystal volume fraction 40% (Barclay et al. 1998) 40–50%
Ls Slip length of brittle failure of melt 0.01 0.01–1.0 m
Pe Excess chamber pressure above 0 MPa 0–20 MPa
lithostatic
Ptop Pressure at conduit exit 0.09 MPa 0.09–4.5 MPa
qc Average density of crystal assemblage 2700 kg m−3 (Burgisser et al. 2550–3200 kg
2010) m−3
qm Density of pure melt 2380 kg m−3 (Burgisser et al. –
2010)
T Magma temperature 1150 K(Devine et al. 2003) 1100–1150 K
Tdiff Amount of cooling at conduit wall 200 K (Collier and Neuberg, 100–200 K
2006)
ss Melt shear strength – 105–107 Pa
W% Initial dissolved water content of magma 4.5 wt% (Barclay et al. 1998) 3–8 wt%
w, d, r Variables that define the conduit shape See Fig. 1 See Fig. 2
and size
Crystals, Bubbles and Melt: Critical Conduit Processes … 157

2 The Model 2.1 Governing Equations

In order to assess the effect of altering the model Conduit flow is computed with a finite element
parameters, a standard or “reference” model is approach within the code COMSOL Multi-
defined. This reference model is based on data physics®, and modelled in an axial symmetric
available in the literature that refers to Soufrière domain space through the compressible formu-
Hills Volcano, and is outlined in Fig. 1 and lation of the Navier-Stokes equation:
Table 1. The general dimensions of the modelled
conduit, shown in Fig. 1 are inferred from geo- @u

q þ qu
ru ¼ rp þ r
fg ru þ ðruÞT
chemical and observational data from Soufrière @t
2
Hills Volcano (Barclay et al. 1998; Sparks et al.  g½r
uIg þ F ð1Þ
2000), placing minimum depth constraints of 3
and the continuity equation:
5–6 km for the position of the magma cham-
ber and width estimates of 30–50 m for the @q
conduit. þ r
ðquÞ ¼ 0 ð2Þ
@t

Fig. 1 Cartoon of the modelled volcanic system. Bubble nucleation and brittle failure depth vary with the model
parameters considered
158 M. E. Thomas et al.

where q is density, u the velocity vector, p the M

pressure, η the dynamic viscosity and F the n¼ ; ð5Þ
m
volume force vector (gravity). There is no time
dependency in the model as they are solved to a where M is the molar mass of water and m is the
@p
steady state, so the terms q @u
@t and @t in Eqs. (1)
mass of water present. Thus, combining Eqs. (7)
and (2) are neglected. and (8) and considering a unit volume we get:
mp
qg ¼ ð6Þ
RT
2.2 Magma Composition
In the reference model, a single magma tem-
The properties of the magma are modelled as the perature is used with the exception of the tem-
averaged properties of its constituents: melt, perature across a thermal boundary layer
crystals and gas. For the reference model, the (TBL) defined adjacent to the conduit wall.
general composition of the melt is taken as rhy- A linear temperature drop is applied across the
olitic, using the groundmass analysis of TBL, to simulate the cooling of the magma
Montserrat dome rocks undertaken by Barclay abutting the country rock in a well-established
et al. (1998). However, several melt composi- conduit (Collier and Neuberg 2006). The gas
tions, were ultimately considered to assess the volume fraction (vg) is calculated by determining
effect of melt composition on the modelled how much water remains dissolved within the
eruption dynamics (Table 2). Crystal content (vc) melt at a particular pressure using the solubility
and density (qc) are fixed as we assume a con- of H2O in rhyolitic melts presented by Liu et al.
stant temperature and that the conduit ascent (2005). At significant pressures, all the water is
times are orders of magnitudes faster than the dissolved within the melt fraction and vg is ini-
time required for crystal growth by decompres- tially zero. However, as the pressure decreases,
sion, meaning only the phenocrysts present in the water begins to exsolve out of the melt and forms
magma chamber are accounted for and growth of bubbles. The absolute volume of exolved gas
microlites and microphenocrysts is not consid- (V) can be calculated from the ideal gas law (4).
ered. The expression for the bulk density of the This absolute volume of gas is then used to
magma is given by: calculate the gas volume fraction of the bulk
    magma constituted by the gas phase.
q ¼ qm vm 1  vg þ qg vg þ vc qc 1  vg ; ð3Þ

where vm is the initial fraction of melt (1 − vc),

2.3 Magma Viscosity—The
qm is the melt density and vg is the gas volume
Contribution of Crystals,
fraction (Table 1). For the gas phase, water is
Bubbles and Melt
assumed as the only volatile species present and
the gas density (qg) is calculated directly from
The bulk magma viscosity (η) is determined by
the ideal gas law with the assumption that bubble
first calculating the viscosity of the pure melt
growth in is equilibrium with the conduit
phase (ηm). This is done using a model for the
pressure:
viscosity of magmatic liquids presented by
pV ¼ nRT; ð4Þ Giordano et al. (2008), that predicts the viscosity
of silicate melts as a function of temperature and
where V is the volume of gas, R the ideal gas melt composition. It is important to note that the
constant and T the temperature. The number of composition used in the viscosity model is that of
moles of water, n, is related to density by: the pure melt phase (rhyolitic) not the overall
Crystals, Bubbles and Melt: Critical Conduit Processes … 159

Table 2 Compositions of melt used in the numerical simulations

Compositiona SiO2 Al2O3 TiO2 FeO MgO MnO CaO Na2O K2O
a 71.41 13.58 0.28 2.78 1.64 0.13 4.86 3.73 1.6
b 76.97 11.21 0.29 1.89 0.26 0.12 1.29 4.07 2.37
c 77.10 9.83 0.18 1.17 0.22 0.10 1.52 4.14 1.72
d 78.66 11.20 0.39 1.93 0.30 0.10 1.48 3.57 2.38
a
Compositions determined through, (a) rastered electron microprobe analysis of groundmass (Barclay et al. 1998);
(b) Matrix glass composition (Rutherford and Devine 2003); (c) Quartz hosted melt inclusion (Devine et al. 1998);
(d) Cameca SX50 microprobe analysis of interstitial glass (Burgisser et al. 2010). All melts are rhyolitic and
composition (a) is used in the defined reference model

magma composition. The whole rock composi- Manga, 2005). Whether a bubble is in an
tion of recent Soufrière Hills Volcano magma is un-deformed or deformed state is represented by
andesitic (Edmonds et al. 2010), but this includes the capillary number:
the contribution of the crystals. The viscosity
model is only used to calculate the actual vis- gm rE
Ca ¼ ð8Þ
cosity of the liquid component (the melt), on C
which the crystals (the solid) have no bearing.
where r is the un-deformed bubble radius, C, the
When the effect of crystals within the melt is
bubble surface tension and E, a function of the
considered, the effective viscosity of the melt
strain rate within the magma flow defined below.
(liquid) and crystal (solid) mixture (ηmc) increa-
If Ca > 1 then the bubbles can be considered
ses, and can be represented by the Einstein-
deformed. The value of Ca will vary as a function
Roscoe equation:
of shear strain rate and elongation strain rate
  (Thomas and Neuberg 2012), meaning bubbles
vc 2:5
gmc ¼ gm 1  max ; ð7Þ can be deformed within the model through either
vc
shear or extension. To account for strain acceler-
ation or deceleration the dynamic capillary num-
where vc max is the volume fraction of crystals at
ber (Cd) is required (Llewellin and Manga 2005).
which the maximum packing is achieved and a
This compares the timescale over which the bub-
commonly adopted value for this is 0.6 (Marsh,
bles can respond to changes in their strain envi-
1981), which is used within this study. Although
ronment with the timescale over which the strain
this value was proposed for randomly packed
environment changes. If this value is large, the
spheres, and it has been shown by Marti et al.
flow is termed unsteady and the bubbles are unable
(2005) that vc max tends to decrease as the particle
to deform independently in response to the flow.
(crystal) shape becomes less isotropic. Ishibashi
However, for the models considered here, condi-
(2009) demonstrated that this value is a good
tions of unsteady flow are found only in a very
approximation as the effect of particle shape on
small area near the exit of the conduit. Accounting
vc max is offset by effects of size heterogeneity and
for this within the models resulted in no noticeable
crystal alignment.
change in the derived flow parameters, hence the
The presence of bubbles also affects the vis-
computation of Cd is not considered.
cosity. If the bubbles within the magma remain
Depending on the value of Ca, η is calculated
un-deformed they act to increase viscosity, whilst
using the suggested ‘minimum variation’ of
if deformed (elongated in the direction of flow),
Llewellin and Manga (2005):
they act to decrease visosity (Llewellin and
160 M. E. Thomas et al.

(  1 shear strength (ss), and has been represented as a

\1 g ¼ gmc 1  vg
Ca ¼  5=3 ð9Þ brittle failure criterion (e.g. Tuffen et al. 2003):
[1 g ¼ gmc 1  vg
g_e
[1 ð13Þ
By assuming the homogeneous nucleation of ss
a number of bubbles in a unit volume of melt as a
single event, which is determined experimentally where e_ is the shear strain rate. This criterion
through the initial bubble number density (bni) holds true under the assumption that during
(e.g. Hurwitz and Navon 1994), the bubble un-relaxed deformation the accumulation of shear
radius (Lensky et al. 2002) is given by: stress in the melt obeys the Maxwell model:
" #1=3 g @rs
S0 3 qm ðC0  Cm Þ rs ¼ ¼ g_e ð14Þ
r¼ ; ð10Þ l @t
qg
where rs is the shear stress and l the shear modulus.
where C0 and Cm are the initial and remaining The magma composition as discussed in
amount of water dissolved in the melt respec- Sects. 2.2 and 2.3 is considered without the
tively and S0 is the initial size of the melt shell effects of microlite growth, the reasons for doing
from which each bubble grows. S0 is related to so are outlined in Sect. 2.2. However, it is worth
the instantaneous bubble number density (bn) noting, that if considered, the influence of
through the expression: microlite growth would possibly increase the
bulk viscosity significantly.
3
S0 3 ¼ : ð11Þ
4pbn

bn is used rather than the initial value (bni) 2.5 Boundary Conditions
because homogeneous nucleation is assumed.
Therefore, the bubble number density must Flow within the system is driven by a pressure
remain constant with respect to the volume of the gradient defined by boundary conditions at the
melt fraction in Eq. 3. This also accounts for top and bottom of the conduit. The top boundary
bubble coalescence and bn is given by: is set to atmospheric pressure at the altitude of the
conduit exit plus any overburden load from an
bni
  emplaced lava dome. The bottom boundary is set
bn ¼ vm  1  v g ð12Þ
vm to lithostatic pressure (assuming a homogeneous
country rock density of 2600 kg m−3) plus any
imposed overpressure (Pe). Both the top and
bottom pressure conditions are held constant
2.4 Brittle Failure of Melt
throughout the model run. Initial boundary con-
ditions along the length of the conduit are defined
It is now well established that magma, or more
as no slip. When brittle failure of melt is con-
specifically the melt component of a magma can
sidered within a model run, at the regions of the
fail in a brittle manner (e.g. Goto 1999). This is
conduit wall where the brittle failure criterion was
likely to generate low-frequency (LF) earthquakes
exceeded, the boundary conditions are changed to
(e.g. Neuberg et al. 2006) and effect the overall
a tangential slip velocity (Du) defined by:
flow dynamics. In order to account for these
effects, it is necessary to define conditions under 1
which the melt may fracture. Shear failure of melt Du ¼ rs ; ð15Þ
b
occurs when the shear stress ðg_eÞ exceeds the
Crystals, Bubbles and Melt: Critical Conduit Processes … 161

where rs is the tangential shear stress to the 3 Critical Conduit Processes

conduit wall and the coefficient b is a function of
the slip length (Ls): 3.1 Using Magma Ascent Rates
to Assess Model Sensitivity
g
b¼ ; ð16Þ
Ls The ascent rate is a key parameter in under-
standing volcanic hazard because it has been
The model is then re-run to account for the
directly linked to eruptive behaviour (e.g.
effect of changing boundary conditions at the
Gonnerman and Manga 2007). By gaining a
conduit walls. Where this results in an increase in
better understanding of which model parameters
the predicted failure depth, an iterative approach
have the greatest effect on ascent rates, we can
is used and the model is re-run with the new
achieve an insight into which are the most
depth until the depth at which brittle failure of
important parameters controlling explosivity, and
the melt stabilises. For the purposes of this study,
the likely severity of the volcanic hazard. For the
the failure depth is considered to have converged
purpose of comparing the various models we use
if the depth increase between iterative runs is less
two velocities, defined as V and V2500, where V is
than 10% of the previous observed increase.
the average accent velocity taken along a vertical
The size and shape of the conduit is also an
profile through the centre of the conduit, and
important factor in influencing the accent rate, so
V2500 is the average accent velocity taken along a
for the purpose of assessing the potential mag-
horizontal profile at a depth of 2500 m within the
nitude of this influence, several possible conduit
conduit.
shapes were modelled. While the types of
The ascent rate has also been linked to mon-
boundary conditions discussed above do not
itoring data such as seismicity (e.g. Thomas and
change, the relative locations of the boundaries
Neuberg 2012) or deformation (e.g. Zobin et al.
do (Fig. 2). Case (a) is the simplest geometry
2011), therefore, it is possible to link the changes
change and represents just a change of the con-
in model parameters to recorded monitoring data.
duit radius (r). Case (b) represents a narrowing of
In addition, there are physically observed varia-
the conduit. Case (c) represents a widening of the
tions in ascent rate estimated from a variety of
conduit. The extent to which the geometry of the
methods, ranging from studying mineral reaction
conduit is changed within the models is dis-
rims around phenocrysts within erupted magma
cussed further in the next Section.

Fig. 2 Schematic diagram of the alternative conduit geometries modelled, showing a a constant conduit radius; b a
narrowing conduit and c a widening conduit
162 M. E. Thomas et al.

(Rutherford and Devine 2003) to interpreting tension. Modifying the parameters involved in
lava dome morphology (Sparks et al. 2000). This the brittle failure of the melt (magma shear
places constraints on the magnitude of changes to strength and slip length) has a negligible effect
the modelled ascent rate engendered by altering on ascent rates and these results have not been
the model input parameters we can consider plotted on Fig. 3. However, the contribution of
realistic. the brittle failure of the melt to observed geo-
Matching the absolute values of physically physical signals is considered very important,
observed and calculated ascent rates is currently and will be discussed in Sect. 3.3.
beyond the scope of the model, however we can It is unsurprising that the group of model
use the magnitude of the observed variations to parameters that appear to have the greatest effect
provide upper and lower bounds to the extent to on the magma ascent velocity, as seen in Fig. 3b
which the model input parameters are varied. (water content, temperature, crystal content, and
Any changes that produce increases in ascent rate chemical composition), also have the greatest
greater than two orders of magnitude over the effect on the magma viscosity (Sect. 2.3). Ulti-
reference model are not considered realistic in mately, modelling the ascent of magma is a fluid
this work. This may seem at first an arbitrary flow problem, and the properties that have the
discrimination, but there is a good reason biggest effect on the fluid (magma) properties
that the observed or calculated ascent rates pre- will have the biggest effect on the overall
sented in the literature (e.g. Rutherford and dynamics of the system. All other parameters
Devine 2003; Castro and Gardner 2008) are have a much smaller direct effect on the fluid
“slow” (<5  10−2 m s−1). Faster ascent rates, properties, and although they may be important
while likely to exist in nature, would almost to specific small scale magmatic processes when
certainly result in substantial fragmentation of considered in isolation, with respect to the
the magma, making it very difficult to observe or overall magma ascent they appear insignificant.
calculate the actual magma ascent rate below the For example, altering the properties of the bub-
initial point of fragmentation. Fragmentation bles within the magma, bni and C, the effect is to
dynamics are not considered within the current change the shape and the number of bubbles.
model, hence no valid inferences or conclusions Previous work has heavily focused on this area
can be gained from studying the model runs that (e.g. Llewellin and Magna 2005) but the effect on
exhibit extremely fast ascent rates. the overall flow modelled here is minimal. The
indication from this is that it is the total volatile
content (water in this case) which is available
3.2 The Critical Model Parameters that is more important to governing the overall
flow dynamics, rather than how exactly it is
Figure 3 summarises the sensitivity of ascent rate stored in the magma. This particular observation
to the different model parameters presented in is a key point as new estimates from Cassidy
Table 1. The single parameter (within the mod- et al. (2015) suggest that basaltic South Soufrière
elled ranges) which has the strongest effect on the Hills magmas (and by extension, possibly other
ascent velocities is the initial dissolved water basaltic arc magmas) have the potential to be
content of the magma. This parameter affected extremely volatile-rich, containing up to >6 wt%
both V and V2500 to a large degree. In contrast H2O prior to eruption. Firstly, this validates the
there are several model parameters which have use of high initial water contents used in the
little effect on the modelled ascent velocities. range of parameters modelled, and secondly,
These include the thermal boundary layer thick- given the range of accent velocities generated
ness and the temperature drop across it, as well as within the models as a result of just changing the
the bubble number density and bubble surface dissolved water content (the dark blue bars in
Crystals, Bubbles and Melt: Critical Conduit Processes … 163

Fig. 3 a The ascent rate, V for each value of the changes that caused increases in the ascent rates plot
parameters altered. The black bar represents the reference upward from the horizontal line, while parameter changes
model with parameters as listed in Table 1. b The same that caused decreases in ascent rates plot downwards.
data as (a) plotted relative to the reference model, which c V2500 for each value of the parameters altered, plotted
is represented by the horizontal black line. Parameter relative to the reference model

Fig. 3), demonstrates that it is vitally important parameters modelled which have the largest
to obtain an accurate understanding of the influence on the modelled ascent rate are the
magma components at the volcano of interest, chamber overpressure and the conduit geometry.
rather than assuming “typical” values represen- These are particularly important points when
tative of a broad compositional category. considering volcanoes entering periods of unrest
Outside of water content, temperature, crystal following long periods of quiescence. It is
content, and chemical composition, the two problematic to achieve an accurate understanding
164 M. E. Thomas et al.

of the magma components highlighted above at value was chosen because it has been highlighted
all volcanoes under these circumstances due to a by Rutherford and Devine (2003) as an ascent
likely lack of monitoring (a problem highlighted rate which may indicate a transition between
in Parts 1 and 2 of this book). Unless there has effusive and explosive behaviour. This value has
been long-term measurement of deformation been obtained from quantifying the breakdown
occurring at the volcano now exhibiting signs of hornblende in ascending magma, and while
unrest, it will be extremely difficult to estimate this technique is not an accurate barometer for
any likely overpressures in the chamber, and defining an exact ascent velocity required for
attempting to define the conduct geometry of a explosive eruptions, the rates calculated for
system that has not yet erupted would be almost non-explosive eruptive activity at Soufrière Hills
impossible. It is therefore paramount that as volcano between the period of November 1995–
much information as possible of all potentially September 2002 were below this value. Figure 4
active volcanic systems is routinely gathered shows that several model runs produced ascent
before signs of unrest are detected. rates of <0.02 ms−1 (by altering the melt com-
position or vc) and several other runs produced
ascent rates very close to this value (by altering
3.3 Matching Observations— W%, T, and conduit geometry), indicating that by
Explosivity and Seismicity altering just single parameters within the system
this theoretical threshold of accent rate can be
Although, as previously mentioned, matching the crossed.
absolute values of physically observed and cal- Conduit flow is treated as a closed system, so
culated ascent rates is currently beyond the scope no outgassing is considered in the model, as a
of the model, key to giving the models real sig- result the ascent velocities are overestimated
nificance is determining whether the changes to (Thomas and Neuberg 2014). It is therefore
important parameters highlighted in Sect. 3.1 can predictable that if this process was included, far
be theoretically linked to physical observations at more of the model runs would result in ascent
real volcanic systems. Figure 4 shows values of velocities that straddle the baseline in Fig. 4.
V2500 for all of the modelled parameters in a This suggests that the ability for the ascent rate
manner similar to that presented in Fig. 3c, but in within the conduit to fluctuate, either side of
this case, the data are plotted relative to a base- values that have been linked to explosive erup-
line accent value of 0.02 ms−1. This base line tions in response to small changes in the system

Fig. 4 The ascent rate, V2500 for each value of the parameters altered, plotted relative to an ascent velocity of
0.02 ms−1
Crystals, Bubbles and Melt: Critical Conduit Processes … 165

parameters is genuine. The requirement to accu- which since mass must be conserved also has the
rately understand and model outgassing pro- effect of increasing the flow velocity.
cesses (see Sect. 4) is therefore an important To resolve this discrepancy between model
capacity that is currently lacking. and observations we introduce a constriction
One major discrepancy between physical within the conduit as a plausible explanation for
observations and the model is the model results brittle failure at greater depths. We test its effect
suggested that the brittle failure of the melt within the reference model by including a bot-
(Sect. 2.4) and the related LF seismicity would tleneck region at a depth of 1500 m, reducing the
occur as a shallow process (in agreement with the conduit diameter from 15 to 10 m. This bottle-
work of Holland et al. (2011)).The physical neck is 100 m in length, which equates to only
observations place the location of this type of 1/50 of the total conduit length. Figure 5 shows
seismicity at Soufrière Hills at depths of *1500 ascent velocity and shear strain rate profiles from
m below the conduit exit (Neuberg et al. 2006). the bottleneck region compared to values from
This is because under normal conditions the shear the same location of the conduit in the unmodi-
stress required to break the melt (Eq. 14) can only fied reference model. By altering this relatively
be reached where the melt is extremely viscous, small region of the conduit, the shear strain rate
which occurs near the surface. In order to reach increases by a factor of four. Crucially, with the
the higher shear stresses required to break the exception of small changes in the magma rheol-
melt at greater depths, where the viscosity is ogy caused by the induced pressure gradients
lower, the shear strain rate ð_eÞ needs to increase. within the bottleneck, the magma viscosity has
Since e_ is equal to the lateral velocity gradient not been altered. Due to the increased value of
within a cylindrical conduit or dyke: shear strain rate the brittle failure ratio (13) will
increase by the same factor. By introducing such
dv asperities into the conduit and increasing the
e_ ¼ ð17Þ
dx strain rate it is possible to drive the brittle frac-
ture of the melt to deeper levels in the conduit
the simplest way to increase e_ is to increase the
that match the location of recorded LF seismicity
velocity of the magma flowing within the con-
at Soufrière Hills volcano. This further
duit, or reduce the area through which it flows,

Fig. 5 a Plot of the shear strain rate within a simple taken at the same depth for velocity and shear strain
bottleneck of 100 m length at intervals of 0.005. The rate for the case of the unmodified reference model
values of shear strain rate are seen to be increases over the (solid line) and a conduit containing a bottleneck (dashed
entire length of the bottleneck. b Cross conduit profiles line)
166 M. E. Thomas et al.

emphasises the importance of understanding the k

possible conduit geometry previously high- u¼ ðrP þ qgrzÞ ð18Þ
gg
lighted in Sect. 3.2. This is just one possible
solution of generating seismicity deeper in the @
ðqeÞ þ r
ðquÞ ¼ 0 ð19Þ
conduit. Any process that acts to increase the sear @t
strain rate, such as strain localisation between " #
@ k
crystals in a magma with a high crystal fraction ðqeÞ þ r
q  ðrP þ qgrzÞ ¼ 0 ð20Þ
(as discussed in Chap. 10 of this book) would be @t gg
a possible explanation.
The gas velocity (u) is then determined by using
the pressure gradient within Darcy’s law (18).
4 Pathways for Outgassing In contrast, with the models for ascent rate,
time dependency is included in this model in order
The role of volatiles has been identified as the to understand the changes in the system through
primary controlling parameter governing ascent time, in response to a permeability (k) increase in
rate within the conduit, but what was not con- response to brittle failure at the conduit-wall
sidered was that the gas, once exsolved from the margin. Due to its abundance in volcanic systems,
melt has the ability move independently to the water is the only volatile considered here. The gas
melt. This is a commonly assumed process in density (q) is calculated using the mean molar
basaltic systems, but not in silicic magmas of a mass within the ideal gas law (6), and the gas
relatively high viscosity as it is assumed that viscosity (l) assumed constant at 1.5  10−5 Pa
bubbles of gas cannot rise with significant speed (Collinson and Neuberg 2012).
within the magma. However fractures generated Bulk permeabilities are set such that the
by brittle failure of the melt (Sect. 3.3) conduit has a higher permeability (10−10 m2)
may provide ideal outgassing pathways for than the wallrocks (10−12 m2), and initially, the
exsolved gas. conduit-wall margin is “sealed” with a very low
The behaviour of this exsolved gas is sepa- permeability of 10−16 m2. The fracturing is ini-
rated from the problem of magma ascent, and tiated at 1500 m, in accordance with measure-
considered independent of any other parameters, ments by Neuberg et al. (2006), and propagate
through additional numerical modelling in vertically towards the surface, as an increase in
Comsol Multiphysics (Collinson and Neuberg, permeability from 10−16 to 10−6 m2.
2012). In order to consider all possible out- Figure 6a shows the initial system, before
gassing pathways, including vertically through fracturing, where the gas loss is predominantly
the conduit, or laterally through the walls, we vertical, through the conduit. In Fig. 6b, the
model the gas response to brittle failure using a fracture zone has propagated vertically up to
simplistic “block-style” model, with a central 700 m depth. Consequently, the pressure has
conduit and adjacent wall-rocks. Brittle failure is increased through the conduit and corresponding
explicitly modelled as an increase in the perme- wall margins where the fractures have developed.
ability within narrow regions either side of the This is due to the regions of increased perme-
conduit. The problem is simplified to considering ability being confined by areas of lower perme-
permeable flow through a static media. Conse- ability above. Thereby, providing a suitable
quently, the equations for Darcy’s law (18), and environment for gas storage, which due to the
the continuity equation (19) are amalgamated to increased pressurisation, may force exsolved
derive a partial differential equation (20), which volatiles back into solution within the melt. This
is solved for pressure (P): change in pressurisation has resulted in a
Crystals, Bubbles and Melt: Critical Conduit Processes … 167

Fig. 6 a The initial system in equilibrium, before resumed equilibrium. Velocity range of arrows (ms−1):
fracturing has commenced. b The systems of fractures a max: 0.30, min: 1.3  10−3; b max: 0.38, min:
has propagated upwards to a depth of 700 m. c The 8.8  10−5; c max: 0.33, min: 6.8  10−4; d max:
fracture zones have reached the surface. d The system has 0.30, min: 9.5  10−4

corresponding change to the gas velocity pattern, associated geophysical signals (e.g. LF seismic-
which the lower conduit having a very low gas ity) within the conduit through moderating the
velocity due to the low pressure gradient. How- amount of gas stored in the system.
ever, at shallow levels, the higher pressure gra-
dient forces a much higher gas velocity towards
the surface. On reaching the surface (Fig. 6c), the 5 Summary and Implications
high permeability fractures result in a decrease in
pressure throughout the conduit and a rapid In the introduction to this chapter, we asked the
expulsion of the stored gas. Equilibrium condi- question of whether it is possible that small
tions resume approximately 4 h after the fracture changes in the composition of properties of the
zones reached the surface (Fig. 6d). This model magma could cause changes significant enough to
shows a propagating fracture zone is an effective be recorded in monitoring data or even visual
mechanism for degassing the conduit and wall observations. It is clear that these small changes,
margins. A key observation is the possibility for particularly when considering changes in the
this mechanism to produce periods of cyclic water content or conduit diameter, can have large
activity which are observed at many silicic vol- effects of the ascent velocity of the magma. These
canoes (e.g. Holland et al. (2011)), which can be effects are large enough that conceivable they
directly related to observed degassing patterns or could be simply observed as an increased extru-
through controlling the ascent rate and the sion rate at the surface. At this point however it
168 M. E. Thomas et al.

may be too late. If increased extrusions rates are Acknowledgements The authors are grateful for inspir-
being observed at the surface a critical threshold ing discussions with Geoff Kilgour (GNS New Zealand)
during J. Neuberg’s study leave at GNS Research Centre
of ascent rate may have already been surpassed. in Wairakei, and contributions to research ideas form
More useful would be to observe these former Ph.D. students within the Volcanic Studies Group
potential changes through monitoring before the at the University of Leeds.
magma was physically observed to be extruding
faster at the surface. Throughout this chapter we
have discussed the importance of shear stress, References
seismicity and outgassing, and it is to monitoring
data relating to these processes that we would Barclay J, Rutherford M, Carroll M, Murphy M, Devine J
(1998) Experimental phase equilibria constraints on
look for an indication of potential changes.
pre-eruptive storage conditions of the Soufrière Hills
Increases in ascent velocity may lead to a magma. Geophys Res Lett 25:3437–3440
change in eruption style, but they also potentially Burgisser A, Poussineau S, Arbaret L, Druitt TH, Gia-
cause changes in monitoring data. As discussed chetti T, Bourdier J-L (2010) Pre-explosive conduit
conditions of the 1997 Vulcanian explosions at
in the chapter, any increase in ascent velocity of
Soufrière Hills Volcano, Montserrat: I. Pressure and
the magma will cause an increase in the shear vesicularity distributions. J Volcanol Geoth Res
stress experience by the magma, and will likely 194:27–41
lead to an increase in rate of seismicity, or alter Castro JM, Gardner JE (2008) Did magma ascent rate
control the explosive-effusive transition at the Inyo
its location. It is now generally accepted that
volcanic chain, California? Geology 36:279–282
shear stress within the conduit also causes a Collier L, Neuberg J (2006) Incorporating seismic
significant deformation signal (e.g. Neuberg et al. observations into 2D conduit flow modelling. J Vol-
2018). Any changes in the number or location of canol Geoth Res 152:331–346
Collinson ASD, Neuberg JW (2012) Gas storage, trans-
low frequency earthquakes, or changes in the
port and pressure changes in an evolving permeable
near-field deformation around the conduit could volcanic edifice. J Volcanol Geoth Res 243–244:1–13
therefore be inferred as changes in the magma Cluzel N, Laporte D, Provost A, Kannewischer I (2008)
properties. In addition, the single most important Kinetics of heterogeneous bubble nucleation in rhy-
olitic melts: implications for the number density of
parameter identified within this study was chan-
bubbles in volcanic conduits and for pumice textures.
ges in the water content. The fractures generated Contrib Miner Petrol 156:745–763
by the seismicity, as a result of the increases in Devine JD, Murphy MD, Rutherford MJ, Barclay J,
ascent velocity, have also been shown to be Sparks RSJ, Carroll MR, Young SR, Gardner JE
(1998) Petrologic evidence for pre-eruptive pressure-
important outgassing pathways. An increase in temperature conditions, and recent reheating, of
water content would lead to an increase in andesitic magma erupting at the Soufrière Hills
volatiles and these would be more easily out- Volcano, Montserrat, W.I. Geophys Res Lett
gassed along the created fractures. Conceptually, 25:3669–3672
Devine JD, Rutherford MJ, Norton GE, Young SR (2003)
a clear signal of an increase in ascent velocity Magma storage region processes inferred from geo-
(and a potential early warning of a change in chemistry of Fe–Ti oxides in andesitic magma,
eruption style) caused by increase in water con- Soufrière Hills Volcano, Montserrat, WI. J Petrol
tent of the magma, could be an increase and 44:1375–1400
Edmonds M, Aiuppa A, Humphreys M, Moretti R,
deepening of low frequency earthquakes, Giudice G, Martin RS, Herd RA, Christopher T (2010)
accompanied by an increase in the deformation Excess volatiles supplied by mingling of mafic magma
signal followed by significant outgassing event. at an andesite arc volcano. Geochem Geophys Geosyst
In this chapter we look at a single volcanic 11:Q04005
Giordano D, Russell JK, Dingwell DB (2008) Viscosity
setting, but by changing the model parameter of magmatic liquids: a model. Earth Planet Sci Lett
these effects could be assessed at any volcano, 271:123–134
and regardless of the volcano studied the relative Gonnermann HM, Manga M (2007) The fluid mechan-
importance of the parameters considered should ics inside a volcano. Annu Rev Fluid Mech
39:321–356
remain unaltered.
Crystals, Bubbles and Melt: Critical Conduit Processes … 169

Gotto A (1999) A new model for volcanic earthquake at low-frequency earthquakes on Montserrat. J Volcanol
Unzen Volcano: melt rupture model. Geophys Res Geoth Res 153:37–50
Lett 26:2541–2544 Neuberg JW, Collinson ASD, Mothes PA, Ruiz CM,
Holland ASP, Watson IM, Phillips JC, Caricchi L, Aguaiza S (2018) Understanding cyclic seismicity and
Dalton MP (2011) Degassing processes during lava ground deformation patterns at volcanoes: Intriguing
dome growth: Insights from Santiaguito lava dome, lessons from Tungurahua volcano. Ecuador, Earth
Guatemala. J Volcanol Geoth Res 202:153–166 Planet Sci Lett 482:193–200
Hurwitz S, Navon O (1994) Bubble nucleation in rhyolitic Rutherford MJ, Devine JD (2003) Magmatic conditions
melts: experiments at high pressure, temperature, and and magma ascent as indicated by hornblende phase
water content. Earth Planet Sci Lett 122:267–280 equilibria and reactions in the 1995–2002 Soufrière
Ishibashi H (2009) Non-Newtonian behavior of hills magma. J Petrol 44:1433–1454
plagioclase-bearing basaltic magma: subliquidus, vis- Sparks RSJ, Murphy MD, Lejeune AM, Watts RB,
cosity measurement of the 1707 basalt of Fuji volcano, Barclay J, Young SR (2000) Control on the emplace-
Japan. J Volcanol Geoth Res 181:78–88 ment of the andesite lava dome of the Soufrière Hills
Lensky NG, Lyakhovsky V, Navon O (2002) Expansion volcano, Montserrat by degassing-induced crystalliza-
dynamics of volatile-supersaturated liquids and bulk tion. Terra Nova 12:14–20
viscosity of bubbly magmas. J Fluid Mech 460:39–56 Thomas ME, Neuberg JW (2012) What makes a volcano
Liu Y, Zhang Y, Behrens H (2005) Solubility of H2O in tick—A first explanation of deep multiple seismic
rhyolitic melts at low pressures and a new empirical sources in ascending magma. Geology 40:351–354
model for mixed H2O-CO2 solubility in rhyolitic Thomas ME, Neuberg JW (2014) Understanding which
melts. J Volcanol Geoth Res 143:19–235 parameters control shallow ascent of silicic effu-
Llewellin E, Manga M (2005) Bubble suspension rheol- sive magma. Geochem Geophys Geosyst 15:4481–4506
ogy and implications for conduit flow. J Volcanol Tuffen H, Dingwell DB, Pinkerton H (2003) Repeated
Geoth Res 143:205–217 fracture and healing of silicic magma generate
Lyakhovsky V, Hurwitz S, Navon O (1996) Bubble flow banding and earthquakes? Geology 31:1089–1092
growth in rhyolitic melts: experimental and numerical Wadge G, Robertson REA, Voight B (eds) (2014) The
investigation. Bull Volc 58:19–32 eruption of Soufrière Hills Volcano, Montserrat from
Marsh B (1981) On the crystallinity, probability of occur- 2000 to 2010. Geol Soc, London, Memoirs, p 39
rence, and rheology of lava and magma: contributions to Zobin VM, Ramirez JJ, Santiago H, Alatorre E,
mineralogy and petrology. Bull Volcanol 78:85–98 Navarro C (2011) Relationship between tilt changes
Marti I, Höfler O, Fischer P, Windhab EJ (2005) Rheology and effusive-explosive episodes at an andesitic vol-
of concentrated suspensions containing mixtures of cano: the 2004–2005 eruption at Volcan de Colima,
spheres and fibers. Rheol Acta 44:502–512 Mexico. Bull Volc 73:91–99
Neuberg JW, Tuffen H, Collier L, Green D, Powell T,
Dingwell D (2006) The trigger mechanism of

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter are
the Creative Commons Attribution 4.0 International included in the chapter's Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter's Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 When Does Magma Break?

Fabian B. Wadsworth, Taylor Witcher, Jérémie Vasseur,

Donald B. Dingwell and Bettina Scheu

Abstract
Geophysical signals arriving at the Earth’s surface originate from a source
mechanism at depth but are not necessarily directly observable. Therefore,
well-posed experiments can provide insights into source mechanics and,
importantly, the parameters required to model aspects of the sources of
unrest signals. In this Chapter we detail one such example of how
experimental laboratory work has improved our understanding of unrest
signals. We focus on the failure of single- and multi-phase magmas,
demonstrating that the liquid viscosity, and therefore the temperature and
volatile content of a magma of a given composition, is the limiting
parameter in determining whether a magma will ascend viscously or
whether it can fracture during ascent. This critical threshold is character-
ized by a Deborah number, the ratio of the timescale of relaxation to the
timescale of local flow. We show that for single-phase magmatic liquids
and for vigorously vesiculating magmas, a local Deborah number of 10
2
is the limit above which mixed viscoelastic behaviour including fracture
propagation can be expected, and a Deborah number of 1 is the limit
above which magma is dominantly elastic and responds in a brittle manner
to applied stresses. These thresholds can be understood in terms of the
onset and peak of the Debye relaxation process for viscoelastic liquids.
The apparent validity of a Maxwell model permits us to predict the
maximum stress that can be supported by a volcanic liquid deforming in
the high Deborah number range. We use these constraints to provide a
map of timescales on which we contour dominant system responses from
viscous to purely brittle; valid for all magmatic liquids. Finally, we
explore the scaling necessary to extend these conceptual insights to
crystal- and bubble-bearing magmas valid under specific conditions.

F.B. Wadsworth (&)  T. Witcher  J. Vasseur

D.B. Dingwell  B. Scheu
Ludwig-Maximilians-Universität, Theresienstr.41,
80333 Munich, Germany
e-mail: [email protected]

Advs in Volcanology (2019) 171–184

DOI 10.1007/11157_2017_23
© The Author(s) 2017
Published Online: 18 October 2017
172 F.B. Wadsworth et al.

The competing timescales of deformation and relaxation in magma are

relevant to unrest source mechanisms that originate from magma
deformation, such as long-period seismic signals that are used to predict
eruption timing.

Keywords
  
Rheology Strain rate Experimental volcanology Glass transition 
 
Failure forecasting Low frequency earthquakes Viscous dissipation

Glossary
Rheology The study of the response of a material to an applied
stress or deformation. In the volcano-sciences this
typically refers to magma-rheology which is dominated
by the rheology of the liquid phase (a viscous or
viscoelastic melt) and the additional effect of suspended
phases (crystals and gases).
Viscoelasticity A material response to an applied stress in which both
elastic and viscous components of the deformation are
observed. In magma, as the temperature decreases or
the local strain rate increases, the elastic component
becomes more dominant. When the viscous component
is negligible, purely elastic behaviour and
material-rupture can readily occur. See Rheology.

predict an eruption or dome collapse event with

1 Introduction
reasonable accuracy (Salvage and Neuberg
2016). The utility of the low-frequency event
There are a wide variety of observable unrest
type has been solidified since physical hypothe-
signals at active volcanoes. The key unrest sig-
ses for the source mechanism have been put
nals are those that are diagnostic of a new regime
forward (Neuberg et al. 2006) and tested in
of behaviour or of how the system may evolve in
scaled laboratory expeirments (Benson et al.
the future. One such family of events are the
2008; Tuffen et al. 2008). The phenomenological
low-frequency earthquakes at volcanoes, which
observation of the use of certain signal types over
are thought to directly represent magma move-
others is useful and has led to accurate retro-
ment at moderate to shallow depths (Chouet et al.
spective eruption forecasts (Voight 1988; Chouet
1994; Chouet 1996; Neuberg et al. 2006; Salvage
1996). However, knowledge of the physics of the
and Neuberg 2016). Such events are powerful
source leads to a more diverse range of tools
tools particularly because they are closely asso-
being deployed to understand the evolution of
ciated with surface activity (Miller et al. 1998).
volcanic systems into the future. In the context of
Accelerating event rates of low frequency
low-frequency earthquakes at volcanoes, once a
earthquakes can be used to retrospectively
When Does Magma Break? 173

source mechanism is identified (Tuffen et al. predict that if a shear strain is imposed on a
2003; Neuberg et al. 2006), numerical models of viscoelastic liquid, the resultant shear stress will
magma ascent in the conduit beneath volcanoes rise rapidly to a peak value si and will relax over
can be used to reproduce the depths of the source time t according to sðtÞ ¼ si expð
t=kr Þ. This
from first principles (Thomas and Neuberg 2012) simple concept has proved powerful in describ-
and could eventually make forecasts of other ing the transition between Newtonian and vis-
observables at the surface that would be consis- coelastic behaviour in volcanic liquids (Dingwell
tent with impending eruption. 1995, 1996). Indeed, this model is invoked to
In this Chapter we explore the source mech- explain the transition between volcanic liquids
anism of low-frequency earthquakes at volcanoes and volcanic glasses on cooling (e.g. Stevenson
from a physical perspective using a compilation et al. 1995), the fragmentation of fluid and bub-
of data from scaled laboratory experiments. We bly magma undergoing decompression (Alidi-
use these datasets to demonstrate that the critical birov and Dingwell 1996; Kameda and Kuribara
threshold for fracture propagation in a vis- 2008) and the conditions under which extensive
coelastic fluid such as magma is universal. We shear fracture networks can form at or near vol-
propose that this potentially simplifies the inputs canic conduit margins (Gonnermann and Manga
required for effective modelling of source 2003; Tuffen et al. 2003; Kendrick et al. 2014;
mechanics of low-frequency events. More than Hornby et al. 2015). In all cases, it is useful to
this, we hope that this provides a good example define a Deborah number De, which is the
of how laboratory work can provide valuable dimensionless ratio between the Maxwell relax-
insight into the physical feasibility of models ation time and the timescale of deformation k.
proposed to explain unrest signals at volcanoes. For simple (viscometric) shearing flow, the latter
timescale is 1=_c and thus

2 Scaling the Viscous-to-Brittle kr l_c

De ¼ ¼ ð1Þ
Transition in Magmas k G1

2.1 Single-Phase Magmatic Liquids Our hypothesis is that a Deborah number of

unity separates the regime between viscous
Newtonian viscous liquids deform and flow ðDe  1) and elastic (De  1Þ responses to
under applied stresses such that the rate of shear shear stresses, or equivalently, between a system
strain c_ resulting from an applied shear stress s is that can relax shear stresses efficiently from one
proportional to the viscosity l via s ¼ l_c. in which shear stresses can accumulate elasti-
However, as large shear stresses are applied to cally. In two different experimental types, both
viscous liquids, they can exhibit viscoelastic Webb and Dingwell (1990a) and Cordonnier
behaviour, which manifests as an apparent et al. (2012a, b) showed that in fact the first
non-Newtonian relationship between s and c_ and evidence of behaviour that is not purely viscous
can result in fracture propagation locally in the occurs at De  10
2 and not exactly at unity.
liquid. Of interest here is the transition from This observation will be discussed later.
simple, apparently Newtonian viscous behaviour The power of Eq. 1 is that simple parameters
to complex viscoelastic behaviour. that can be measured in the laboratory can be
Maxwell described the simplest viscoelastic used to predict where this transitional point of
model in which a liquid has a characteristic time De ¼ 1 will be. This critical Deborah number
0
required to relax an applied shear stress, termed can be termed De : In what follows, we give
the Maxwell relaxation timescale kr , which is examples of how l and G1 have been found for
given by kr ¼ l=G1 where G1 is the shear a range of volcanic liquid compositions at mag-
modulus in the purely elastic regime. The Max- matic temperatures and volatile contents before
well model of viscoelasticity permits us to exploring the validity and applicability of the
174 F.B. Wadsworth et al.

scaling in Eq. 1. The dimensionless nature of Unlike viscosity, the shear modulus in the
Eq. 1 means that it can be assessed for any sys- elastic regime does not vary significantly in
tem so long as k can be defined. siliciate liquids and is not strongly dependent on
At the core of the experimental toolkit is the composition or temperature. Indeed, a short sur-
determination of the fundamental quantities that vey of the values for G1 in silicate glasses and
are required to model volcanic processes liquids at a range of temperature and a huge
involved in unrest. The most variable physical range of composition, provided by Dingwell and
quantity in volcanic systems is the viscosity of Webb (1989) shows that G1 ¼ 10100:5 Pa. Here
the liquid phase. Using a range of techniques, we use this range in order to fully parameterize
this viscosity can be determined with prodigious De (by predicting kr Þ as a function of tempera-
accuracy and significant effort has been expen- ture (via Eq. 2) for any silicate liquid in the
ded in mapping the full range of composition, shallow crust.
temperature (see Giordano et al. 2008) and
volatile content (Hess and Dingwell 1996) rele-
vant to shallow magmatic settings. Multi- 2.2 Extensions to Multiphase
component models have been proposed such Magmas
that the viscosity of any composition of silicate
volcanic liquid on Earth can now be predicted as Except in rare circumstances such as
a function of temperature including the range of obsidian-forming eruptions, volcanic liquids are
conditions between shallow magma storage and not often erupted without some proportion of
the surface (Hess and Dingwell 1996; Giordano suspended pore space (either as isolated bubbles
et al. 2008). This provides parameterization of or connected networks) and rigid crystals. In
lðTÞ for use in Eq. 1. The most commonly used either case, the utility of Eq. 1 requires additional
form for lðTÞ is the non-Arrhenian attention. We posit that for the liquid phase
Vogel-Fulcher-Tammann equation of the form between the pores or the crystals, De given by

 Eq. 1 holds. However, we acknowledge that
B estimation or measurement of the rate of shear
l ¼ A exp ; ð2Þ
T
C strain locally between pores or crystals would be
difficult (Deubelbeiss et al. 2011). Therefore, a
where A; B; and C are coefficients that are more robust criterion for the viscous-to-brittle
experimentally determined and then parameter- transition in crystal- or bubble-bearing magma
ized as a function of families of oxides in the would require that we scale the bulk strain rate c_ b
liquid structure (Giordano et al. 2008) or as a on the system for the effect of the suspended
function of the dissolved water content (for load.
calc-alkaline rhyolites: Hess and Dingwell 1996), For crystals, we show some first-order scaling
to give two examples. attempts for their effect on the critical threshold
To give examples of the range of viscosities for the onset of brittle behaviour in multiphase
of interest in volcanic scenarios, we give end magma. The simplest view of the local flow of
members in Fig. 1 for a calc-alkaline rhyolite liquid in crystal-bearing magma under constant
using the model from Hess and Dingwell (1996), shear stress is that the rate of shear strain between
and for the basaltic liquid composition provided the crystals should scale approximately with c_ ¼
in Zhang et al. (1991) using the model from
c_ b ð1
/x =/m Þ
1 where /x is the suspended
Giordano et al. (2008), both contoured for a
crystal volume fraction and /m is a jamming
range of water contents.
fraction above which no more crystals can be
When Does Magma Break? 175

10 4/ T [K− 1]
8 9 10 11 12
10 23 10 13
Calc-alkaline rhyolite
(Model from Hess & Dingwell, 1996)
10 20 10 10

Relaxation timescale, λ r [s]

Liquid viscosity, μ [Pa.s] Basalt (Composition from Zhang et al., 1991;
Model from Giordano et al., 2008)
10 17 10 7

10 14 H 2O
10 4
O
wt% H2
0.01 % H 2O
H 2
O
t %
wt w
10 11 0.1 %
wt 0.1 % H 2O
10 1
0 1 wt
10 8 H 2O 10 − 2
t%
1w

10 5 10 − 5

10 2 10 − 8
1300 1200 1100 1000 900 800
Temperature, T [K]

Fig. 1 The viscosity of end-member magmatic liquids 1991) calculated using the multicomponent viscosity
with variable dissolved water concentrations. Plotted are model (Giordano et al. 2008) for 0–1 wt% water. The
the results for calc-alkaline rhyolite from the general relaxation timescale is calculated assuming a composition
viscosity model for hydrous silicic liquids (Hess and independent value for G1 of 1010 Pa (Dingwell and
Dingwell 1996) for 0.01–1 wt% water, and a typical Webb 1989)
basaltic composition (composition from Zhang et al.

added to the flowing system. This scaling would system, which is at odds with the scaling of strain
imply that the Deborah number for a rate for local liquid effects only. Nonetheless, this
crystal-bearing magma Dex would be approach described their data satisfactorily and
the conceptual insight that crystals locally


l /x
1 increase the liquid rate of shear strain relative to
Dex ¼ c_ 1
; ð3Þ
G1 b /m the bulk value is robust, with the implication is
that the whole suspension will begin fracturing at
where /m is a function of crystal shape and lower bulk rates of strain relative to a
roughness (Mueller et al. 2010; Mader et al. single-phase liquid of the same composition as
2013). A critical value of Dex , termed De0x ; is crystals are added. It may be that additional
10
2 ; consistent with the limiting De0 at / ¼ 0. second-order effects are important at high /x ,
A new definition of the bulk failure criterion De0 which are not accounted for in this simple
can now be made, which decreases as / ! /m analysis.
0
so that De0 ¼ Dex ð1
/x =/m Þ. In Fig. 2 we For the case of porous magmas, there are two
demonstrate how this concept predicts a linear considerations: (1) The growth of bubbles exerts
relationship between Dex and /x =/m , which in a rate of shear strain in the liquid concentrated at
turn shows that lower bulk strain rates are the bubble walls, which is broadly independent
required to induce brittle behaviour in of any bulk shearing deformation, and (2) like for
crystal-bearing magmas. the crystal case, the presence of bubbles changes
This hypothesis was found to hold for simple the partitioning of the bulk rate of shear strain in
two-phase systems by Cordonnier et al. (2012a), the liquid between the bubbles. In the case of
but where those authors required that l be scenario (1), the Rayleigh-Plesset equation can
replaced by the suspension viscosity of the whole be used to relate the bubble growth rate to the gas
176 F.B. Wadsworth et al.

Fig. 2 The critical Deborah

number required for 0.010 Dex
fracturing De0 is reduced

Critical Deborah number, De

when crystals are suspended
in magma. To a first-order, the 0.008
De0 value is reduced
proportional to
De0 ¼ De0x ð1
/x =/m Þ
0.006

0.004

0.002

0.000
0.0 0.2 0.4 0.6 0.8 1.0
Normalized crystal volume fraction, φ x /φ m

pressure in the bubble, relative to the hydrostatic 2013) but remains untested in the low Deborah
pressure (Sparks 1978). This can be augmented number regime where bubble deformation may
for the diffusion-controlled gradient of viscosity be important.
in the immediate liquid shell around a growing
bubble (Prousevitch et al. 1993; Lensky et al.
2001). The component of the rate of shear strain 2.3 Apparent Non-newtonian Effects
tangential to the bubble wall c_ h can then be
computed throughout bubble growth (Ichihara In single-phase liquids at Deborah numbers
et al. 2002). If we were then to input c_ h into our below the critical threshold at which fracturing is
computation of De, we would predict under observed, there is evidence that during steady
which conditions the system would meet the state shearing flow, there is a non-Newtonian
criterion of De0 ¼ 10
2 locally at the bubble rim. relationship between applied shear stress and
These conditions would be best cast in terms of resultant shear rate of strain (Simmons et al.
the critical rate of pressure or temperature 1982; Dingwell and Webb 1989; Webb and
change, or bulk initial volatile content that would Dingwell 1990a; Cordonnier et al. 2012b). In the
allow the bubble to grow sufficiently fast that onset of this non-Newtonian, the onset of this
fractures could be propagated at the bubble wall. non-Newtonian behaviour appears to be well
In case (2), the effect of bubbles on the bulk described by the value above which viscous
Deborah number is less clear. We speculate that dissipation of heat is active on the system length
at De  1, the stress required to fracture a scale of interest (Costa and Macedonio 2005;
bubbly system will be analogous to that required Hess et al. 2008). This can be scaled by the
to fracture a vesicular glass. In this case, models Brinkman number Br
for the effect of spherical cavities on the stress
required for fracturing are valid and they predict Ug l_c2
Br ¼ ¼ ; ð4Þ
that the stress is reduced significantly as the bulk Ul kq
gas volume fraction increases (Sammis and
Ashby 1986). This has been confirmed in the where Ug and Ul are the gain and loss power
high Deborah number regime for porous liquids densities, respectively, q is the areal heat flux out
analogous to volcanic systems (Vasseur et al. of the system and k is the thermal conductivity of
When Does Magma Break? 177

the material. Ug ¼ l_c2 represents the amount of smooth sample inflation was punctuated by vis-
energy produced in a system of a given volume ible fractures opening. In these decompression
by viscous dissipation of heat, while Ul ¼ kq experiments there additionally was a violent
represents the energy lost due to diffusive ther- rupture mode in which the fracturing was per-
mal equilibration. When Br  1, heat is effi- vasive and shattered the sample in a vigorous
ciently produced and inefficiently lost from the fragmentation event. These data are selected here
system, resulting in a bulk temperature increase because the liquids used have a known relaxation
in the liquid. This would manifest itself as an time under the conditions used in the experi-
apparent shear thinning rheology if the temper- ments and care was taken by the authors who
ature increase where not locally accounted for, originated the work to investigate the shear rates
and would be most likely to be operative at high of strain local to the bubbles (discussed above).
viscosities and high shear strain rates. We note In Fig. 3 we map these experimental results as
that unlike the Deborah number, the Brinkman a ratio of the deformation timescale to the relax-
number is scale dependent (as q depends on the ation timescale, which permits us to contour the
area available for heat transfer out of the system) plot for critical Deborah numbers. We
and so should be assessed for each system scale colour-code the data according to the bulk mode
separately. of response of the sample to the deformation
using green to represent purely viscous relaxed
behaviour, orange to represent brittle unrelaxed
3 The Universal Breaking behaviour and red to represent the complete
Timescales of Volcanic Liquids violent rupture of the sample. These data suggest
that the viscous relaxed behaviour transitions to
Laboratory data related to the viscous-to-brittle unrelaxed brittle behaviour at a Deborah number
transition in magmas has been collected in a of 10
2 , and are consistent across a huge range of
variety of geometries. Using single-phase liquids, experimental conditions and across both the
there are two dominant geometries: (1) thin fibers single-phase compression and vesicular decom-
of silicate liquid of basaltic, andesitic, phonolitic pression experiment types. Furthermore, we see
and rhyolitic composition were stretched under that a Deborah number of unity consistently
constant load in tension until the fibers snapped separates the experiments for which the bulk
in a singular fracture event (Webb and Dingwell response was unrelaxed and brittle from those for
1990b), and (2) cylinders of synthetic borosili- which the response was violent rupture, frag-
cate liquid were compressed under constant load mentation or complete failure. It appears that
and the bulk temporal evolution of the rate of these two thresholds, De ¼ 10
2 and De ¼ 1 are
axial shortening was determined as viscous (re- universal even when comparing analogue room
laxed) if it continuously increased to a steady or temperature liquids with high temperature silicate
near-steady value, or brittle (unrelaxed) if the liquids deformed in a variety of ways. This lends
axial shortening rate jumped due to fracturing power to the scaling provided by the dimension-
events (Cordonnier et al. 2012b). In another type less Deborah number and implies that we need
of experiment, analogue vesicular liquids were only to define the liquid viscosity and a charac-
decompressed at different rates from pressure teristic rate of deformation in order to predict
(Kameda and Kuribara 2008; Kameda et al. whether a system will flow viscously or rupture
2013). In this type of decompression experiment, violently. For example, the working viscosity of
if the dominant response was viscous (relaxed), the analogue fluid used in Kameda and Kuribara
then the sample was seen to grow due to (2008) is 100  l  1010 Pa.s, which extends to
decompression-driven bubble growth, and if the much lower used in Cordonnier et al. (2012b),
dominant response was brittle (unrelaxed), then and yet the scaling with the Deborah number
178 F.B. Wadsworth et al.

10 6 Webb & Dingwell, 1990

Cordonnier et al., 2012
Kameda et al., 2008

Deformation timescale, λ [s]

Kameda et al., 2013
10 4
xed −2
Rela 0
=1 ed
ax
10 2 De rel
Un
1
Br =
ing
h eat ian) =1 nt
10 0 us ton De ole
sco ew Vi pture
Vi on-N ru
(n

10 − 2

10 − 4
10 − 5 10 − 4 10 − 3 10 − 2 10 − 1 10 0 10 1 10 2 10 3
Relaxation timescale, λr [s]
Fig. 3 A map of behaviour arising from deformation and 1, which separate experiments with a purely viscous
experiments on natural (Webb and Dingwell 1990b) and response from those with an unrelaxed brittle response or
synthetic (Cordonnier et al. 2012b) silicate liquids and on a violent rupture response, respectively. Additionally
analogues of magmatic liquids undergoing vesiculation marked is the Brinkman number Br of 1, which separates
(Kameda and Kuribara 2008; Kameda et al. 2013). The the boundary between isothermal viscous behaviour
experimental results were obtained by applying a charac- (above the line) and viscous behaviour in which the
teristic timescale of deformation k of a liquid with a material heats up due to viscous dissipation as heat (below
characteristic timescale of stress relaxation kr . Marked are the line)
ratios kr =k; equivalent to Deborah numbers De, of 10
2

holds across a broad range simply because the at much higher deformation timescales for the
deformation timescale was also smaller in the same relaxation timescale (the curve would shift
former example. up in Fig. 3). This implies that in the natural
In the viscous field, Cordonnier et al. (2012b) case, viscous dissipation of heat may be far more
additionally recorded whether samples hosted a important than shown here for the sample length
measurable temperature increase due to viscous scale (Costa and Macedonio 2005; Mastin 2005;
dissipation of heat during the experiment, or not. Costa et al. 2007). Nevertheless, the Deborah
In Fig. 3, we plot those with a measureable number limits discussed appear to be universal
temperature increase as filled symbols and those and, importantly, are scale-independent.
without this feature as unfilled symbols. To To gain helpful physical insight into why
explain this, we plot the threshold dimensionless 10
2  De  100 is the transitional window
Brinkman number of unity Br ¼ 1, using an between a purely viscous and a purely elastic
estimation of the loss power density for the fur- response of a liquid to a deformation, we provide
nace and sample size used Ul ¼ 104:5 W.m
2 low-strain, high frequency oscillatory rheological
(Cordonnier et al. 2012b). We note that the measurement data for similar liquids (Fig. 4).
Brinkman number curve consistently divides the Here, rods of single-phase liquid are subjected to
regimes of purely isothermal experiments and a low-amplitude oscillatory strain with a forcing
those with measurable heat gain. On this map, frequency x at a range of temperatures similar to
the position of Br ¼ 1 is non-unique and depends those used in experiments presented in Fig. 2.
on sample size, such that on the scale of a vol- Here, x is normalized with kr , yielding a
canic conduit, for example, Br ¼ 1 would occur dimensionless frequency or, equivalently, a
When Does Magma Break? 179

Deborah number (frequency domain), De

−3
10 10 − 1 10 1 10 3 10 5 10 7
0.6
Na 2 Si2 O 5 ; Webb, 1991
Normalized imaginary component,
Rhyolite; Webb, 1992
0.5 Rhyolite; Bagdassarov & Dingwell, 1993
HPG8; Bagdassarov et al., 1993

0.4

0.3

0.2

0.1

0.0

10 − 3 10 − 1 10 1 10 3 10 5 10 7
Normalized frequency, ωλ r
00
Fig. 4 The response of synthetic (Webb 1991; Bagdas- shear modulus G =G1 as a function of the frequency of
sarov and Dingwell 1993a) and natural (Webb 1992; applied oscillation normalized by the relaxation timescale
Bagdassarov and Dingwell 1993b) silicate liquids to low xkr , which is equivalent to a Deborah number De in the
strain, high frequency oscillatory deformation tests. This frequency domain. We show the viscoelastic prediction
is shown as a normalized imaginary component of the for a Maxwell liquid for comparison (solid curve)

00
frequency-domain version of the Deborah num- overprediction of G around xkr 1 for the
ber. The metric that we choose to track is the rhyolitic liquids, which is poorly understood.
imaginary component of the complex elastic Nevertheless, we point out that the first onset of a
00
modulus G normalized by G1 . When this value measurable elastic component to the response
00
G =G1 is close to zero in the low Deborah of the liquid to deformation occurs at
number limit, the system is dominated by liquid De ¼ xkr
10
2 . Similarly, the point above
behaviour and when it rises from zero at which the majority of the response to deforma-
00
increasing Deborah number, there is an increas- tion is elastic (the peak of G ) occurs at De ¼
ing component of the response to the forced xkr
1 (Fig. 3). We use this observation to
oscillation that is elastic. We present collated validate the two thresholds found in Fig. 3. This
data for a sodium disilicate synthetic composition implies that fractures can propagate in silicate
(Webb 1991), rhyolitic compositions (Webb liquids when there is even a small component of
1992; Bagdassarov and Dingwell 1993b) and a elastic behaviour (De  10
2 ) and that those
synthetic composition used as an analogue for fractures can propagate vigorously when the
type calc-alkaline rhyolite systems (Bagdassarov elastic behaviour dominates over the viscous
and Dingwell 1993a). This imaginary component behaviour (De  1). An interpretation might also
can be described by the generalized Debye model be that 10
2  De  1 is the range in which
for viscoelasticity for systems with a single fracture propagation is competing with viscous
00
characteristic relaxation time as G ¼ xkr = relaxation of stress and therefore the fractures are
h i
ðxkr Þ2 þ 1 . The data reproduce the broad unlikely to be long, sharp-tipped or pervasive.
And that at De  1, stress dissipation by fracture
shape of the Debye model, albeit with an
180 F.B. Wadsworth et al.

propagation can be localized onto longer, sharp scale scaled with natural constraints of in-conduit
and pervasive fracture networks. volcanic fracture systems and natural frequencies
The fact that a viscous limit to the expansion of low-frequency volcano-seismicity, indicating
of vesicular magma also scales with our Deborah that these viscoelastic fracturing events are
number criterion is tantalizing. Microphysically, indeed the likely source mechanism for
in this case it is not fracture of a deforming low-frequency earthquakes at volcanoes. This
homogeneous liquid, but fractures propagating at confirmed the conclusion of Neuberg et al.
bubble walls as the local rate of strain in the (2006), who showed that these low-frequency
liquid induced by expansion of the bubble meets events were most likely to be associated with
the Deborah number criterion for fracture prop- repetitive fracturing of magma at a given depth in
agation (c.f. Kameda and Kuribara 2008). For the conduit during ascent. Neuberg et al. (2006)
complete rupture of the vesicular material— further modelled magma ascent in confined
which is a fragmentation event or violent rupture geometry and showed that a De 1 is met at a
in Fig. 3—the fractures propagating from the depth of 830 m below the surface using param-
bubbles must interact. Presumably in the range eters typical for recent eruptions at Soufriere
10
2  De\1, fractures propagate but the frac- Hills volcano and the magma thereof. Thomas
ture tips are blunted during competing viscous and Neuberg (2012) predicted a deeper source of
relaxation such that brittle behaviour can be 1500 m using the same model approach but by
observed but is not catastrophic to the system. invoking a conduit restriction (see Chap. 9),
Then, as the local strain rate at the bubble wall consistent with inversions for low-frequency
increases further and De approaches 1, the frac- sources using seismic data. Therefore, vis-
tures propagate in a dominantly elastic medium coelastic magma fracturing in the high Deborah
and can span the inter-bubble distances, inter- number regime appears to be a consistent model
acting to produce violent rupture. Therefore, it is for the source mechanics of low-frequency vol-
clear that a scaling of the Deborah number con- cano seismicity, often used for eruption
cept to bubble wall dynamics would provide a forecasting.
fragmentation criterion for viscoelastic vesicu- Vasseur et al. (2015) showed that the fore-
lating magma (c.f. Namiki and Manga 2005; castability of full sample rupture scales with the
Koyaguchi et al. 2008; Namiki and Manga heterogeneity of the system—cast most simply as
2008). a porosity (Fig. 5). The implication is that the
less vesicular the magma undergoing deforma-
tion, the less likely that accurate forecasts based
4 Laboratory-Scale Unrest Signals on accelerated precursory signals can be made,
with up to *120% error on the timing of the
When magma breaks in a laboratory experiment, rupture event for single-phase homogeneous
acoustic emissions—packets of acoustic energy liquids in the high De limit. It may be that the
—are released and can be recorded (Benson et al. magma vesicularity, pore-network structure
2008; Lavallee et al. 2008). These signals appear (Vasseur et al. 2017), crystallinity and textural
to represent large total released amounts of anisotropy, play key roles in determining fore-
energy when k is short (high c_ Þ, compared to casting success based on low-frequency earth-
when k is long (low c_ Þ (Lavallee et al. 2008). quakes. More experimental work is clearly
This supports our posit that fracture networks are required in this area, along with more rigorous
more likely to be pervasive and large when De is scaling between acoustic and seismic events that
large than when De is small. For single phase originate from magma failure.
silicate liquids deformed at high De, Tuffen et al. An implication of the models explored here,
(2008) showed that the experimental acoustic encapsulated by Eq. 1, is that the peak stress
event frequency range and sample fracture length supported by a liquid rm can be predicted as a
When Does Magma Break? 181

1.4
Sintered glass beads;

Error on magma failure forecast

Vasseur et al., 2015; 2017
1.2 Welded volcanic debris;
Vasseur et al., 2017

1.0

0.8

0.6

0.4

0.2

0.0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40
Gas volume fraction, φ
Fig. 5 In the high-De regime, the error on a prediction of The failure times are recorded as a large stress drop during
failure times scales with the porosity of the material such uniaxial loading, and the approach to failure is monitored
that low porosity magmas are unpredictable and high using acoustic emissions generated by pre-failure
porosity magmas are predictable (Vasseur et al. 2015). micro-fracturing events

Fig. 6 The peak stress that 10 11

can be supported by a
shearing liquid in the low De
limit is given by rm ¼ G1 De: 10 10 G∞
And in the high De limit, this
Maximum stress, σm [Pa]

value appears to 10 9
asymptotically approach
rm ¼ 10
2 G1 which is
equivalent to the strength of 10 8 σg
glassy materials in that same
regime (Simmons et al. 1982) 10 7

10 6

10 5

10 4
10 − 6 10 − 4 10 − 2 10 0 10 2 10 4 10 6
Deborah number, De

function of De by rm ¼ G1 De. Once De  10
2 , perhaps significant that the average strength of
however, this linear relationship appears to be glass in the high De regime is 108 Pa (Sim-
invalid as fractures can form in the liquid. It is mons et al. 1982; Vasseur et al. 2013), consistent
182 F.B. Wadsworth et al.

with a rupture threshold of 10
2 G1 . These Dingwell DB (1995) Relaxation in silicate melts; some
applications. Rev Mineral Geochemistry 32:21–66
relationships are explored in Fig. 6.
Dingwell DB (1996) Volcanic dilemma: flow or blow?
Here we summarize just one example of how Science (80-) 273:1054–1055. doi:10.1126/science.
targeted laboratory experiments and dimensional 273.5278.1054
analysis have demonstrated the most likely source Dingwell DB, Webb SL (1989) Structural relaxation in
silicate melts and non-Newtonian melt rheology in
mechanism of unrest signals commonly used to
geologic processes. Phys Chem Miner 16:508–516
monitor magma movement and predict impend- Giordano D, Russell JK, Dingwell DB (2008) Viscosity
ing eruptions. Clearly, with scaling arguments for of magmatic liquids: a model. Earth Planet Sci Lett
the applicability of an experimental set up—such 271:123–134
Gonnermann HM, Manga M (2003) Explosive volcanism
as we show here with the Deborah number anal-
may not be an inevitable consequence of magma
ysis—laboratory-based work can provide new fragmentation. Nature 426:432–435
insights into unrest mechanisms that facilitate Hess KU, Dingwell DB (1996) Viscosities of hydrous
more accurate forward-modelling of geophysical leucogranitic melts: a non-Arrhenian model. Am
Mineral 81:1297–1300
signals.
Hess K-U, Cordonnier B, Lavallee Y, Dingwell DB
(2008) Viscous heating in rhyolite: an in situ exper-
imental determination. Earth Planet Sci Lett 275:121–
126. doi:10.1016/j.epsl.2008.08.014
References Hornby AJ, Kendrick JE, Lamb OD, Hirose T, De
Angelis S, Aulock FW, Umakoshi K, Miwa T, Henton
Alidibirov M, Dingwell DB (1996) Magma fragmentation De Angelis S, Wadsworth FB, Hess KU (2015) Spine
by rapid decompression. Nature 380:146–148 growth and seismogenic faulting at Mt. Unzen, Japan.
Bagdassarov N, Dingwell DB, Webb SL (1993) Effect of J Geophys Res: Solid Earth 120(6):4034–4054
boron, phosphorus and fluorine on shear stress Ichihara M, Rittel D, Sturtevant B (2002) Fragmentation
relaxation in haplogranite melts. Eur J Miner 3:409– of a porous viscoelastic material: implications to
425 magma fragmentation. J Geophys Res: Solid Earth
Bagdassarov NS, Dingwell DB (1993) Frequency depen- 107(B10)
dent rheology of vesicular rhyolite. J Geophy Res: Kameda M, Kuribara H, Ichihara M (2008) Dominant
Solid Earth 98(B4):6477–6487 time scale for brittle fragmentation of vesicular magma
Benson P, Vinciguerra S, Meredith P, Young R (2008) by decompression. Geophy Res Lett 35(14)
Laboratory simulation of volcano seismicity. Science Kameda M, Ichihara M, Shimanuki S, Okabe W, Shida T
(80-) 322:249–252 (2013) Delayed brittle-like fragmentation of vesicular
Chouet BA (1996) Long-period volcano seismicity: its magma analogue by decompression. J Volcanol Geoth
source and use in eruption forecasting. Nature Res 258:113–125
380:309–316. doi:10.1038/380309a0 Kendrick JE, Lavallée, Y, Hirose T, Di Toro G, Hornby AJ,
Chouet BA, Page RA, Stephens CD et al (1994) De Angelis S, Dingwell DB (2014) Volcanic drumbeat
Precursory swarms of long-period events at Redoubt seismicity caused by stick-slip motion and magmatic
Volcano (1989–1990), Alaska: their origin and use as frictional melting. Nat Geosci 7(6):438
a forecasting tool. J Volcanol Geotherm Res 62:95– Koyaguchi T, Scheu B, Mitani NK, Melnik O (2008) A
135. doi:10.1016/0377-0273(94)90030-2 fragmentation criterion for highly viscous bubbly
Cordonnier B, Caricchi L, Pistone M et al (2012a) The magmas estimated from shock tube experiments.
viscous-brittle transition of crystal-bearing silicic melt: J Volcanol Geotherm Res 178:58–71
direct observation of magma rupture and healing. Lavallee Y, Meredith PG, Dingwell DB et al (2008)
Geology 40:611–614 Seismogenic lavas and explosive eruption forecasting.
Cordonnier B, Schmalholz SM, Hess K, Dingwell DB Nature 453:507–510. doi:10.1038/nature06980
(2012b) Viscous heating in silicate melts: an exper- Lensky NG, Lyakhovsky V, Navon O (2001) Radial
imental and numerical comparison. J Geophys Res variations of melt viscosity around growing bubbles
117:B02203 and gas overpressure in vesiculating magmas. Earth
Costa A, Macedonio G (2005) Viscous heating effects in Planet Sci Lett 186:1–6
fluids with temperature-dependent viscosity: trigger- Mader HM, Llewellin EW, Mueller SP (2013) The
ing of secondary flows. J Fluid Mech 540:21–38 rheology of two-phase magmas: a review and analysis.
Costa A, Melnik O, Vedeneeva E (2007) Thermal effects J Volcanol Geoth Res 257:135–158
during magma ascent in conduits. J Geophys Res 112: Mastin LG (2005) The controlling effect of viscous
B12205. doi:10.1029/2007JB004985 dissipation on magma flow in silicic conduits. J Vol-
Deubelbeiss Y, Kaus BJ, Connolly JA, Caricchi L (2011) canol Geoth Res 143(1):17–28
Potential causes for the non‐Newtonian rheology of Miller AD, Stewart RC, White R et al (1998) Seismicity
crystal‐bearing magmas. Geochem Geophy Geosys 12(5) associated wtih dome growth and collapse at the
When Does Magma Break? 183

Soufriere Hills Volcano, Montserrat. Geophys Res sources in ascending magma. Geology 40:351–354.
Lett 25:3401–3404 doi:10.1130/g32868.1
Mueller S, Llewellin EW, Mader HM (2010) The Tuffen H, Dingwell DB, Pinkerton H (2003) Repeated
rheology of suspensions of solid particles. Proc R fracture and healing of silicic magma generate flow
Soc A Math Phys Eng Sci 466:1201–1228 banding and earthquakes? Geology 31:1089–1092.
Namiki A, Manga M (2005) Response of a bubble bearing doi:10.1130/g19777.1
viscoelastic fluid to rapid decompression: implications Tuffen H, Smith R, Sammonds PR (2008) Evidence for
for explosive volcanic eruptions. Earth Planet Sci Lett seismogenic fracture of silicic magma. Nature
236(1):269–284 453:511–514
Namiki A, Manga M (2008) Transition between frag- Vasseur J, Wadsworth FB, Lavallée Y et al (2013)
mentation and permeable outgassing of low viscosity Volcanic sintering: timescales of viscous densification
magmas. J Volcanol Geoth Res 169(1):48–60 and strength recovery. Geophys Res Lett 40:5658–
Neuberg JW, Tuffen H, Collier L et al (2006) The trigger 5664
mechanism of low-frequency earthquakes on Montser- Vasseur J, Wadsworth FB, Lavallée Y, Bell AF, Main IG,
rat. J Volcanol Geotherm Res 153:37–50. doi:10. Dingwell DB (2015) Heterogeneity: the key to failure
1016/j.jvolgeores.2005.08.008 forecasting. Sci Rep 5
Prousevitch AA, Sahagian DL, Anderson AT (1993) Vasseur J, Wadsworth FB, Heap MJ, Main IG, Lavallée
Dynamics of diffusive bubble growth in magmas: Y, Dingwell DB (2017) Does an inter-flaw length
isothermal case. J Geophys Res Solid Earth 98:22283– control the accuracy of rupture forecasting in geolog-
22307 ical materials? Earth Planet Sci Lett
Salvage R, Neuberg JW (2016) Using a cross correlation Voight B (1988) A method for prediction of volcanic
technique to refine the accuracy of the Failure Forecast eruptions. Nature 332:125–130
Method: application to Soufrière Hills volcano, Webb SL (1991) Shear and volume relaxation in Na2 Si2 O5.
Montserrat. J Volcanol Geotherm Res 324:118–133. Am Mineral 76(9–10):1449–1454
doi:10.1016/j.jvolgeores.2016.05.011 Webb SL (1992) Low-frequency shear and structural
Sammis CG, Ashby MF (1986) The failure of brittle relaxation in rhyolite melt. Phys Chem Mineral 19
porous solids under compressive stress states. Acta (4):240–245
Metall 34:511–526 Webb SL, Dingwell DB (1990a) The onset of
Simmons JH, Mohr RK, Montrose CJ (1982) non-Newtonian rheology of silicate melts. Phys Chem
Non-Newtonian viscous flow in glass. J Appl Phys Miner 17:125–132
53:4075–4080 Webb SL, Dingwell DB (1990b) Non-Newtonian rheol-
Sparks RSJ (1978) The dynamics of bubble formation and ogy of igneous melts at high stresses and strain rates:
growth in magmas: a review and analysis. J Volcanol experimental results for rhyolite, andesite, basalt, and
Geotherm Res 3:1–37 nephelinite. J Geophys Res Solid Earth 95:15695–
Stevenson RJ, Dingwell DB, Webb SL, Bagdassarov NS 15701
(1995) The equivalence of enthalpy and shear stress Zhang Y, Stolper EM, Wasserburg GJ (1991) Diffusion of
relaxation in rhyolitic obsidians and quantification of a multi-species component and its role in oxygen and
the liquid-glass transition in volcanic processes. water transport in silicates. Earth Planet Sci Lett
J Volcanol Geotherm Res 68:297–306 103:228–240
Thomas ME, Neuberg J (2012) What makes a volcano
tick-a first explanation of deep multiple seismic
184 F.B. Wadsworth et al.

Open Access This chapter is distributed under the terms The images or other third party material in this chapter
of the Creative Commons Attribution 4.0 International are included in the work’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in the credit line; if such
which permits use, duplication, adaptation, distribution material is not included in the work’s Creative Commons
and reproduction in any medium or format, as long as you license and the respective action is not permitted by
give appropriate credit to the original author(s) and the statutory regulation, users will need to obtain permission
source, a link is provided to the Creative Commons from the license holder to duplicate, adapt or reproduce
license and any changes made are indicated. the material.
 Volcano Seismology: Detecting
Unrest in Wiggly Lines

R.O. Salvage, S. Karl and J.W. Neuberg

Abstract
Seismology is a useful tool to gain a better understanding of volcanic
unrest in real time as it unfolds. The generation of seismic signals in a
volcanic environment has been linked to a number of different physical
processes occurring at depth, including fracturing of the volcanic edifice
(producing high frequency seismicity) and movement of magmatic fluids
(producing low frequency seismicity). Further classification of seismic
signals according to their waveform similarity, in addition to their
frequency content, allows greater detail in temporal and spatial changes of
seismicity to be detected. At Soufrière Hills volcano, Montserrat, one of
the target volcanoes of the VUELCO project, families of similar
waveforms provided valuable insight into evaluating the significance of
ongoing unrest. In June 1997 over 6000 more events were able to be
identified over a 5 day period of interest (22 to 25 June) by using families
of seismic events, rather than a standard amplitude-based detection
algorithm. In total, 11 families were identified, with the events clustering
into a number of swarms, suggesting a repeating and non destructive
cyclic source mechanism. Since each family is believed to represent a
distinct source location and mechanism, identifying 11 coexisting families

R.O. Salvage (&)

Observatorio Vulcanológico y Sismológico
de Costa Rica, Universidad Nacional,
Apartado Postal: 2386-3000, Heredia, Costa Rica
e-mail: [email protected]
S. Karl
Shell, NAM offices, Schepersmaat 2,
9405 TA Assen, The Netherlands
e-mail: [email protected]
J.W. Neuberg
School of Earth and Environment, Institute of
Geophysics and Tectonics, University of Leeds,
Leeds LS2 9JT, United Kingdom
e-mail: [email protected]

Advs in Volcanology (2019) 185–201

DOI 10.1007/11157_2017_11
© The Author(s) 2017
Published Online: 02 August 2017
186 R.O. Salvage et al.

reflects the complex diversity of physical processes which act simulta-

neous at this volcano. In July 2003, conditions at the volcano had clearly
changed since only one family of seismicity was identified. The source
location of this family appeared to shift with time from 8 July (when no
events from the family were identified) to 12 July (where most events had
a cross correlation coefficient over 0.9). In addition, the use of families
appears to greatly aid hindsight forecasting attempts for the large scale
dome collapses of 1997 and 2003 using the Failure Forecast Method.
Knowledge of the temporal and spatial extent of seismicity during periods
of unrest, its source mechanism and its relationship to physical processes
at depth is essential for decision and policy makers for risk mitigation.
However, the source mechanisms of such volcanic seismicity is still much
debated and appears to often be misinterpreted because of compromising
assumptions used in the numerical modelling of inverting such sources.
Use of a spatially extended source such as a ring fault structure, rather than
a single point for determining the origin of low frequency seismicity, is
now thought to be more realistic for the mechanism of such events since it
more accurately represents the movement of magma through a conduit.
However, use of this spatially extended source instead of a simple single
point results in a large underestimation of slip from P-wave amplitudes,
which may lead to an underestimation in magma ascent rates, with large
consequences for eruption forecasting. Additionally, the P-wave radiation
patterns exhibited by these two mechanisms are remarkably similar, and
can only be distinguished if the small radial radiation lobes can be
determined. In a volcanic environment this is extremely difficult due to
large uncertainties in earthquake source depth locations, and the
implementation of small aperture seismic networks.

Español
La sismología es una herramienta geofísica valiosa que brinda infor-
mación en tiempo real, permitiendo una mejor comprensión del compor-
tamiento de sistemas volcánicos que inician un proceso de reactivación o
de intensificación de la actividad. La generación de señales sísmicas en
ambientes volcánicos se ha relacionado con un número diverso de
procesos geofísicos que ocurren en el interior de los volcanes, incluyendo
fracturamiento del edificio volcánico (produciéndose sismicidad de alta
frecuencia) y movimiento de fluidos magmáticos (produciéndose sismi-
cidad de baja frecuencia). La clasificación de señales sísmicas basada en la
similitud de las formas de onda, además del contenido de frecuencias, ha
permitido detectar cambios temporales y espaciales de la sismicidad con
mayor detalle.
En el volcán Soufriere Hills en Monserrat, uno de los volcanes
investigados como parte del Proyecto VUELCO, el reconocimiento de
familias de señales sísmicas con formas de onda similares proveyó un
entendimiento valioso en la evaluación de la significancia de la
Volcano Seismology: Detecting Unrest in Wiggly Lines 187

reactivación de su actividad volcánica. En junio de 1997, se pudieron

identificar más de 6000 eventos ssmicos dentro de unperiodo particular de
4 días (entre el 22 y el 25 de junio) mediante la determinación de familias
de eventos sísmicos que fueron determinados por un algoritmo estándar de
detección basado en la amplitud de la señal sísmica. En total, 11 familias
de eventos sísmicos fueron identificadas, con grupos de eventos confor-
mando enjambres sísmicos, lo que sugiere un mecanismo cíclico repetitivo
de una fuente sísmica no destructiva. Ya que se considera que cada familia
representa una fuente con ubicación espacial y mecanismo distinto, la
identificación de 11 familias refleja la compleja diversidad de los procesos
geofísicos que operan simultáneamente en este volcán en particular. En
julio del 2003, el régimen del volcán cambió definitivamente ya que
solamente 1 familia de eventos sísmicos fue identificada. La ubicación de
la fuente sísmica de esta familia parece haber migrado con el tiempo entre
el 8 de julio (cuando ningún evento sísmico de esta familia fue
identificado) y el 12 de julio (cuando la mayoría de eventos tuvo un
coeficiente de cros-correlación superior a 0,9). Además, el establecimiento
de familias de sismos parece haber sido de gran ayuda en los intentos de
pronosticar los colapsos del domo en gran escala de 1997 y el 2003
utilizando el método determinístico de pronóstico de rompimiento por
fatiga (Failure Forecast Method, FFM). El conocimiento de la extensión
temporal y espacial de la sismicidad durante periodos de reactivación
volcánica, el mecanismo de la fuente, y su relación con los procesos
geofísicos a niveles profundos son aspectos esenciales para los tomadores
de decisiones y ejecutores de políticas relacionadas con la mitigación de
riesgos.
En general, los mecanismos de la fuente generadora de la sismicidad
volcánica continúan aún bajo gran debate y parecen ser con frecuencia mal
interpretados debido a simplificaciones comprometedoras usadas en los
modelos numéricos de inversión de dichas fuentes. El uso de una fuente
espacial extendida tal como una falla estructural tipo anular, en lugar de un
solo punto para determinar el origen de sismicidad de baja frecuencia, se
piensa es más realista para visualizar el mecanismo de este tipo de eventos
ya que representa con más exactitud el movimiento del magma a través de
un conducto. Sin embargo, el uso de este modelo resulta en la
subestimación significativa del deslizamiento en la falla determinada a
través de las amplitudes de las ondas P, lo que podría llevar a una
subestimación de las tasas de ascenso del magma, con consecuencias
cruciales para el pronóstico de erupción inminente. Por otra parte, los
patrones de radiación de la onda P mostrados por estos dos mecanismos
son marcadamente similares, y pueden ser solamente distinguidos si los
pequeños lóbulos radiales de radiación pueden ser determinados. En un
contexto volcánico, esta es una tarea extremadamente difícil debido a la
gran incertidumbre inherente a la localización de la profundidad de la
fuente sísmica, y a la implementación de redes sísmicas de pequeña
extensión.
188 R.O. Salvage et al.

Keywords


Volcano seismology Families of similar seismic events
Magmatic


source mechanism Eruption forecasting

Index Terms
Volcano seismology

Soufrière Hills

Chiles-Cerro Negro Low





frequency seismicity Families Cross correlation Forecasting Failure




forecast method Source mechanisms Point source CLVD Ring





fault Spatially extended source P-wave radiation patterns

remotely analysed in real-time, often by an

Volcanic Unrest
automated system, if a number of sensors are
placed around the volcano. Deviation from the
The monitoring of active volcanoes for the pro-
background level is also often easier to determine
tection of society has excelled in recent decades
for seismicity than for other signals.
fuelled by rapid technological advances, which
have allowed the development and deployment
of more cost-effective monitoring solutions. It is
now possible to detect geophysical and geo-
Seismic Event Characterisation
chemical signals from a volcano which previ-
A wide variety of seismic signals exist within
ously would have been below the detection
volcanic settings, associated with magmatic and
threshold. The routine monitoring of volcanoes
hydrothermal fluid movement at depth, pressur-
during periods of quiescence is crucial, although
ization of the volcanic edifice, and/or the surface
not always feasible, in order to assess back-
manifestation of the interaction of these pro-
ground levels of activity at volcanoes, and thus
cesses. The variety of signals is a reflection of the
more rapidly detect the onset of future unrest.
number of different processes and the great
Most simply, volcanic unrest is defined as a
structural heterogeneities found within this con-
deviation from background levels of activity
tex. The characterisation of seismicity can be
towards a level which is a cause for concern over
based upon waveform similarities, but is tradi-
short time scales of hours to days. Volcanic
tionally based upon the signals’ time and fre-
unrest does not necessarily lead to eruption,
quency characteristics: different bands of
although this is the most likely outcome, with
frequency relate to different active source pro-
64% of 228 volcanoes that have experienced
cesses at depth, which can then be distinguished
unrest since the year 2000 culminating in erup-
from one another, although the frequency bands
tive events (Phillipson et al. 2013).
associated with each process may overlap (Lahr
Current monitoring efforts at volcanoes can be
et al. 1994).
grouped into three categories: measurements of
The identification of seismic events in vol-
surface degassing; deformation; and seismic
canic settings can assist with the detection of the
activity, which are all thought to result from the
onset and cessation of unrest, while the spatial
movement of magmatic fluids at depth, and
and temporal patterns of occurrence may be
therefore may provide key insights into an
informative of magma movement and changes in
impending eruption. Unrest must be detected on
stress at depth, as well as the spatial extent of
short timescales which is appropriate for
concern. An understanding of the physical pro-
decision-making, and therefore seismicity has
cesses occurring at depth is essential if accurate
remained a primary monitoring tool, as it can be
Volcano Seismology: Detecting Unrest in Wiggly Lines 189

5
x 10 High Frequency Event

Velocity (m/sec)
2
(a) (b)

Amplitude
2000
0
1000

−2 0
0 10 20 30 40 0 5 10 15
Time (seconds) Frequency (Hz)
5
x 10 Low Frequency Event
Velocity (m/sec)

1
(c) (d)

Amplitude
2000
0
1000

−1 0
0 10 20 30 40 0 5 10 15
Time (seconds) Frequency (Hz)
4
x 10 Rockfall Event
Velocity (m/sec)

2
(e) (f)
Amplitude
2000
0
1000

−2 0
0 10 20 30 40 0 5 10 15

Fig. 1 Examples of waveforms and their frequency waveform with clear phase arrivals. c, d Low frequency
content seen in volcanic environments taken from waveforms with an emergent onset. Waveform filtered
Soufrière Hills Volcano, Montserrat in 1997. Soufrière between 0.5 and 5 Hz. e, f Rockfall event with classic
Hills Volcano was a target volcano identified by the “cigar” shape
VUELCO project for investigation. a, b High frequency

and timely forecasts of volcanic eruptions, and 2003), similar to the generation of tectonic
developing unrest scenarios, are to be made. earthquakes.
Low frequency (LF) seismic signals (Fig. 1c,
d) usually occupy the spectral range of 0.2 to
Classification by Frequency Content 5 Hz (Chouet and Matoza 2013), and are fre-
quently characterized by emergent P-wave onsets
Seismic signals originating from processes and lack of S-wave arrivals. It has been sug-
occurring at depth within the volcanic system are gested that the occurrence of low frequency
usually split into high- and low-frequency waveforms is linked to resonance of seismic
end-members, although in reality a continuum energy trapped at a solid-fluid interface either
across the spectrum exists between the two within a crack (e.g. Chouet 1988), or a volcanic
(Chouet and Matoza 2013). High frequency conduit (e.g. Neuberg et al. 2000). The trigger
seismic signals (Fig. 1a, b), also known as mechanism of such seismic energy is further
Volcano-Tectonic (VT) events, have energy disputed, with suggestions that it may be gener-
concentrated in the frequency range of 1 to ated by: (1) a stick-slip motion along the conduit
20 Hz, generally peaking between 6 and 8 Hz walls as magma ascends (e.g. Iverson et al.
(Lahr et al. 1994). They are characterized by 2006); (2) the brittle failure of magma itself
clear, impulsive P- and S-wave arrivals, followed either through an increase in viscosity and strain
by a short coda. High frequency seismicity is rates (Lavallée et al. 2008), which may be due to
usually attributed to brittle failure within the an increase in the ascent rate of magma through
volcanic edifice, where magmatic processes cre- the conduit (Neuberg et al. 2006), changes in the
ate enough elastic strain to force the surrounding crystal and/or bubble concentration in the magma
rocks into failure (Arciniega-Ceballos et al. (Goto 1999), or through a change in the
190 R.O. Salvage et al.

geometry of the conduit (Thomas and Neuberg events) form a “cigar shaped’’ waveform with an
2012); (3) the interaction between the magmatic emergent onset (Fig. 1e, f), whereby there is an
and hydrothermal system at depth (e.g. Nakano initial increasing amplitude of the waveform as
and Kumagai 2005); or (4) through slow rupture the amount of material falling down slope
and failure of unconsolidated material on vol- increases. Pyroclastic flow signals are distin-
canic slopes (Bean et al. 2014). guishable from rockfalls since their waveforms
Many volcanic seismic events fall between are at least an order of magnitude larger and they
these two end-member categories and are termed often occur over a longer duration since larger
“hybrid” events. Hybrid events are characterised amounts of material are involved moving down
by a high frequency onset with a long resonating slope (De Angelis et al. 2007), however the two
low frequency coda, therefore distributing energy are likely to exist on a continuum.
across a wider frequency spectrum (Chouet and
Matoza 2013). Hybrid and LF events are often
classified in the same group of volcanic seis- Classification by Waveform Similarity
micity, since source and path effects can result in
a LF event recorded at one station being recorded Seismic waveforms can also be classified
as a hybrid event at another. according to their similarity with other detected
With the deployment of broadband sensors in seismic events. The frequency content of seismic
many volcanic environments, it is now possible waveforms is indicative of the active processes
to detect seismicity within a much wider fre- that may be occurring within the volcanic envi-
quency band, up to 120 s periods (Chouet and ronment and the source mechanism involved in
Matoza 2013), known as Very Long Period the generation of such seismicity. The further
(VLP) earthquakes. VLP events occupy the classification of seismic events into families
spectral range below 0.01 Hz. The generation of which all have a similar waveform shape, as well
these waveforms is not yet fully understood, in as the same frequency content, allows the
particular how such long wavelengths can be depiction of temporal and spatial changes in the
generated in apparently small source volumes, source mechanism and the source location on a
although it has been linked to perturbations in the much smaller scale (e.g. Thelen et al. 2011;
flow of fluid or gas within pressurized volcanic Salvage and Neuberg 2016). For example, the
conduits or cracks (e.g. Dawson et al. 2011). relative relocation of families of similar seismic
Their large wavelength, sometimes over one events at Soufrière Hills volcano, Montserrat has
hundred kilometers, means few path effects on produced very precise source locations (e.g. De
the waveform and as such if identified, these Angelis and Henton 2011). By definition, fami-
waveforms provide an excellent choice for per- lies of seismic events must be generated by the
forming waveform inversion techniques to iden- same mechanism and at the same location in
tify source characteristics (Chouet and Matoza order for the detected waveforms to have the
2013). same shape at the seismometer, and therefore
Furthermore, seismicity can be generated by changes in either of these parameters affect the
surface processes, such as landslides, rockfall similarity of waveforms. In many instances it is
events, pyroclastic flows and lahars (Fig. 1e, f). assumed that families of seismic events are
These are particularly dominant during dome generated by the same mechanism and within a
building eruptions and at volcanoes with glaciers similar source location, estimated at between one
during the spring and summer months due to quarter and one tenth of the wavelength (Geller
partial melting of the ice (McNutt 2005). These and Mueller 1980; Neuberg et al. 2006).
signals can be exploited to determine the size and Waveform similarity in terms of shape and
magnitude of such events, their location and their duration can be evaluated by cross correlation.
direction of travel (e.g. De Angelis et al. 2007). Identical signals will result in a cross correlation
Typically rockfall events (small free falling rock coefficient of 1 or −1, dependent upon their
Volcano Seismology: Detecting Unrest in Wiggly Lines 191

0 1

50
0.95

Event Number (Temporal Order) 100

0.9
150

200 0.85

250
0.8
300

350 0.75

400
0.7
450

0.65
0 100 200 300 400
Event Number (Temporal Order)

Fig. 2 Cross Correlation Matrix of events identified at greater than 0.7 are shown on a colour scale, with those
Soufrière Hills volcano, Montserrat at a single station in close to one being more similar. The autocorrelation of
June 1997. A total of 486 events were identified on 24 each event with itself is shown in dark red along the
June 1997 and are shown in temporal order along the x diagonal and is equal to a cross correlation coefficient of 1
and y axis. Events with a cross correlation coefficient of

relative polarity. Signals with no correlation each individual seismic event with every other
result in a cross correlation coefficient of 0. seismic event. The result are typically presented
A threshold must be chosen above which wave- as a similarity matrix, as seen in Fig. 2, where
forms can be considered similar. Waveforms events which are deemed to be similar are shown
which are deemed similar can be grouped toge- on the colour spectrum. However, such a matrix
ther into a family of events. The choice of simi- may include a number of families of similar
larity threshold is important: if it is too low there events since it only determines whether each
is a risk of placing events which are not similar event shows similarity to any of the other
into the same family; if it is too high similar earthquakes analysed. In order to identify fami-
events can be missed. Green and Neuberg (2006), lies of similar events, events with a high cross
Thelen et al. (2011) and Salvage and Neuberg correlation coefficient as decided by the user are
(2016) suggest a cross correlation coefficient grouped together and removed from the matrix.
threshold of 0.7, since this is significantly above This procedure is repeated across the entire
the correlation coefficient that can be produced investigated time period until all events have
from random correlations between noise and a been classified into a family, or have been
waveform. Higher cross correlation coefficient removed from the matrix. A master event is then
thresholds can be used to identify families of determined from each family of events as the
almost identical waveforms, however Petersen average of the stack of similar waveforms. This
(2007) suggests that this is probably not appro- is representative of the family in terms of
priate in volcanic settings due to additional noise waveform shape.
in this environment. Families of similar seismicity have been
The similarity between identified seismic identified at a number of active volcanoes around
events can be determined by cross correlating the world, including Redoubt volcano, Alaska
192 R.O. Salvage et al.

Dome collapse: 1655 UTC

Dominant Master
Master 001

Master 010

Master 014

Master 100

Master 106

Master 121

Master 136

Master 141

Master 210

Master 291

22nd 23rd 24th 25th 26th

Time

Fig. 3 Comparison of the timing and duration of swarms purpose of clarity on the plot and does not represent time
of families of events identified at a single station at or dominance; each master event is simply drawn below
Soufrière Hills volcano, Montserrat in June 1997. The the last so that all can be compared. Each coloured
timing of the dome collapse is represented by the vertical rectangular box represents the times when the families
line on the 25 June 1997. The y axis is only an indication were active during the 22–25 June analysis period
of each of the families present separated in space for the

(e.g. Buurman et al. 2013); Mt. St. Helens, USA before a lava dome collapse. The cross correlation
(Thelen et al. 2011); Colima, Mexico (Arámbu- technique identified 7653 similar seismic events
la-Mendoza et al. 2011); Merapi, Indonesia during the same time period, offering a five-fold
(Budi-Santoso and Lesage 2016); Katla, Iceland increase in the number of detected earthquakes
(Sgattoni et al. 2016); and Soufrière Hills volcano, (Salvage and Neuberg 2016).
Montserrat (Green and Neuberg 2006; Ottemӧller Low frequency families of seismicity identi-
2008; Salvage and Neuberg 2016). The identifi- fied during this unrest period at Soufrière Hills
cation of families rather than simply detecting volcano were followed by a dome collapse on 25
seismic events and classifying them according to June 1997. Soufrière Hills Volcano was chosen
their frequency content is advantageous as subtle by the VUELCO project as a target volcano due
temporal and spatial patterns can be identified, to the longevity of its dome building and collapse
allowing detailed source information to be cycles which have been ongoing since 1995,
uncovered. In addition, this technique dramati- providing a wealth of associated geophysical data
cally increases the number of identified events (Sparks and Young 2002; Wadge et al. 2014). In
from the continuous seismic record, since low total, 11 distinct seismic sources (i.e. 11 families
amplitude events and closely spaced events can of seismicity) were identified during this period of
still be identified. For example, using a standard unrest [Fig. 3; Green and Neuberg (2006); Sal-
amplitude-based detection algorithm, 1435 seis- vage and Neuberg (2016)], which all broadly
mic events were identified at Soufrière Hills vol- follow the same temporal pattern in the number of
cano, Montserrat between 22 and 25 June 1997, a identified swarms present over this time period.
period of interest due to increased seismicity However, the timing and duration of these
Volcano Seismology: Detecting Unrest in Wiggly Lines 193

Fig. 4 The evolution of the 1

dominant cross correlation
coefficient with time at
Soufrière Hills volcano, 0.9
Montserrat in July 2003.
A dome collapse event 0.8

Cross Correlation Coefficient

occurred at the time of the
vertical line (13:30 on 12 July
2003). The temporal gaps in 0.7
the data represent drops in the
seismometer recordings rather
0.6
than a change in the cross
correlation coefficient,
indicated by the white space 0.5
on the x-axis
0.4

0.3

0.2
8th 9th 10th 11th 12th 13th
July 2003

swarms can be seen to be different for each family of heightened seismicity at Soufrière Hills.
of similar seismic events. Low frequency seis- A migrating source mechanism can be identified
micity is associated with the movement of mag- from changing amplitudes in seismic events
matic fluid at depth, and therefore in this case within the same family, however this character-
would suggest cyclic flow dynamics to generate istic cannot be identified from analysing cross
such swarm-like behaviour. The source process correlation coefficients alone. The amplitudes of
for the generation of this seismicity must be stable seismic events within the single family identified
and non-destructive in order to be repeatable in July 2003 were relatively constant (Ottemӧller
(Green and Neuberg 2006; Petersen 2007), and 2008), suggesting the changing cross correlation
must be able to occur at a number of different coefficient is a consequence of a migrating source
locations and/or by a number of different sources location at depth, rather than an evolving source
at the same time in order to generate a number of mechanism. The generation of families ceased
active families of events, reflecting the complex immediately prior to the dome collapse event,
diversity of seismic sources and physical pro- and no similar earthquakes were detected after
cesses which act simultaneously at this volcano. the collapse (Fig. 4). This suggests that the
The identification of families can also be used physical conditions required for the generation of
to understand evolving seismicity with time. An families were not met in the hours before, and
evolving cross correlation coefficient with time, after, the collapse event.
if not an artefact of data processing, may be The analysis of families of seismicity in the
indicative of a migrating source location or time domain may also allow for the identification
source mechanism. This was observed at Sou- of spatial patterns in seismicity. Families detec-
frière Hills volcano, Montserrat in July 2003 ted at Chiles-Cerro Negro, a volcano within the
(Fig. 4; Salvage and Neuberg (2016)). The lar- Northern Andes on the border between Ecuador
gest dome collapse to date observed at this vol- and Colombia in October 2014, suggests distinct
cano occurred on 12 July 2003, with removal of temporal and spatial patterns of seismicity. The
210  106 m3 of material (Herd et al. 2005), last eruption of the volcanic complex of
following a 4 day period from 8 to 12 July 2003 Chiles-Cerro Negro is believed to have been
194 R.O. Salvage et al.

Cross Correlation Coefficient Matrix: CH1 Julian Day 293

1

50
0.95
Event Number (Temporal Order) 100

150
0.9
200

250 0.85
300

350 0.8

400
0.75
450

500
0.7
550

0.65
100 200 300 400 500
Event Number (Temporal Order)

Fig. 5 Cross Correlation Matrix of events identified being more similar. The autocorrelation of each event
using a simple amplitude based detection algorithm at with itself is shown in dark red along the diagonal and is
Chiles-Cerro Negro on 20 October 2014, at a single equal to a cross correlation coefficient of 1. Distinct
station. A total of 597 events were identified and are clusters of similar events can be identified, thought to
shown in temporal order along the x and y axis. Events suggest a temporal evolution in the dominant similar
with a cross correlation coefficient of greater than 0.7 are seismicity
shown on a colour scale, with those close to one (red)

3400 years ago, although seismicity has since October 2014 suggested not only the occurrence
been detected in the area, thought to be related to of families, but also their occurrence in distinct
an active hydrothermal system (Ruiz et al. 2013). temporal patterns. The clustering of similar
The volcanic complex is dissected by a large seismic events around the diagonal in a number
fault system, believed to have been active as of box-like formations within a similarity matrix
recently as 1868, when two large seismic events suggests that a number of sources were active for
occurred (Mw 6.6 and 7.2) (Beauval et al. 2010). discrete periods of time generating families of
Seismic activity increased in 1991 and then again seismicity (Fig. 5). The distinct clusters of sim-
in July 2013. However, a dramatic increase in ilar seismic waveforms may relate to a changing
seismicity from less than 50 events a day to over source location or mechanism at depth (Salvage
150 events occurred in October 2014, concen- 2015). Over this time period, no significant
trated beneath the summit of Chiles volcano at changes in the amplitude of events (indicative of
depths of less than 10 km (Ruiz et al. 2013). a changing source mechanism) were identified.
Although originally not a target volcano for the Since similar seismicity is thought to be gener-
VUELCO project, the volcanic complex of ated through a similar source mechanism and a
Chiles-Cerro Negro is an excellent example of a similar source location, distinct cluster of similar
re-awakening volcano, having shown no signs of seismicity is most likely related to its own dis-
magmatic unrest in recent history until the events tinct spatial region, which generated seismicity
of 2014. Analysis of seismicity identified in during distinct periods of time.
Volcano Seismology: Detecting Unrest in Wiggly Lines 195

The Source Mechanisms of Low assumptions such as slip along a single, unbent
Frequency Earthquakes surface, which is believed to be unrealistic.
An example of a spatially extended source
Since low frequency events, and in particular generating seismicity within a volcanic environ-
families, appear to be important in detecting ment is a volcanic conduit through which mag-
changes in unrest at volcanoes, it is important to matic fluids move. In these instances, the
understand their mechanism of generation. Syn- generation of low frequency seismicity may be
thetic modelling and moment tensor inversions of related to the brittle failure of magma itself
low frequency seismic wavefields are powerful (Neuberg et al. 2006; Lavallée et al. 2008;
tools for gaining information on the source Thomas and Neuberg 2012) or through a
mechanisms underlying volcanic earthquakes. stick-slip motion at the conduit edge (Iverson
Once instrument response and path effects have et al. 2006). In either case, shallow source depths
been accounted for, real data can be compared to (1–2 km) and short epicentral distances to seis-
synthetic models, and on the basis of a best-fit mic receivers (a few kilometers) suggest that a
approach the obtained model parameters allow spatially extended source is more realistic than a
insights into the nature and geometry of the single point source.
source (Chouet 1996; Shuler et al. 2013). How- The occurrence of slip (i.e. the generation of
ever, as the fundamental assumptions behind the the seismic energy itself) of spatially extended
commonly used moment tensor inversions are sources may either be instantaneous along two or
based on plane surface geometries which are more slip surfaces, or may occur on different slip
believed to be too simple to explain the genera- surfaces at different times, offset by a given time
tion of low frequency events in a volcanic envi- increment, delta t. A ring fault structure is a
ronment, the application to more complex seismic numerical description of seismogenic slip of
sources has so far been inconclusive. In the magma along all of the conduit walls within a
framework of the VUELCO project, slip along volcanic edifice, and can be numerically mod-
bent surfaces (a complex source) was thought of elled by considering a cylinder representing the
as the underlying physical motion responsible for volcanic conduit with instantaneously slipping
generating seismic energy. This novel way of double couple (single point) sources bounding
investigating low frequency earthquakes can the circumference (Fig. 6). Upward movement
explain several features of the earthquakes under inside the cylinder and downward movement
investigation without introducing compromising outside represents the movement of magma

Fig. 6 Schematic representation of a ring fault structure planar surface is host to a single double couple source i.e.
with movement directed upwards within the cylinder to a point source (labelled 1–8)
represent the flow of magma through a conduit. Each
196 R.O. Salvage et al.

through this channel. Other spatially extended

sources which may evoke the generation of low
frequency seismicity include: the upward move-
ment of magma through a narrow dyke, numer-
ically modelled by two oppositely directed
double couple sources; slip along distinct seg-
ments of the volcanic conduit i.e. due to geom-
etry changes (Thomas and Neuberg 2012),
numerically modelled with double couple sour-
ces on distinct segments; or the generation of a
number of seismic swarms of families of earth-
quakes occurring simultaneously (Salvage and
Neuberg 2016), which can be numerically mod- Fig. 8 Zoom of smaller radial lobes of P-wave radiation
elled by movement on two or more simultane- pattern for a spatially extended ring fault structure.
ously acting ring fault structures. Compressional lobes are red, dilatational lobes are blue
Green and Neuberg (2006) suggested that
accelerated magma movement at Soufrière Hills
volcano can be linked to observed deformation slip occurring and the slip rates. In the case of
cycles through low frequency seismic swarms, slip along a ring fault, P-wave amplitudes are
and that seismicity is only generated if significant greatly reduced due to destructive interference, in
magma movement takes place. Considering slip comparison to simple double couple (point)
through extended sources brings us one step sources. As a result, observed amplitudes yield
closer to estimating magma ascent rates. Once an underestimation of actual slip by more than a
calibrated, the link between observed waveform factor of 3 if interpreted as a point source, when
amplitudes and the amount of seismogenic slip in reality a spatially extended source acts at
occurring during a seismic swarm will yield depth. This underestimation in seismic moment
magma ascent rates and will ultimately con- consequently may lead to an underestimation of
tribute to forecasting volcanic eruptions more magma flow rate at depth, which in turn has
accurately. Point and extended source models severe implications for eruption forecasting (Karl
yield great differences in observed P-wave 2014).
amplitudes and waveforms, leading to remark- Furthermore, the P-wave radiation pattern for
able differences when interpreting the amount of a spatially extended ring fault structure (i.e. for
modelling the movement of magma within the
entire conduit made up of a number of double
couple sources) shows remarkable similarity to
the P-wave radiation pattern for a compensated
linear vector dipole (CLVD) source, which
instead is a conservation of energy (Fig. 7). The
derived ring fault radiation pattern shows rota-
tional symmetry around the depth axis, and
consists of a large compressional lobe directly
(a) CLVD source (b) Ring fault structure
above the source and an inversely polarised,
Fig. 7 P-wave radiation pattern generated for a CLVD dilatational lobe with the same amplitude below
source, and for an extended ring fault structure source. it. Only if the small radial P-wave radiation lobes
The red lobes are compressional, the blue lobes are can be determined for the ring fault structure
dilatational. In both radiation patterns a large compres-
sional lobe is found above the source, with dilatational (Fig. 8) is it possible to distinguish between these
lobes extending radially which can lead to confusion for two source mechanisms. Seismic networks with
interpretation of first motion polarity patterns small apertures typical in volcanic settings and
Volcano Seismology: Detecting Unrest in Wiggly Lines 197

uncertainties in earthquake source depth loca- potential correlation with the movement of
tions will likely lead to difficulties in distin- magmatic fluid at depth.
guishing between the two radiation patterns, The Failure Forecast Method (FFM) is based
since both can explain observed first motion on an empirical power-law relationship, which
polarity patterns of low frequency seismicity on relates the acceleration of a precursor ðd 2 X=dt2 Þ
volcanoes (Karl 2014). to the rate of that precursor ðdX=dtÞ (Voight
1988) method by:
 a
Forecasting Eruptive Activity d2 X dX
¼K ð1Þ
dt2 dt
The ability to forecast the timing, intensity and
type of volcanic activity is one of the key issues where K and a are empirical constants. X can
facing volcanologists today. The most notable represent a number of different geophysical pre-
instances of successful volcanic forecasting use cursors, for example low frequency seismic event
precursory activity at andesitic-dacitic volcanoes. rate (Salvage and Neuberg 2016), event rate of
The cataclysmic eruption of Mt. Pinatubo, all recorded seismicity (Kilburn and Voight
Philippines on 15 June 1991 was preceded by at 1998), or the amplitude of seismic events (Ortiz
least two months of heightened seismicity (Har- et al. 2003). The parameter a is thought to range
low et al. 1996). With increases in seismicity and between 1 and 2 in volcanic environments
an alarming sudden drop in SO2 , scientists were (Voight 1988), or may even evolve from 1
able to successfully evacuate over 45,000 local towards 2 as seismicity proceeds (Kilburn 2003).
people and 14,500 military personnel to safety by a has also been calculated in hindsight as high as
14 June, such that less than 300 people were 3.3 for accelerating seismicity in 1991 at Mt.
killed in the ensuing volcanic activity on 15 June. Pinatubo, although this extreme value appears
More recently, the 2010 eruption of Merapi, rare and was calculated with only a small amount
Indonesia, on 26 September was preceded by of seismic data (Smith and Kilburn 2010). An
approximately 6 weeks of precursory activity infinite dX=dt suggests an uncontrolled rate of
(Budi-Santoso et al. 2013): rates of seismicity change (a singularity) and in this environment is
and SO2 during this time were comparable to, or associated with an impending eruption. The
higher than, the highest rates observed during inverse form of dX=dt is linear if a ¼ 2, and
previous (smaller) Merapi eruptions (1992– therefore in this case the solution for the timing
2007), and rapid deformation was observed. of failure is a linear regression of inverse rate
Consequently, one day prior to the explosive against time, with the timing of failure relating to
eruption, several tens of thousands of people the point where the linear regression intersects
were evacuated from a radius extending 10 km the x-axis (Voight 1988).
from the volcano, resulting in a greatly lowered Although assuming that a ¼ 2 is the simplest
death toll of 35. method to estimate the timing of an eruption
Volcanic eruptions are often preceded by through a linear regression and therefore the most
accelerating geophysical signals, associated with common application of the FFM in hindsight
the movement of magma or other fluid towards analysis, some authors have suggested that it
the surface. Of these precursors, seismicity is at may not be an appropriate assumption for use
the forefront of forecasting volcanic activity with the FFM (e.g. Bell et al. 2011). Addition-
since it is frequently observed and the change ally, some authors have argued that a may evolve
from background level can be observed in real with time as precursory sequences develop,
time. Since forecasting of volcanic eruptions which is not detailed in the FFM (Kilburn 2003).
relies on the ability to determine the timing of As the FFM follows a least squares regression
magma reaching the surface, low frequency analysis when a is equal to 2, the residual error
seismicity may act as a forecasting tool due to its between the observed event rate and the mean
198 R.O. Salvage et al.

event rate of seismicity should follow a typical events followed a linear regression trend for
Gaussian distribution (Bell et al. 2011). Green- 4 days prior to a dome collapse event on 2 Jan-
hough and Main (2008) have suggested that uary 1990. Due to this trend, and the fact that the
since earthquake occurrence is a point process, seismic intensity was far above background
the rate uncertainties are best described by a levels, the Alaskan Volcano Observatory issued a
Poisson distribution. In this instance, a gener- “formal warning’’ of an impending eruptive
alised linear model (GLM) where a ¼ 1, rather event on the morning of the 2 January, a few
than a least squares regression model ða ¼ 2Þ hours before the eruption began, although the
may be more appropriate, since it can allow for a FFM calculations suggested an eruption was
distribution of data that is non-Gaussian (Bell likely within 0.5–2 days. A similar, if not clearer
et al. 2011). At Soufrière Hills volcano, however, trend, that supported the forecast was found
the use a GLM to forecast the timing of eruptive using the same precursory sequence but only
events in 1997 and 2003 failed to generate an using seismic events within the spectral range of
appropriate forecast (Salvage and Neuberg 1.3–1.9 Hz (Cornelius and Voight 1994), sug-
2016). Hammer and Ohrnberger (2012) sug- gesting an increased accuracy in forecasts when
gested that this may be related to the fact that a focusing on a single source process at depth.
Poisson process, and therefore the GLM, is a Swarms of seismic events, i.e. a number of
memoryless system, meaning that past events do similar events within a short period of time, with
not influence future patterns. A memoryless typical swarm durations of hours to day, are not
system is not consistent with the fundamental observed at all volcanoes, but have been com-
assumptions of the FFM, since previous geo- monly observed at Soufrière Hills volcano (e.g.
physical observables form the basis of such a Green and Neuberg 2006) and Redoubt volcano
forecast. (e.g. Buurman et al. 2013). Using precursory
One of the first instances of real time fore- seismicity and the FFM, Salvage and Neuberg
casting using the FFM was at Redoubt volcano, (2016) forecast in hindsight the timing of a dome
when the inverse average amplitude of seismic collapse event on 25 June 1997 at Soufrière

(a) (b)
1
Inverse average number of events in swarm (per 10 minutes)

0.9
Inverse Average event rate for swarm/10minutes

0.6
0.8

0.5 0.7

0.6
0.4
0.5
0.3
0.4

0.2 0.3

0.2
0.1
0.1

0 0
22nd 23rd 24th 25th 26th 22nd 23rd 24th 25th 26th
June 1997 Time

Fig. 9 Application of the FFM: the inverse average graphical representation of the FFM is depicted by the
event rate per 10 min within swarms from 22 to 25 June linear regression (it is assumed that a ¼ 2 for simplicity)
1997 at a single station at Soufrière Hills volcano. Each and the forecasted timing of failure can be read off the
data point represents the inverse average event rate for x-axis at the point where the linear regression crosses it.
each individual identified swarm of seismicity. The a All triggered low frequency seismicity. b Single family
vertical line represents the known timing of dome of similar seismicity
collapse on the 25 June 1997 at 16:55 UTC. The
Volcano Seismology: Detecting Unrest in Wiggly Lines 199

Hills, based upon accelerating rates of low fre- essential to ensure that the magma flow rate is
quency earthquakes which occurred in swarms, estimated accurately and therefore an accurate
rather than simply the number of low frequency forecast can be generated for the timing of
events over the precursory time period (Fig. 9). eruption. Using seismicity in combination with
Using the average event rate per swarm showed a other monitoring tools, we are now closer to
clearer accelerating pattern over the entire seis- gaining a better understanding of evolving
mic sequence, rather than using the traditional magmatic systems at depth.
method of binning data in units of time. More
accurate forecasts were determined when using Acknowledgements The past and present staff at the
Montserrat Volcano observatory are fully acknowledged
only one single family of similar events to fore- for their ongoing support in the upkeep and maintenance
cast the dome collapse, rather than all low fre- of the seismic network, and the sharing of data. All staff at
quency seismicity mixed together. A dome the Instituto-Geofisico in Ecuador are also fully
collapse on 12 July 2003 at Soufrière Hills vol- acknowledged for providing data from Chiles-Cerro
Negro volcano, for useful discussions regarding the vol-
cano was also more accurately forecast when canic complex, and for their continued monitoring efforts
using a single family of similar events, rather of all volcanoes in Ecuador. We thank two anonymous
than all low frequency seismicity, which occur- reviewers for their detailed comments and suggestions
red during the period of unrest (Salvage and which greatly enhanced the quality of this manuscript.
Muchas gracias también a Maria Martinez-Cruz y Javier
Neuberg 2016). Consequently, the use of families Pacheco por su ayuda con la traducción al español.
of seismicity, and therefore concentration upon a
single active system at depth, may allow a more
accurate forecast of the timing of an eruptive References
event in these instances.
Arámbula-Mendoza R, Lesage P, Valdés-González C,
Varley N, Reyes-Dávila G, Navarro C (2011) Seismic
activity that accompanied the effusive and explosive
Summary eruptions during the 2004–2005 period at Volcán de
Colima, Mexico. J Volcanol Geoth Res 205(1):30–46
Seismology is a powerful tool which can be used Arciniega-Ceballos A, Chouet B, Dawson P (2003)
to understand processes occurring at depth, and Long-period events and tremor at Popocatepetl vol-
cano (1994–2000) and their broadband characteristics.
their relationship to the surface at volcanoes, Bull Volc 65(2):124–135
especially since seismic events are easily detec- Bean CJ, De Barros L, Lokmer I, Métaxian J-P, O’Brien
ted and the deviation from the background level G, Murphy S (2014) Long-period seismicity in the
is often notably pronounced. An increase in shallow volcanic edifice formed from slow-rupture
earthquakes. Nat Geosci 7(1):71–75
seismicity may be an indication of volcanic Beauval C, Yepes H, Bakun WH, Egred J, Alvarado A,
unrest, since it relates to a number of physical Singaucho J-C (2010) Locations and magnitudes of
processes at depth including brittle failure and historical earthquakes in the Sierra of Ecuador (1587–
fracturing of the conduit or of the surrounding 1996). Geophys J Int 181(3):1613–1633
Bell A, Naylor M, Heap M, Main I (2011) Forecasting
edifice (high frequency seismicity) and the volcanic eruptions and other material failure phenom-
movement of magmatic fluids at depth (low fre- ena: an evaluation of the failure forecast method.
quency seismicity). Using families of seismicity Geophys Res Lett 38:L15304
as an indicator of a single active system at depth, Budi-Santoso A, Lesage P, Dwiyono S, Sumarti S,
Jousset P, Metaxian J-P et al (2013) Analysis of the
temporal and spatial patterns in the seismic seismic activity associated with the 2010 eruption of
events can be used to assess the potential Merapi Volcano, Java. J Volcanol Geoth Res
migration of seismicity towards the surface, as 261:153–170
well as to forecast the timing of volcanic eruptive Budi-Santoso A, Lesage P (2016) Velocity variations
associated with the large 2010 eruption of Merapi
events related to the acceleration of seismicity. volcano, Java, retrieved from seismic multiplets and
However, a full understanding of the source ambient noise cross-correlation. Geophys J Int 206
mechanism of the generated seismic events is (1):221–240
200 R.O. Salvage et al.

Buurman H, West ME, Thompson G (2013) The Karl S (2014) The source mechanisms of low frequency
seismicity of the 2009 Redoubt eruption. J Volcanol seismic events on volcanoes. Ph.D. thesis, University
Geoth Res 259:16–30 of Leeds, UK, Online at: http://etheses.whiterose.ac.
Chouet B (1988) Resonance of a fluid-driven crack: uk/id/eprint/8406
radiation properties and implications for the source of Kilburn C (2003) Multiscale fracturing as a key to
long-period events and harmonic tremor. J Geophys forecasting volcanic eruptions. J Volcanol Geoth Res
Res Solid Earth (1978–2012) 93(B5), 4375–4400 125(3–4):271–289
Chouet BA (1996) New methods and future trends in Kilburn CR, Voight B (1998) Slow rock fracture as
seismological volcano monitoring. In: Monitoring and eruption precursor at Soufriere Hills volcano Montser-
Mitigation of Volcano Hazards. Springer, pp. 23–97 rat. Geophys Res Lett 25(19):3665–3668
Chouet BA, Matoza RS (2013) A multi-decadal view of Lahr J, Chouet B, Stephens C, Power J, Page R (1994)
seismic methods for detecting precursors of magma Earthquake classification, location, and error analysis
movement and eruption. J Volcanol Geoth Res in a volcanic environment: implications for the
252:108–175 magmatic system of the 1989–1990 eruptions at
Cornelius R, Voight B (1994) Seismological aspects of Redoubt Volcano, Alaska. J Volcanol Geoth Res 62
the 1989–1990 eruption at Redoubt Volcano, Alaska: (1):137–151
the materials Failure Forecast Method (FFM) with Lavallée Y, Meredith P, Dingwell D, Hess K-U, Wasser-
RSAM and SSAM seismic data. J Volcanol Geoth Res mann J, Cordonnier B, Gerik A, Kruhl J (2008)
62(1):469–498 Seismogenic lavas and explosive eruption forecasting.
Dawson PB, Chouet BA, Power J (2011) Determining the Nature 453(7194):507–510
seismic source mechanism and location for an explo- McNutt SR (2005) Volcanic seismology. Annu Rev Earth
sive eruption with limited observational data: Augus- Planet Sci 32:461–491
tine Volcano, Alaska. Geophys Res Lett 38(3):L03302 Nakano M, Kumagai H (2005) Response of a hydrother-
De Angelis S, Henton S (2011) On the feasibility of mal system to magmatic heat inferred from temporal
magma fracture within volcanic conduits: constraints variations in the complex frequencies of long-period
from earthquake data and empirical modelling of events at Kusatsu-Shirane Volcano, Japan. J Volcanol
magma viscosity. Geophys Res Lett 38(19):L19310 Geoth Res 147(3):233–244
De Angelis S, Bass V, Hards V, Ryan G (2007) Seismic Neuberg J, Luckett R, Baptie B, Olsen K (2000) Models
characterization of pyroclastic flow activity at Sou- of tremor and low-frequency earthquake swarms on
frière Hills Volcano, Montserrat, 8 January 2007. Nat Montserrat. J Volcanol Geoth Res 101(1–2):83–104
Hazards Earth Syst Sci 7:467–472 Neuberg J, Tuffen H, Collier L, Green D, Powell T,
Geller R, Mueller C (1980) Four similar earthquakes in Dingwell D (2006) The trigger mechanism of
central California. Geophys Res Lett 7(10):821–824 low-frequency earthquakes on Montserrat. J Volcanol
Goto A (1999) A new model for volcanic earthquake at Geoth Res 153(1):37–50
unzen volcano: melt rupture model. Geophys Res Lett Ortiz R, Moreno H, GarcÌ A, Fuentealba G, Astiz M, Peña
26(16):2541–2544 P, Sánchez N, Tárraga M (2003) Villarrica volcano
Green D, Neuberg J (2006) Waveform classification of (Chile): characteristics of the volcanic tremor and
volcanic low-frequency earthquake swarms and its forecasting of small explosions by means of a material
implication at Soufrière Hills Volcano, Montserrat. failure method. J Volcanol Geoth Res 128(1):247–259
J Volcanol Geoth Res 153(1):51–63 Ottemöller L (2008) Seismic hybrid swarm precursory to
Greenhough J, Main I (2008) A poisson model for a major lava dome collapse: 9–12 July 2003, Soufriere
earthquake frequency uncertainties in seismic hazard Hills Volcano, Montserrat. J Volcanol Geoth Res 177
analysis. Geophys Res Lett 35(19):L19313 (4):903–910
Hammer C, Ohrnberger M (2012) Forecasting Petersen T (2007) Swarms of repeating long-period
seismo-volcanic activity by using the dynamical earthquakes at Shishaldin Volcano, Alaska, 2001–
behavior of volcanic earthquake rates. J Volcanol 2004. J Volcanol Geoth Res 166(3):177–192
Geoth Res 229:34–43 Phillipson G, Sobradelo R, Gottsmann J (2013) Global
Harlow DH, Power JA, Laguerta EP, Ambubuyog G, volcanic unrest in the 21st century: an analysis of the
White RA, Hoblitt RP (1996) Precursory seismicity first decade. J Volcanol Geoth Res 264:183–196
and forecasting of the June 15, 1991, eruption of Ruiz G, Cordova A, Ruiz M, Alvarado A (2013) Informe
Mount Pinatubo. Eruptions and lahars of Mount Tecnico de los volcanes Cerro Negro y Chiles. Tech.
Pinatubo, Philippines, Fire and mud, pp 223–247 Rep., IG-EPN, in Spanish
Herd RA, Edmonds M, Bass VA (2005) Catastrophic lava Salvage RO (2015) Using seismic signals to forecast
dome failure at Soufriere Hills volcano, Montserrat, 12– volcanic processes. Ph.D. thesis, University of Leeds,
13 July 2003. J Volcanol Geoth Res 148(3):234–252 UK, Online at: http://etheses.whiterose.ac.uk/12268/.
Iverson R, Dzurisin D, Gardner C, Gerlach T, LaHusen R, Restricted until April 2018
Lisowski M, Major J, Malone S, Messerich J, Moran S Salvage R, Neuberg J (2016) Using a cross correlation
et al (2006) Dynamics of seismogenic volcanic technique to refine the accuracy of the failure forecast
extrusion at Mount St Helens in 2004–05. Nature method: application to Soufrière Hills volcano,
444(7118):439–443 Montserrat. J Volcanol Geoth Res 324:118–133
Volcano Seismology: Detecting Unrest in Wiggly Lines 201

Sgattoni G, Jeddi Z, Gudmundsson Ó, Einarsson P, Thelen W, Malone S, West M (2011) Multiplets: their
Tryggvason A, Lund B, Lucchi F (2016) Long-period behavior and utility at dacitic and andesitic volcanic
seismic events with strikingly regular temporal pat- centers. J Geophys Res Solid Earth (1978–2012) 116
terns on Katla volcano’s south flank (Iceland). J Vol- (B8):B08210
canol Geoth Res 324:28–40 Thomas ME, Neuberg J (2012) What makes a volcano
Shuler A, Ekström G, Nettles M (2013) Physical mech- tick—a first explanation of deep multiple seismic
anisms for vertical-CLVD earthquakes at active vol- sources in ascending magma. Geology 40(4):351–354
canoes. J Geophy Res Solid Earth 118(4):1569–1586 Voight B (1988) A method for prediction of volcanic
Smith R, Kilburn C (2010) Forecasting eruptions after eruptions. Nature 332:125–130
long repose intervals from accelerating rates of rock Wadge G, Voight B, Sparks R, Cole P, Loughlin S,
fracture: the June 1991 eruption of Mount Pinatubo, Robertson R (2014) An overview of the eruption of
Philippines. J Volcanol Geoth Res 191(1):129–136 Soufriere Hills Volcano, Montserrat from 2000 to
Sparks R, Young S (2002) The eruption of Soufrière Hills 2010. Geological Society, London, Memoirs 39(1):1–
Volcano, Montserrat (1995–1999): overview of sci- 40
entific results. Geol Soc Lond Mem 21(1):45–69

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 The Ups and Downs of Volcanic
Unrest: Insights from Integrated
Geodesy and Numerical Modelling

J. Hickey, J. Gottsmann, P. Mothes, H. Odbert, I. Prutkin

and P. Vajda

Abstract
Volcanic eruptions are often preceded by small changes in the shape of the
volcano. Such volcanic deformation may be measured using precise
surveying techniques and analysed to better understand volcanic pro-
cesses. Complicating the matter is the fact that deformation events (e.g.,
inflation or deflation) may result from magmatic, non-magmatic or
mixed/hybrid sources. Using spatial and temporal patterns in volcanic
deformation data and mathematical models it is possible to infer the
location and strength of the subsurface driving mechanism. This can
provide essential information to inform hazard assessment, risk mitigation
and eruption forecasting. However, most generic models over-simplify
their representation of the crustal conditions in which the deformation
source resides. We present work from a selection of studies that employ
advanced numerical models to interpret deformation and gravity data.
These incorporate crustal heterogeneity, topography, viscoelastic rheology
and the influence of temperature, to constrain unrest source parameters at
Uturuncu (Bolivia), Cotopaxi (Ecuador), Soufrière Hills (Montserrat), and
Teide (Tenerife) volcanoes. Such model complexities are justified by

J. Hickey and H. Odbert, previously at School of Earth

Sciences, University of Bristol, UK.

J. Hickey (&) H. Odbert

Camborne School of Mines, University of Exeter, Met Office, Exeter, UK
Exeter, UK
I. Prutkin
e-mail: [email protected]
Institute of Geosciences, Jena University, Jena,
J. Gottsmann Germany
School of Earth Sciences, University of Bristol,
P. Vajda
Bristol, UK
Earth Science Institute, Slovak Academy of
J. Gottsmann Sciences, Bratislava, Slovakia
Cabot Institute, University of Bristol, Bristol, UK
P. Mothes
Instituto Geofísico, Escuela Politecnica Nacional,
Quito, Ecuador

Advs in Volcanology (2019) 203–219

DOI 10.1007/11157_2017_13
© The Author(s) 2017
Published Online: 15 July 2017
204 J. Hickey et al.

geophysical, geological, and petrological constraints. Results highlight

how more realistic crustal mechanical conditions alter the way stress and
strain are partitioned in the subsurface. This impacts inferred source
locations and magmatic pressures, and demonstrates how generic models
may produce misleading interpretations due to their simplified assump-
tions. Further model results are used to infer quantitative and qualitative
estimates of magma supply rate and mechanism, respectively. The
simultaneous inclusion of gravity data alongside deformation measure-
ments may additionally allow the magmatic or non-magmatic nature of the
source to be characterised. Together, these results highlight how models
with more realistic, and geophysically consistent, components can
improve our understanding of the mechanical processes affecting volcanic
unrest and geodetic eruption precursors, to aid eruption forecasting, hazard
assessment and risk mitigation.

Extended Spanish Abstract

La deformación volcánica, caracterizada por pequeños cambios medibles
en la morfología del volcán, a menudo, pero no siempre, precede a una
erupción volcánica. Esta cuestión, sin embargo, se complica por el hecho
de que los eventos de deformación (por ejemplo, inflación o deflacción)
pueden ser el resultado de una fuente magmática, no magmática o de
fuentes mixtas/híbridas. Utilizando tanto la amplitud y patrones
espacio-temporales de datos de deformación volcánica registrados, así
como la utilización de modelos matemáticos, es posible inferir la
ubicación y la fuerza de la fuente impulsora subyacente. Estos métodos
pueden proporcionar información esencial para la evaluación y mitigación
de riesgos, así como para el pronóstico de erupciones volcánicas. Sin
embargo, la mayoría de los modelos genéricos son insatisfactorios en su
representación de las condiciones de la corteza en las que reside la fuente.
En este trabajo presentamos una selección de estudios que emplean
modelos numéricos avanzados para la interpretación de datos de
deformación y gravedad. Dichos datos incorporan la heterogeneidad de
la corteza, la topografía, la reología inelástica y los efectos
termo-mecánicos para constreñir los parámetros asociados a la fuente de
perturbación en cuatro sistemas volcánicos. Las complejidades de estos
modelos están justificadas por limitaciones geofísicas, geológicas y
petrológicas. El estudio realizado en el volcán Uturuncu, localizado en
Bolivia, destaca la importancia de la estructura sub-superficial y de los
procesos dependientes de tiempo en la fuente para explicar los patrones de
deformación espacial-temporal. La combinación de dichos resultados
indica un ascenso del magma de tipo diapírico. En el volcán Cotopaxi,
localizado en Ecuador, los nuevos modelos de inversión que emplean el
Análisis por Elementos Finitos esclarecen la ubicación y el volumen de
una intrusión magmática durante un episodio de actividad asísmico y no
eruptivo con una baja tasa de suministro de magma. Estos modelos
también proporcionan señales observables que podrían estar asociadas con
futura actividad volcánica intrusiva o eruptiva. El análisis de la
deformación intra-eruptiva en el volcán Soufrière Hills, en Montserrat,
The Ups and Downs of Volcanic Unrest: Insights … 205

mostró cómo la inflación registrada podría deberse a una serie de

reservorios magmáticos apilados o un depósito alargado verticalmente,
debido a respuestas termo-mecánicas de la corteza similares. Las
condiciones de falla derivadas para las tasas de suministro de magma y
reservorio son consistentes con las restricciones térmicas y mecánicas
independientes. Utilizando datos gravimétricos y la curiosa falta de
deformación asociada, el estudio en el volcán Teide, en Tenerife, se
identifican, de forma separada, las contribuciones de fuentes poco
profundas como de aquellas con más profundidad. La interpretación de
ambos elementos sugiere que una intrusión magmática profunda activó un
sistema hidrotermal superficial localizado encima de ésta, y por lo tanto
demuestra un efecto causal de un origen mixto. Estos casos de estudio
representan una contribución significativa para la comprensión de los
procesos volcánicos durante los periodos de actividad intra-eruptiva y no
eruptiva. La combinación de estos resultados enfatiza cómo una mejor
parametrización de condiciones mecánicas de la corteza puede alterar
fundamentalmente la forma en que el estrés y la tensión se reparten en la
sub-superficie. Del mismo modo, una topografía compleja, como es el
caso en los estratovolcanes con laderas empinadas, puede afectar la
partición de la deformación superficial. Estos dos efectos impactan la
inferencia de la localización de las fuentes y presiones magmáticas previo
al fallo y la erupción, e indican cómo los modelos genéricos pueden
conducir a interpretaciones engañosas debido a la simplificación de
inferencias acerca de la corteza y a espacios planos y a medias. Resultados
adicionales de la modelización son utilizados para inferir estimaciones
cuantitativas y cualitativas de la tasa de suministro de magma y el
mecanismo, respectivamente, contribuyendo así al entendimiento de la
dinámica del transporte del magma. Además, la inclusión simultánea de
datos de gravedad junto a mediciones de deformación, permiten la
caracterización de la naturaleza magmática o no magmática de la fuente.
Juntos, estos estudios destacan cómo los modelos que cuentan con
componentes más plausibles y geofísicamente consistentes, pueden
mejorar nuestro entendimiento de los procesos mecánicos que afectan la
reactivación volcánica y de precursores geodésicos de erupciones. Estos
también proporcionan un marco para ayudarnos a avanzar en el pronóstico
de una erupción, así como en la evaluación y mitigación de los riesgos,
proporcionando datos cuantitativos derivados de la modelización de
mecanismos físicos adecuados y robustos.

Keywords
Volcano deformation
Gravity
Modelling
Crustal mechanics

Geodesy

Palabras clave




deformación volcánica gravedad modelización mecánica de la corteza

geodesia
206 J. Hickey et al.

Introduction Sparks et al. 2012), and constraints on source

parameters (e.g., pressure or volume changes)
Volcano deformation is a key observable during from previous deformation episodes can help to
periods of volcanic unrest, and one of the main quantitatively improve these forecasts. However,
tools used to monitor developing crises (Sparks making the transition from surface deformation
et al. 2012). Non-magmatic causes of volcanic observations to subsurface source processes
deformation during an unrest period do not requires the use of geodetic models, which
involve movement of new magma and are most demand assumptions about crustal mechanics.
commonly related to active hydrothermal sys- Generic analytical models of volcanic defor-
tems (e.g., Fournier and Chardot 2012; Rouwet mation (e.g., Mogi 1958) are often oversimplified
et al. 2014, and references therein). Magmatic and do not capture the intrinsic complexities of
causes reflect the active migration and accumu- subsurface systems, which are essential for reli-
lation of new magma, with usually deeper origins able assessment of causative processes, and thus
and wider deformation footprints. Hybrid mech- eruption forecasting and hazard assessment. They
anisms have also been proposed, where contri- usually represent the Earth’s crust as a homoge-
butions from magmatic and hydrothermal neous, isotropic, elastic, half-space with a flat, free
systems combine to produce a single complex surface. These assumptions can cause misleading
deformation pattern (Gottsmann et al. 2006a). interpretations when compared to numerical
A variety of both ground and satellite based models that allow for more realistic crustal con-
geodetic monitoring techniques are used to ditions (Masterlark 2007). Numerical models of
assess the spatial and temporal evolution of volcanic deformation are thus becoming increas-
volcanic related surface deformation. The two ingly popular due to this ability to estimate source
most common techniques employed today are the parameters in crustal conditions that are beyond
Global Positioning System (GPS) and Interfero- the analytical realm. These improvements are
metric Synthetic Aperture Radar (InSAR), which synonymous with the recent advances in geodetic
track changes in position due to the ground monitoring, which deserve a more in-depth anal-
deforming and complement each other with ysis than generic analytical models can offer.
advantages in their temporal and spatial cover- The extra complexities that numerical models
age, respectively. Volcano gravimetry studies, can account for relate to both the source itself, as
the monitoring and interpretation of continuous well as the crustal rocks in which it resides. The
or time-lapse spatio-temporal gravity variations, deformation source does not have to be repre-
can also provide additional information on any sented as a point or cavity, but can contain its
density changes associated with the driving own material properties (e.g., Hickey et al. 2013;
mechanism (Battaglia et al. 2008), and thus Gottsmann and Odbert 2014). Moreover,
enable an estimate of its nature (e.g., magmatic numerical models can incorporate inferences
or hydrothermal), which is particularly important from other geophysical, geological and petro-
in cases where there is no significant (observable) logical observables and thus maintain a higher
surface deformation. level of consistency by relaxing some of the
The magnitude, spatial pattern, and temporal assumptions that restrict generic analytical
evolution of volcanic deformation can be used to models. For example, this enables the inclusion
infer the location and ‘strength’ of a causative of subsurface heterogeneity, topography, vis-
subsurface source. These source parameters have coelastic rheology, and temperature-dependent
important implications for volcanic hazards and mechanics (e.g., Hickey et al. 2016, and refer-
risk mitigation. Geodetic data are also a key ences therein). Consequently, more robust source
component in eruption forecasting efforts (e.g., parameters and magma transport processes can
The Ups and Downs of Volcanic Unrest: Insights … 207

Table 1 Summary of geodetic modelling

Volcano Unrest FWM IVM Data Topo CMX TMX Model Ref
period
Uturuncu 1992–2006 ✓ – InSAR – ✓ – 2D Hickey et al. (2013)
Cotopaxi 2001–2002 ✓ ✓ EDM ✓ ✓ ✓ 3D Hickey et al. (2015)
Soufrière hills 2003–2005 ✓ – cGPS ✓ ✓ ✓ 2D Gottsmann and
Odbert (2014)
Central volcanic 2004–2005 – ✓ GPS & ✓ – – 3D Prutkin et al. (2014)
complex gravity
FWM forward modelling, IVM inverse modelling, Topo topography, CMX crustal mechanics, TMX thermomechanics,
2D two-dimensional axisymmetric, 3D three-dimensional, Ref reference

be inferred, thereby improving the understanding transferred through the Earth’s crust, from the
of the links between deformation and eruption. source to the surface. In this regard, as briefly
In this chapter we summarise the key findings mentioned above, generic analytical models are
from investigations of a variety of unrest epi- limited by their necessary assumptions of
sodes at a selection of the VUELCO target vol- homogeneous and elastic conditions throughout
canoes; Cotopaxi (Ecuador), Soufrière Hills the entire model domain. In reality, the Earth’s
(Montserrat, British West indies) and the Central crust is known to be layered, and volcanic
Volcanic Complex on Tenerife (Spain), as well regions in particular can have wide-ranging
as at Uturuncu volcano in Bolivia (Table 1). regions of stiff (high Young’s Modulus) and
These examples highlight not only different soft (low Young’s Modulus) rocks relating to the
timescales of unrest and spatial patterns of vol- type of volcanic deposit that formed them, e.g.,
cano deformation, but also the influence of lava flows (stiff) compared to tuffs (soft) (Gud-
crustal mechanical heterogeneity, thermal effects, mundsson 2011). This is what we call subsurface
and topography, on observed signals. The unrest heterogeneity. Where stiff and soft regions are
episodes are investigated using both forward and adjacent in the crust, complex subsurface parti-
inverse numerical modelling procedures and tioning alters the way stress and strain are
demonstrate the process of interpreting different transferred to the surface. The outcome is a dif-
geodetic data sets to constrain realistic source ferent surface deformation pattern compared to
parameters. Thus, the purpose of this chapter is one that would be seen if the crust was in fact
not to provide a detailed treatment of the math- homogeneous. Consequently, generic analytical
ematical and numerical approaches. Instead, it is models restricted to homogeneous crustal
intended to provide the reader with a general mechanics will not adequately represent likely
overview of the use of advanced numerical subsurface conditions, and the inferred source
models to infer volcanic unrest driving mecha- parameters from these models can be misleading.
nisms with realistic source characteristics. Seismic studies are capable of delineating
areas of relatively high and low seismic veloci-
ties. They can be used to estimate the dynamic
Implementing Complex Crustal Young’s Modulus, ED , Poisson’s Ratio, m, and
Mechanics density, q, of the crust (Brocher 2005):
"
 #,"
 #
One of the most important aspects of a defor- VP 2 VP 2
mation model is the representation of crustal m ¼ 0:5  2 1 ð1Þ
VS VS
mechanics. This has a fundamental control on the
way in which stress and strain is distributed and
208 J. Hickey et al.

q ¼ 1:6612VP  0:4721VP2 þ 0:067VP3 elastic mechanical behaviour it can be difficult to

 0:0043VP4 þ 0:000106VP5 ð2Þ constrain realistic source processes, particularly
if there is a complex temporal deformation pat-
VP2 qð1 þ mÞð1  2mÞ tern that can not be reproduced with the instan-
ED ¼ ð3Þ
ð1  mÞ taneous elastic stress-strain relationship, and may
be better represented by a non-instantaneous
where VP and VS are the primary and shear viscous stress response. A model that incorpo-
seismic velocities, respectively. The static rates temperature-dependent mechanics can
Young’s Modulus, E, is usually a factor of 2–9 overcome these difficulties. This is achieved
smaller than ED , and is more appropriate for using the Finite Element Method with a two-step
deformation studies as the dominant processes procedure (Hickey and Gottsmann 2014). First, a
taking place are significantly slower than the spatially-variable, stationary, temperature distri-
propagation of seismic waves (e.g., Gud- bution is derived, primarily dependent on the
mundsson 2011, and references therein). local geothermal gradient(s) and an assumed
Numerical approaches can incorporate these (magmatic) source temperature. This temperature
spatially-variable calculations of Young’s distribution, T, is then used to define a crustal
Modulus and Poisson’s Ratio by allowing the viscosity, g, for example through an Arrhenius
mechanical parameterisation of the crust to vary relationship, where:
within the model (Hickey and Gottsmann 2014).

This can be achieved in one-dimension, where H
g ¼ Ad exp ð4Þ
the values are only changed with depth, or in RT
three-dimensions, where the mechanical repre-
sentation additionally changes in the two hori- and Ad is the Dorn Parameter, H is the activa-
zontal axes. The Finite Element Method is the tion energy, and R is the universal gas constant.
most common technique used when incorporat- The viscosity is used in a viscoelastic material
ing subsurface heterogeneity. With this approach, representation (Hickey and Gottsmann 2014), so
a heterogeneous model domain can be built using where the viscosity is high the material response
multiple different sized homogeneous ‘blocks’ of is dominated by elastic behaviour, and when the
varying material properties (Hickey et al. 2013), viscosity is lower the rocks show an increas-
with continuous interpolations as a function of ingly higher tendency for viscous behaviour. To
depth or distance (Gottsmann and Odbert 2014), demonstrate the effect of temperature-dependent
or with a three-dimensional coordinate interpo- mechanics, and subsurface heterogeneity, the
lation (Hickey et al. 2016). following case studies include examples of
A further advantage of a numerical approach both.
using the Finite Element Method is the ability to
include a spatially variable inelastic rheology that
is dependent on an estimated temperature distri- Case Studies
bution. Rocks do not always behave in an elastic
manner (Ranalli 1995). Instead, where crustal Uturuncu
conditions are hotter than the brittle-ductile
transition, or subject to long-term loading, Uturuncu volcano is located in southern Bolivia,
rocks can deform like very high viscosity fluids, within the Altiplano-Puna Volcanic Complex. It
and so viscoelastic effects are more likely has been steadily inflating since at least 1992
(Ranalli 1995). This is especially important when (Pritchard and Simons 2002), and combined with
a hot magmatic component is assumed to be shallow seismicity and near-summit active
present, or where elevated geothermal gradients fumaroles represents a volcanic system showing
are observed. Also, when models are restricted to significant signs of unrest (Sparks et al. 2008).
The Ups and Downs of Volcanic Unrest: Insights … 209

Hickey et al. (2013) focused on the mechanism causing the observed spatial and temporal surface
driving the 70 km wide region of ground uplift uplift, whereas the previous models could only
between 1992 and 2006. The aim was to con- infer that a simple-geometry source was being
strain first-order source parameters that explain pressurised somewhere within the APMB
both the observed uplift rate of 1–2 cm/year and (Pritchard and Simons 2002). Hence, this also
the large spatial deformation footprint (Pritchard demonstrates how a pressure-time function plays
and Simons 2002). Stress and strain from pres- a first-order role in explaining time-dependent
surised finite sources were solved numerically deformation.
using Finite Element Analysis, accounting for Simultaneous work on an extended InSAR
both homogeneous and heterogeneous subsur- data set shows how the central uplift region at
face structure in elastic and viscoelastic rheolo- Uturuncu is surrounded by a ‘moat’ of subsi-
gies. Crustal heterogeneity was constrained from dence (Fialko and Pearse 2012). To explain this
seismic velocity data, which indicates a perva- observation they also constrained a model where
sive large low-velocity zone *17 km below the magma rises out and up from the APMB with a
surface. This is deduced to represent one of the diapiric-type ascent mechanism. Further evi-
world’s largest known regions of partial-melt: the dence for this magmatic process is available
Altiplano-Puna Magma Body (APMB) (Sparks through a complementary gravimetric study (del
et al. 2008). Potro et al. 2013). Therefore, this highlights how
The comparison between crustal heterogene- combinations of geodetic data and numerical
ity and homogeneity highlights the significant models can not only constrain more plausible
effect of a mechanically weak source-depth layer deformation source parameters, but can also infer
(Fig. 1). The weak layer, with a lower Young’s magma transport dynamics.
Modulus, alters surface deformation patterns by
accommodating more of the subsurface strain
than its surrounding layers, thereby acting as a Cotopaxi
mechanical buffer. Continuous and regular
time-dependent deformation, the long-lived nat- Cotopaxi is a large, glacier-clad stratovolcano
ure of the source, and an anomalously high situated in the Eastern Cordillera of the
regional crustal heat-flux break the assumption of Ecuadorian Andes. The 1 km3 glacier presents a
elastic conditions (e.g., Ranalli 1995), so a vis- substantial lahar risk to people in the surrounding
coelastic crustal rheology was tested, using the areas, and particularly to the 100,000 inhabitants
standard linear solid representation (e.g. Hickey that reside in the path of the 1877 lahar which
and Gottsmann 2014). The elastic models could descended the Inter-Andean Valley (Pistolesi
also only account for the spatial component of et al. 2013).
the observed uplift so their results were used Unrest was detected at Cotopaxi in 2001 and
solely to guide the parameters tested in the vis- 2002. There was a significant increase in the
coelastic models. A range of possible source amount of volcanic seismicity in the NE quadrant
geometries were assessed, but spherical and of the volcano, and this was originally interpreted
oblate shapes were rejected on the grounds of to represent a dyke intrusion with subsequent
their depth below the APMB and likely unsus- gas-release and resonance of the crack (Molina
tainable pressurisation given the expected crustal et al. 2008). Hickey et al. (2015) revisited this
mechanics. This left a prolate shaped source, unrest period, but approached it from a volcano
whose minimum size was determined using deformation viewpoint. Analysis of an electronic
maximum laboratory values for host-rock tensile distance meter (EDM) network over the 2001–
strength. The final preferred model suggests that 2002 period also indicated an asymmetric infla-
temporally-continuous pressurisation of a magma tion of the edifice that accompanied the recorded
source protruding from the top of the APMB is seismicity (Fig. 2). However, the irregular
210 J. Hickey et al.

(a) (b) A
B

D
E

5 km Homogeneous 5 km Heterogeneous

0.0 0.4 0.8 1.2 1.6 2.0

Radial Displacement [m]

8
(c) Homogeneous
Heterogeneous
7

6
LOS Displacement (cm)

0
0 5 10 15 20 25 30 35
Radial Distance (km)

Fig. 1 A comparison of the effect crustal heterogeneity into the softer layer (D), compared to the homogeneous
for the same prolate source geometry and depth. a, b The medium, which exhibits a concentric deformation pattern.
effect of a source depth soft layer on the subsurface c Modelled surface displacement profiles from the
deformation of a source. Both panels show the same homogeneous (a) and heterogeneous (b) models. The
source, embedded in a homogeneous (a) or heterogeneous subsurface layering in b alters the displacement pattern
domain (b). Colours relate to the radial displacement, and produced at the surface as the soft layer modifies the
the white shape shows the exaggerated outline of the subsurface strain partitioning. The blue shaded area
deformed source after the pressure is applied. In the represents the observed InSAR data and its estimated
heterogeneous model the source preferentially deforms error bounds

acquisition in time of the EDM data prevented volcanic deformation inversion study to explic-
any systematic comparison between the two data itly account for both subsurface heterogeneity
sets. To solve for the optimum deformation and surface topography while searching for a
source parameters, Hickey et al. (2015) imple- best-fit solution with a range of source shapes.
mented a novel numerical inversion procedure The method works by solving for the predicted
using Finite Element models. This is the first EDM deformation with an initial model and
The Ups and Downs of Volcanic Unrest: Insights … 211

source configuration. It then continually changes The inversion models converge on a shal-
the source location and/or overpressure, within low source beneath the SW flank. The indi-
some predefined parameter limits, to minimise vidual best-fit model is inferred to represent a
the misfit to the recorded EDM data (Fig. 2). small oblate-shaped magmatic reservoir,
After each inversion, the parameter limits (e.g., X approximately 4–5 km beneath the summit,
and Y coordinates) are reduced around the pre- with a volume increase of roughly
vious solution to produce a set of decreasing size 20  106 m3. A deformation source location in
‘Russian-doll-like’ parameter constraint grids the SW is substantially different to the NE
that ensure a robust solution. Within this work- location proposed by Molina et al. (2008)
flow, the Finite Element model geometry and when explaining the recorded seismicity.
mesh are automatically rebuilt, removing the Despite this, when the deformation source was
need for repeated manual editing. restricted to the NE quadrant the predicted

(a) 40
00 (b)
N BOSQ 2.5
−0.60°
LOMO
40
00
2.0
4000
00 5
40 40
00
3500 4
6
Objective [cm]

NASA 1.5
450
0
00

1 2 3
45

55

TAMB
0

7
00

500

−0.70° 1.0

4000
00 0.5
3500 40
1 - MARN
2 - PEDE
3 - HIQU
4 - RUDI
5 - JUAN 0
6 - LUIS 0 10 20 30 40 50 60 70 80 90
7 - LAVA 3 km Iteration
−78.50° −78.40°
(c) (d)
8
NASA−HIQU Data
10 NASA−MARN Model
NASA−PEDE 7
BOSQ−RUDI
LOMO−JUAN 6
LOMO−LUIS
5 TAMB−LAVA
Baseline Change [cm]

5
Contraction [cm]

4
0
3

2
−5
1

0
−10

−1
0 10 20 30 40 50 60 70 80 90 NH NM NP BR LJ LL TL
Iteration Baseline

Fig. 2 Deformation and modelling results from Coto- b The variation, and eventual reduction, of the misfit
paxi volcano. a Map of the EDM network operational objective function with each iteration from the final,
between 2001 and 2002 around the summit of the best-fit Finite Element inversion model. c The modelled
volcano, with the best-fit source from the Finite Element values from the final inversion for the EDM baseline
inversions indicated by a red star. The yellow squares changes converge to their near-absolute values after
represent the EDM base-stations and the orange triangles approximately 25 iterations. d The best-fitting model
are the reflecting prisms. The inset map and arrow shows (red circles) fits six out of seven of the EDM observations
the location of Cotopaxi within Ecuador, South America. (black circles with error bounds)
212 J. Hickey et al.

EDM measurements had a very poor fit to the period preceded a restart of eruptive activity and
observed data. renewed dome extrusion in August 2005.
To clarify the difference in source location Gottsmann and Odbert (2014) developed
between the seismic and geodetic studies, Hickey numerical models to test for the influence of tem-
et al. (2015) applied the best-fit source parame- perature- and time-dependent stress evolution in a
ters from the elastic inversion models in a suite mechanically heterogeneous crust to explain the
of temperature-dependent viscoelastic forward deformation data. Full details on the model setup
models to assess the rheology of the host-rock for and parameter derivation are given in Gottsmann
a range of thermal parameters. The results indi- and Odbert (2014) and not repeated here. They
cated the most likely subsurface conditions implemented two types of magma reservoir model.
would have promoted a large component of The first explored a series of pressurising,
viscous deformation and that the deformation vertically-stacked reservoirs as proposed by
source in the SW would have consequently been Hautmann et al. (2010), while the second explored
pressurised aseismically. Fluid migration from the time-dependent pressurisation of a single,
the SW along existing NNE-SSW trending faults vertically-elongated reservoir. Due to a similar
could have then caused the observed seismicity temperature distribution from both the stacked and
in the NE due to mass transport and excess pore single reservoirs, and similar resultant rheological
pressures. The lack of eruption following this crustal properties, both suites of models provided
2001–2002 unrest period and the aseismic nature equally good fits to the observed ground defor-
of the event suggests that the magma supply rate mation. The study could hence not discriminate
for this period was low. A higher magma supply between pressurisation in a magmatic plumbing
rate during a future unrest period would be more system consisting of either a single vertically
likely to produce seismicity around the reservoir, elongated reservoir or a series of stacked reser-
and could possibly indicate a level of unrest that voirs. Reservoir pressure changes between 4 and
signifies an increased likelihood of a forthcoming 7 MPa, for volumes between 60 and 100 km3 and
eruption. magma compressibility between 4  10−11 and
1  10−9 Pa−1, provided plausible thermome-
chanical model parameters to explain the defor-
Soufrière Hills mation data. The associated magma volume fluxes
are between 0.015 and 0.021 km3/year and match
Starting its latest eruptive episode in 1995 and those derived from thermal modelling of active
displaying a remarkable diversity of eruptive sub-volcanic systems (Annen 2009). Introducing a
activity, Soufrière Hills volcano (SHV) on deep-crustal hot zone in the model, which modu-
Montserrat (British West Indies) is one volcano lates the partitioning of strain into the hotter
where a correlation between observed deforma- underlying crust beneath the reservoir(s), promotes
tion and subsequent eruption can be directly a further reduction in reservoir overpressures to
established. Ground deformation data at SHV values of around 1–2 MPa upon reservoir failure
indicate cyclic behaviour of the andesitic mag- (Fig. 3). These pressure changes are significantly
matic system from periods of several hours to a lower than those derived from models assuming a
few years (Odbert et al. 2014). This section mechanically homogeneous and elastic crust. The
focuses on the analysis of intra-eruptive unrest deduced overpressures match those for sudden and
associated with island-wide ground uplift rapid transcrustal reservoir activation prior to
observed via a network of continuous GPS explosions at SHV from the analysis of volumetric
receivers between 28/07/2003 and 01/08/2005 strain data (Hautmann et al. 2014). The emerging
(Odbert et al. 2014). After a major lava dome eruption model at SHV hence involves the periodic
collapse in 2003, which marked the end of the failure of a compressible magma mush column
second phase of dome extrusion at SHV, this beneath the volcano.
The Ups and Downs of Volcanic Unrest: Insights … 213

Las Cañadas

The Las Cañadas caldera (LCC) hosts the central

volcanic complex (CVC) on Tenerife (Canary
Islands), which includes the twin-volcanoes of
Pico Viejo (PV) and Pico Teide (PT). Volcanic
unrest on Tenerife started in April 2004 and
ended a period of quiescence after the most
recent magmatic eruption on the island in 1909.
The unrest was geophysically characterised by
anomalous seismic activity including a number
of felt earthquakes and significant changes in the
acceleration of gravity, yet, an absence of sig-
nificant ground deformation (Gottsmann et al.
2006b). The initial interpretation of the gravity
changes by Gottsmann et al. (2006b) pointed
towards shallow (*2 km depth) fluid migration
as the main source of the unrest, possibly related
to a magma intrusion at greater depth. As part of
the VUELCO project, Prutkin et al. (2014)
revisited the microgravity data recorded between
May 2004 and July 2005 and inverted the
spatio-temporal residual gravity changes for the
gravitational attraction of three-dimensional line
segments.
The line segments are defined by 7 parame-
ters: two end point coordinates (X, Y, Z) and a
Fig. 3 a The temperature distribution (in K) of the 2-D line strength (see Table 2). The latter is a proxy
axisymmetric model domains, caused by a combination of
for the amplitude of sub-surface mass change,
a basal heat flux, the heat from the plumbing system
(magma reservoir plus feeder pipe), and a deep-crustal hot whereby a “weak” line segment represents a
zone. Background thermal distribution in the far field is small amount of added mass compared to a
caused exclusively by a basal heat flux of 0.09 W/m2. “strong” line segment of greater added mass. An
Arrows indicate deformation of the medium by reservoir
initial non-linear inversion yielded three line
pressurisation. Substantial deformation is accommodated
by hot and ductile parts of the mid and lower crust. b The segments of different strengths at depths between
model parameter space for reservoir volume, volume 1 km a.s.l. and 2 km b.s.l. (i.e. between 1 and
change, and magma compressibility (shown in log units 4 km beneath the surface). Two of the segments
on the colour bar) from the best fit model solutions to the
were located to the NW of the PV-PT complex,
2003–2005 intra-eruptive deformation time series at
Soufrière Hills Volcano. The overall fit quality decreases while a third segment was located below the SW
substantially toward higher compressibility (order rim of the LCC (sources marked Sh 1–3 in
10−9 Pa−1) and toward smaller reservoir volumes Fig. 4). The locations of the segments corre-
(<50 km3). The preferred parameter space is encompassed
spond broadly to zones of heightened seismic
between the mapped upper limit of compressibility (ULC)
and the lower limit of the magma reservoir volume activity and the unrest source location identified
(LLRV) earlier by Gottsmann et al. (2006b).
214 J. Hickey et al.

Table 2 UTM coordinates X (Easting in m), Y (Northing in m), and Z (elevation in m with respect to sea level) of the
end points of the modelled shallow (Sh) and deep (Dp) sources represented by line segments (see Fig. 4)
Line segment X1 Y1 Z1 X2 Y2 Z2 ΔM
Sh1 327,500 3,137,930 1280 330,330 3,137,260 600 0.31
Sh2 336,560 3,132,050 590 334,210 3,134,040 420 0.83
Sh3 335,460 3,120,730 1380 332,100 3,121,210 380 0.53
Dp1 334,890 3,133,910 −5440 334,630 3,134,200 −5680 7.25
Dp2 334,310 3,134,590 −5840 334,630 3,134,200 −5680 8.24
Mass additions (ΔM) are given in 1010 kg. Data Prutkin et al. (2014)

Fig. 4 Location of line segments from gravity data amount of added mass. The mass added to the deep
inversion. The figure shows the surface projections of sources is more than 10 times higher than to the shallow
three shallow sources (line segments Sh1–3 in turquoise) sources. Line segments Sh1–3 are interpreted to represent
and the two deep sources (line segments Dp1 and 2 in near-surface sources in the NW and SW part of the Pico
red) superimposed over a Google Earth image of the Teide (PT; 3718 m a.s.l.) and Pico Viejo (PV; 3135 m a.s.
Central Volcanic complex of Tenerife island (Spain). The l.) volcanic complex, associated with fluid migration as a
source locations are derived by a nonlinear inversion of result of an intrusion of magma at around 5.8 km b.s.l.
spatiotemporal residual gravity changes shown by (Dp sources). See Table 2 for details on segments and
coloured contours (in lGal) observed between 2004 and associated mass changes. The southern caldera rim (CR)
2005 (see Prutkin et al. (2014) for details on inversion is shown for reference and the inset shows a digital
routine). The thickness of a line segment is indicative of elevation model of Tenerife, with the study area identified
the “strength” of the deduced mass change; i.e., the by a white rectangle

However, closer inspection of the initial for the separate inversions of the two fields using
inversion results revealed that the line segments the three-dimensional line segment approxima-
represent the superposition of deep and shallow tion. The deep field inversion constrained two
seated sources. The decomposition of the gravity connected and strong line segments at a depth of
change data into shallow and deep fields (see about 5.8 km b.s.l. (marked Dp1 and Dp2 in
Prutkin et al. 2014 for details) provided the basis Fig. 4), while the inversion of the shallow field
The Ups and Downs of Volcanic Unrest: Insights … 215

identified three similarly weak line segments From this, it also follows that a horizontal vari-
situated at near-surface depths (<2 km from ation in mechanical properties would influence
ground surface; see Fig. 4 and Table 2). the horizontal location of a deformation source
The most plausible interpretation of the (e.g., Hickey et al. 2016).
inversion results is that the weak line segments A second effect of subsurface heterogeneity
represent sources dominated by hydrothermal relates to source pressure requirements. It is
fluids. In contrast, the deeper-seated sources can common with homogeneous crustal mechanics to
be interpreted as parts of an intrusion of new require unrealistically high source pressurisation
magma. This intrusion may have released fluids (>100’s MPa) when attempting to fit an observed
which consequently migrated towards shallower deformation signal, yet there is no petrological or
depths where they excited a shallow hydrother- mechanical evidence for such conditions. Theo-
mal system. The geophysical signals resulting retical work from a mechanical viewpoint sug-
from these coupled magmatic-hydrothermal gests that the maximum overpressure a source
processes point towards a hybrid source nature can sustain without failing is equal to the tensile
for the unrest on Tenerife in 2004–2005. The strength of the host rock (e.g., Gudmundsson
identified link between deep and shallow unrest 2011). Above this limit, a magmatic reservoir
sources suggests the presence of permeable would trigger a dyke intrusion. Work at Utu-
pathways for shallow fluid migration at the CVC. runcu and SHV shows that elastic heterogeneous
deformation models bring source over-pressure
requirements more in line with both in situ and
Discussion laboratory values of tensile strength. This is due
to a relative reduction in the average Young’s
The Effect of Crustal Mechanics Modulus above the source, compared to a higher
on Stress, Strain and Pressure homogeneous Young’s Modulus through the
entirety of the crust, and thus a greater amount of
The presented investigations at Uturuncu, Coto- deformation for a given pressure increment.
paxi and SHV all incorporate subsurface Furthermore, with a maximum value for the
heterogeneity, however the effects are somewhat source over-pressure, constraints can be placed
different. At Cotopaxi the final inferred defor- on the minimum size of a deformation source,
mation source is shallow, at a level where the given the two are directly linked. This allows, for
model does not incorporate substantial hetero- example, an estimate to be placed on the volume
geneity due to the limited amount of available of a magma reservoir. When multiple magmatic
seismic data. Hence, with this model configura- sources might be present numerical models can
tion, the heterogeneity has not played a signifi- additionally account for stress interactions
cant role in altering the location of the source. It between the two, something not considered by
is likely, however, that in reality the volcanic generic analytical models (Pascal et al. 2013).
edifice and shallow subsurface does have a cer- As a next step to subsurface heterogeneity, the
tain level of heterogeneity. It is therefore possible case studies at Cotopaxi and SHV incorporated
that had this been taken into account the inferred temperature-dependent mechanics to evaluate the
deformation source might be located more cen- effect of a viscoelastic rheology. At Uturuncu,
trally beneath the edifice. Conversely, at Utu- investigations were also carried out using a vis-
runcu and SHV subsurface heterogeneity played coelastic rheology without temperature-
a crucial role in determining the deformation dependence. In all three cases, a viscoelastic
source locations. In both cases, the inclusion of medium reduced the over-pressure requirements
vertical layering in the Young’s Modulus distri- further when compared to the purely elastic
bution altered the inferred depth of the source; models. This is due to the viscous expansion that
the heterogeneous models predict deeper sources follows an initial elastic inflation. The effect of
than the generic homogeneous analytical models. thermomechanics is greater than just modifying
216 J. Hickey et al.

source over-pressure requirements, however. and non-magmatic unrest, and plays a major role
Prolonged magma emplacement over thousands in probabilistic eruption forecasting (Rouwet
to millions of years in active volcanic areas et al. 2014).
builds up a significant thermal legacy within the
crust. This results in elevated geothermal gradi-
ents, and in the case of continued active mag- Application to Eruption Forecasting
matism, deep crustal hot-zones (Annen 2009).
These thermal perturbations are significant for The primary objective of the case studies was to
the transfer of stress and strain. For example, at develop advanced geodetic models to interpret
SHV, the combined thermomechanical effects of spatial and temporal deformation monitoring
a deep-crustal hot zone and hot encasing rocks signals, and provide better constraints on the
around a mid-crustal andesitic reservoir funda- subsurface processes causing volcanic unrest.
mentally alter the time-dependent subsurface This has been achieved by incorporating more
stress and strain partitioning upon priming of the plausible model components, such as subsurface
magma reservoir. These effects substantially heterogeneity, topography and temperature-
influence surface strains recorded by volcano dependent mechanics, to relax the assumptions
geodetic monitoring. that hinder analytical models and maintain con-
sistency with inferences from geophysics, geol-
ogy and petrology. Crucially, this highlighted the
Hybrid Unrest and Source importance of pressure-time functions and
Characterisation inelastic rheology in deciphering temporal
deformation patterns. In turn, a more thorough
An intrusion of magma commonly leads to the understanding of how a volcanic system behaves
exsolution of fluids upon decompression, and through time will benefit eruption forecasting, as
fluid migration (and accumulation) in itself can quantitative estimates of key parameters such as
produce measurable geodetic surface signals magma supply rate and mechanism can be
(e.g., Fournier and Chardot 2012; Rouwet et al. deduced.
2014, and references therein). A deep intrusion When only considering the spatial deforma-
of magma, such as the 2004–2005 unrest on tion pattern, this work has further demonstrated
Tenerife, may not necessarily lead to observable some of the pitfalls associated with models
surface deformation, meaning deformation data assuming homogeneous, elastic, half-spaces
on its own can not provide any meaningful (e.g., Mogi 1958). Crustal heterogeneity signifi-
insights on the source process(es). In this case, cantly effects the horizontal and vertical location
the combination of deformation and gravity sur- of a deformation source by altering the subsur-
veys allowed the characterisation of the unrest face strain distribution. Surface strain partition-
sources in much greater detail owing to their ing by complex topography is equally important,
density contrasts, relative depths and mass such as at steep-sided stratovolcanoes. The
additions. This highlights that, especially for the Cotopaxi case highlighted complex partitioning
case of hybrid mechanisms where both magmatic behaviour along deep ravines and adjacent lava
and hydrothermal components are present, flow ridges. If ground-based geodetic monitoring
multi-disciplinary geodetic surveys can provide sites are positioned at localities with steep topo-
valuable information on source characterisation graphic gradients, surface strain partitioning
to help distinguish the processes driving unrest. plays a first-order role and needs to be accounted
This identification and discrimination of sources for during data modelling. Consequently, esti-
driving volcanic unrest via mathematical mod- mates of source parameters from generic, ana-
elling of surface data is of vital importance for lytical models which cannot account for
hazard and risk characterisation, given the dif- heterogeneities or complex topography may
ferent connotations associated with magmatic skew eruption forecasting and risk mitigation
The Ups and Downs of Volcanic Unrest: Insights … 217

efforts. On the other hand, more reliable defor- Finite Element Method/Analysis a numerical
mation source locations from numerical models modelling technique that subdivides an
can be used to better estimate where and when an entire problem into a set of smaller ‘ele-
eruption might occur, and the associated hazard. ments’. Mathematical problems are then
It should also be pointed out that data limitations solved in each of the elements and com-
at some volcanoes may prevent complex bined together to calculate the response of
numerical models being used, in which case the whole object.
simpler models may be the only option and their Geodesy applied mathematics with the aim of
results should be carefully scrutinised and only measuring the geometric spatial representation
applied cautiously to further analyses. of the Earth and its gravitational field in
time-varying three-dimensions, as well as
their orientation in space. Volcano geodesy
Conclusions specifically deals with recording the spatial
and temporal patterns of crustal deformation
The case studies provide new interpretations of and gravimetric changes in volcanic settings.
volcanic processes during intra-eruptive and Gravimetry field of geophysics devoted to
non-eruptive unrest. They also provide observa- observing, processing and interpreting minute
tions for distinct magmatic settings, thus con- changes in the Earth’s gravitational field.
tributing to the ongoing global comparisons of Hydrothermal system an area in which heat
deformation and unrest, and the process of pat- and fluids from a partially molten magmatic
tern recognition for identification of eruption body interact with a multi-phase groundwater
precursors. The incorporation of system, causing chemical and thermal (heat-
multi-disciplinary data into integrated geodetic ing) perturbations to the water.
models has closed the gap between observations
Mechanics (crustal, thermal) the branch of
and interpretations of volcanic deformation and
physics that studies how stress can effect a
gravimetric changes. This will help to improve
physical object. Crustal mechanics relates to
eruption forecasting as it moves away from a
the values of the material parameters that
qualitative approach towards incorporation of
describe the mechanical behaviour of the
more quantitative data derived from Earth’s crust. Thermal mechanics relates to
well-constrained physical mechanisms. the variation in mechanical material properties
due to changes in temperature.
Acknowledgements Work presented herein has received
funding by the European Commission (FP7; Theme: Model (inverse, forward) a simulation of a
ENV.2011.1.3.3-1; Grant 282759: VUELCO). We are process. Inverse models solve ‘backwards’ to
very grateful to Alvaro Guevara and Irving Munguía
determine optimal model parameters that fit a
Gonzalez for their help translating Spanish, and to two
anonymous reviewers for their constructive comments. set of data. Forward models use predefined
constant input parameters to calculate the
expected model response.
Glossary Terms Rheology (crustal) the behavioural response of
the Earth’s crust to forces that act upon or
Deformation the action of changing the shape within it. See also elastic and viscoelastic.
of a physical object (e.g., a volcano) under the Strain the change in dimension of an object
influence of some stress. (e.g., DX) relative to the original dimension of
Elastic rheological behaviour in which an the object (e.g., X). It has no units, as the units
applied stress causes an immediate strain that cancel: DX=X.
is 100% recoverable when the stress is Stress a measure of force per unit area, with the
removed. See also rheology and viscoelastic. unit of pascals, Pa. 1 Pa = 1 N/m2.
218 J. Hickey et al.

Viscoelastic rheological behaviour in which an Hautmann S, Gottsmann J, Sparks RSJ, Mattioli GS,
applied stress causes an elastic response, fol- Sacks IS, Strutt MH (2010) Effect of mechanical
heterogeneity in arc crust on volcano deformation with
lowed by a delayed (viscous) response. See application to Soufrière Hills volcano, Montserrat,
also rheology and elastic. West Indies. J Geophys Res 115(B9):B09203
Hautmann S, Witham F, Christopher T, Cole P,
Linde AT, Sacks IS, Sparks RSJ (2014) Strain field
analysis on Montserrat (W.I.) as tool for assessing
Index Terms permeable flow paths in the magmatic system of
Soufriere Hills Volcano. Geochem Geophys Geosyst
Magma reservoir, deformation, gravity, GPS, 15:2013G. doi:10.1002/C005087
Hickey J, Gottsmann J (2014) Benchmarking and devel-
InSAR, Cotopaxi, Soufrière Hills, Las Cañadas, oping numerical finite element models of volcanic
Uturuncu, unrest, thermomechanics, modelling, deformation. J Volcanol Geoth Res 280:126–130
finite element, inversion/inverse model, crustal Hickey J, Gottsmann J, del Potro R (2013) The large-scale
mechanics, hybrid unrest, Young’s Modulus, surface uplift in the Altiplano-Puna region of Bolivia:
a parametric study of source characteristics and crustal
Poisson’s Ratio, viscosity. rheology using finite element analysis. Geochem
Geophys Geosyst 14(3):540–555
Hickey J, Gottsmann J, Mothes P (2015) Estimating
References volcanic deformation source parameters with a finite
element inversion: the 2001–2002 unrest at Cotopaxi
volcano, Ecuador. J Geophys Res Solid Earth 120
Annen C (2009) From plutons to magma chambers: (3):1473–1486
thermal constraints on the accumulation of eruptible Hickey J, Gottsmann J, Nakamichi H, Iguchi M (2016)
silicic magma in the upper crust. Earth Planet Sci Lett Thermomechanical controls on magma supply and
284(3–4):409–416 volcanic deformation: application to Aira caldera,
Battaglia M, Gottsmann J, Carbone D, Fernández J (2008) Japan. Sci Rep 6(August):32691
4D volcano gravimetry. Geophysics 73(6):WA3– Masterlark T (2007) Magma intrusion and deformation
WA18 predictions: sensitivities to the Mogi assumptions.
Brocher TM (2005) Empirical relations between elastic J Geophys Res 112(B06419)
wavespeeds and density in the Earth’s crust. Bull Mogi K (1958) Relations between the eruptions of various
Seismol Soc Am 95(6):2081–2092 volcanoes and the deformations of the ground surfaces
del Potro R, Dez M, Blundy J, Camacho AG, Gottsmann J around them. Bull Earthq Res Inst 36:99–134
(2013) Diapiric ascent of silicic magma beneath the Molina I, Kumagai H, Garca-Aristizábal A, Nakano M,
Bolivian Altiplano. Geophys Res Lett 40(10):2044– Mothes P (2008) Source process of very-long-period
2048 events accompanying long-period signals at Cotopaxi
Fialko Y, Pearse J (2012) Sombrero uplift above the Volcano, Ecuador. J Volcanol Geothermal Res 176
Altiplano-Puna magma body: evidence of a ballooning (1):119–133
mid-crustal diapir. Science 338(6104):250–252 Odbert HM, Stewart RC, Wadge G (2014) Cyclic
Fournier N, Chardot L (2012) Understanding volcano Phenomena at the Soufrière Hills Volcano, Montser-
hydrothermal unrest from geodetic observations: rat, The Eruption of Soufrière Hills Volcano, Montser-
insights from numerical modeling and application to rat from 2000 to 2010, vol 39. The Geological
White Island volcano, New Zealand. J Geophys Res Society, London, pp 41–60. doi:10.1144/M39.2
117(B11):B11208 Pascal K, Neuberg J, Rivalta E (2013) On precisely
Gottsmann J, Odbert HM (2014) The effects of modelling surface deformation due to interacting
thermo-mechanical heterogeneities in island-arc crust magma chambers and dykes. Geophys J Int 196
on time-dependent pre-eruptive stresses and the failure (1):253–278
of an andesitic reservoir. J Geophys Res. doi:10.1002/ Pistolesi M, Cioni R, Rosi M, Cashman KV, Rossotti A,
2014JB011079 Aguilera E (2013) Evidence for lahar-triggering
Gottsmann J, Rymer H, Berrino G (2006a) Unrest at the mechanisms in complex stratigraphic sequences: the
Campi Flegrei caldera (Italy): a critical evaluation of post-twelfth century eruptive activity of Cotopaxi
source parameters from geodetic data inversion. Volcano, Ecuador. Bull Volcanol 75(3):698
J Volcanol Geoth Res 150(1–3):132–145 Pritchard ME, Simons M (2002) A satellite geodetic
Gottsmann J, Wooller L, Marti J, Fernandez J, Cama- survey of large-scale deformation of volcanic centres
cho AG, Gonzalez PJ, Garcia A, Rymer H (2006b) in the central Andes. Nature 418(6894):167–171
New evidence for the reawakening of Teide volcano. Prutkin I, Vajda P, Gottsmann J (2014) The gravimetric
Geophys Res Lett 33(20) picture of magmatic and hydrothermal sources driving
Gudmundsson A (2011) Rock fractures in geological hybrid unrest on Tenerife in 2004/5. J Volcanol Geoth
processes. Cambridge University Press, New York Res 282:9–18
The Ups and Downs of Volcanic Unrest: Insights … 219

Ranalli G (1995) Rheology of the Earth. Chapman and Sparks RSJ, Folkes CB, Humphreys MC, Barfod DN,
Hall, London Clavero J, Sunagua MC, McNutt SR, Pritchard ME
Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J, (2008) Uturuncu volcano, Bolivia: volcanic unrest due to
Tonini R, Papale P (2014) Recognizing and tracking mid-crustal magma intrusion. Am J Sci 308(6):727–769
volcanic hazards related to non-magmatic unrest: a Sparks RSJ, Biggs J, Neuberg JW (2012) Monitoring
review. J Appl Volcanol 3:1–17 volcanoes. Science 335(6074):1310–1311

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Fluid Geochemistry and Volcanic
Unrest: Dissolving the Haze in Time
and Space

Dmitri Rouwet, Silvana Hidalgo, Erouscilla P. Joseph

and Gino González-Ilama

Abstract
The heat and gas released by a degassing magma affects the overlying
predominantly meteoric aquifers to form magmatic-hydrothermal systems
inside the solid body of a volcano. This chapter reviews how fluid
geochemical signals help to track the evolution throughout the various
stages of volcanic unrest. A direct view into a degassing magma is
possible at open-conduit degassing volcanoes. Nevertheless, in most cases
gas is trapped (i.e. scrubbed) by abundant water, leading to the loss of the
pure signal the magma ideally provides. Deciphering how magmatic gas
rises through, reacts, and re-equilibrates with the liquids in the
magmatic-hydrothermal system in time and space is the only way to
trace back to the pure signal. The most indicative magmatic gas species
(CO2, SO2–H2S, HCl and HF) are released as a function of their solubility
in magma. The less soluble gas species are released early from a magma at
higher pressure conditions (CO2) (deeper), whereas the more soluble
species are released later, at lower pressures (SO2, HCl and HF)
(shallower depth). When these gases hit the water during their rise
towards the surface, they will be more or less scrubbed. Depending on the
chemical equilibria inside the magmatic-hydrothermal system (e.g. SO2–
H2S conversion, acidity), the gas that eventually reaches the surface will
carry the history of its rise from bottom to top. Tracking volcanic unrest
implies a time frame; the kinetics of magma degassing throughout the
liquid cocktail inside the volcano impose the maximum resolution the
volcano provides and hence the monitoring time window to be adopted for

D. Rouwet (&) E.P. Joseph

Istituto Nazionale di Geofisica e Vulcanologia, Seismic Research Centre, University of the West
Sezione di Bologna, Via Donato Creti 12, 40128 Indies, St. Augustine, Trinidad & Tobago
Bologna, Italy G. González-Ilama
e-mail: [email protected] Centro de Investigaciones en Ciencias Geológicas,
S. Hidalgo Universidad de Costa Rica, San José, Costa Rica
Instituto Geofisico, Escuela Politécnica Nacional,
Quito, Ecuador

Advs in Volcanology (2019) 221–239

DOI 10.1007/11157_2017_12
© The Author(s) 2017
Published Online: 19 July 2017
222 D. Rouwet et al.

each volcano. Gas-dominated systems are “faster” and require a higher

monitoring frequency, water-dominated systems are slower and require a
lower monitoring frequency.

Resumen
El calor y gas liberados por la desgasificación del magma afecta los
acuíferos de origen predominantemente meteórico para formar sistemas
magmático-hidrotermales dentro el cuerpo sólido del volcán. Este capítulo
revisa como la geoquímica de fluidos puede ayudar a trazar la evolución a
través de las varias etapas de “unrest” volcánico. Una visión directa dentro
de un magma en desgasificación es solo posible para volcanes de conducto
abierto. Sin embargo, en la mayoría de las situaciones el gas queda
atrapado (i.e. “scrubbing”) en el agua, que conduce a la pérdida de la señal
pura que el magma idealmente puede proporcionar. Descifrar como el
magma sube a través de los líquidos, y reacciona y re-equilibra con ellos
dentro el sistema magmático-hidrotermal, en un marco de tiempo y
espacio, es la única manera para rastrear el origen de la señal del magma.
La desgasificación de magma se da por cuatro procesos: (1) durante la
subida de magma, (2) por la descompresión debido al eliminar una
porción del edificio volcánico, (3) debido a la convección interna dentro la
cámara magmática, o (4) después de “ebullición secundaria” siguiendo el
enfriamiento y consecuente cristalización. Las especies gaseosas magmáti-
cas más indicativas (CO2, SO2–H2S, HCl y HF) se liberan en función de
su solubilidad en el magma. Las especies menos solubles se liberan antes
del magma, bajo regímenes de presiones más altas (CO2), mientras que las
especies más solubles se liberan después, bajo regímenes de presiones más
bajas (SO2, HCl y HF). En términos espaciales, CO2 se libera a lo largo de
una área espacial más amplia (desde lo más profundo). La presencia de
SO2 es una indicación clara de un magma que sube hacia un ambiente más
somero. La llegada de HCl en la superficie generalmente indica la
presencia de una remesa de magma somera (cientos de metros hasta pocos
kilómetros). Especialmente un aumento en la proporción CO2/SO2 es
indicativo para elucidar un estado de “inquietud” (“unrest”). Una
disminución consecutiva en el CO2/SO2, después de un aumento, es una
indicación de que el magma está cerca de la superficie y es propenso a una
erupción. Cuando estos gases alcanzan el agua durante su ascenso hacia la
superficie, serán más, o menos, absorbidos. Dependiendo de los equilib-
rios químicos dentro el sistema magmático-hidrotermal (e.g. conversión
de SO2–H2S, acidez), el gas que al final llega a la superficie lleva consigo
la historia de su ascenso desde el fondo hasta la superficie. La
desgasificación magmática es un proceso más rápido, mientras que la
dinámica hidrotermal en el sistema rocoso FeO–FeO1.5 es más lenta. Por
eso, el H2S se suele llamar un “gas hidrotermal”, y el SO2 un “gas
magmático”. Trazar “unrest volcánico” implica un encuadramiento de
tiempo más especifico. Si la ventana de tiempo de monitoreo es más largo
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 223

que el tiempo definido por la cinética de la migración de gas, los detalles

de la dinámica de desgasificación se perderán inevitablemente. Contrari-
amente, si la ventana de tiempo de monitoreo es más corto que la ventana
definida por la cinética de la migración de gas, en tal caso, resultará en una
visión demasiado detallada de lo que el sistema magmático-hidrotermal
puede proveer. Estudios recientes han demostrado que acuíferos
extremadamente ácidos pueden “desacelerar” la señal dejada por el
sistema magmático-hidrotermal dominado por el gas (e.g. fumarolas), pero
pueden “acelerar” la señal dejada por el sistema magmático-hidrotermal
dominado por agua (e.g. lagos cratéricos ácidos). Estos hallazgos tienen
implicaciones significativas para el encuadramiento de tiempo en recono-
cer la desgasificación magmática, y por tanto, para la frecuencia del
monitoreo.

Keywords



Fluid geochemistry Magmatic-hydrothermal systems Volcanic unrest

Volcano monitoring

Palabras clave


Geoquímica de fluidos Sistemas magmático-hidrotermales
Inquietud


volcánica Monitoreo volcánico

Understanding how gas absorbs and solutes react

Introduction
in the liquid phase, and how it is eventually
released to the atmosphere, is key to linking
Fluid geochemical monitoring tracks variations
surface manifestations to magma dynamics.
in gaseous species and fluid phases released in
Volcano monitoring largely focuses on how,
various manners from a volcano. Fluids infiltrate,
where and when magma migrates towards the
move, migrate, rise, react and re-equilibrate in
surface, as, intrinsically, every magmatic erup-
the water- and vapor-filled solid body of the
tion is anticipated by magma rise. Nevertheless,
volcano. These processes are invisible as they
volcanoes can become hazardous even without
occur in the subsurface, and can only be deduced
an eruption. A volcano is in a state of magmatic
from measurements at the surface. Nevertheless,
unrest if we recognize signals of a
gas and water are more mobile than rock and,
magma-on-the-move; in any other situation,
when a volcano shifts into the gear of unrest, a
given volcanic unrest, the volcano is in a state of
change in degassing is often the first sign to be
non-magmatic unrest (hydrothermal or tectonic)
detected. As such, fluid geochemistry offers a
(Rouwet et al. 2014a). To recognize volcanic
crucial means to recognize unrest in a timely
unrest, the background behavior of a volcanic
matter.
system should be tracked for a sufficiently long
Most volcanic edifices store a large volume of
period, in order to know when a deviation from
water, as cold or thermal aquifers, with various
this background becomes a cause for concern
hydrogeological architectures. When gas hits
(i.e. unrest, Phillipson et al. 2013). This back-
water in its rise towards the surface, the original
ground behavior and deviations from it are
signature of the gas is mostly lost. This disad-
volcano-specific and can be monitored in several
vantage can however be overcome.
224 D. Rouwet et al.

ways, besides the geochemical approach origin of the water is generally meteoric (i.e.
reviewed here. rain, snow and its melt water). How hot, and
The aim of this chapter is to scan through how gas-rich such magmatic-hydrothermal sys-
magmatic-hydrothermal systems during the pro- tems are depend on the proportion of the water
cess of magmatic degassing, from bottom to top, volume with respect to the heat and gas pro-
and describe how fluids behave in time and space. vided by the magma. The latter depends on the
What are the lessons learned from fluid geo- residing depth of the magma. Boiling-
chemistry throughout the evolution of volcanic temperature fumaroles are a common manifes-
quiescence, re-awakening, volcanic unrest, mag- tation at magmatic-hydrothermal systems
matic unrest and non-magmatic/hydrothermal (Fig. 1c). When the water table of such systems
unrest? intersects the surface, in craters or volcano
flanks, boiling pools appear (Fig. 1d). Such
pools can manifest bubbling degassing, and are
Magmatic-Hydrothermal nothing less than a water-rich fumarole (Fig. 1
Manifestations e). Depending on the dominant gas they exhale,
paired with water vapor, the manifestations are
One of the first signs of re-awakening after pro- called solfataras (S-rich gases) or mofettes (CO2-
longed volcanic quiescence to a state that even- rich gases); depending on the temperature and
tually causes concern is often the appearance of vapor/water proportion they emit they are called
fumarolic exhalations from a crater. This hap- fumaroles (boiling or above boiling steam
pened, for instance, in 1994 during the reawak- vents), thermal springs (liquid water emission)
ening of Popocatépetl (Mexico), and recently, at or geysers (water + vapor jets with a cyclic
Cotopaxi (Ecuador), both VUELCO target vol- behavior). Thermal springs can discharge inside
canoes (De la Cruz-Reyna and Tilling 2008; Hall active craters or on volcano flanks in a degassed
and Mothes 2008) (Fig. 1a). For open-conduit state, without bubbling or boiling (Fig. 1f).
volcanoes the presence of a plume (i.e. a visible Heated water can fill (parts of) craters and form
gas-vapor cloud originating from an open vol- volcanic lakes (Fig. 1g). Depending on the
cano crater) can become the prominent mani- degassing state and depth of the underlying
festation of degassing. magma, degassing features (bubbling or diffuse
Some volcanoes are characterized by decade degassing) and evaporation can occur at the lake
to century long high-temperature fumarolic surface (Fig. 1h).
degassing in a closed-conduit setting (Fig. 1b), The pictures of the degassing manifestations
suggesting the presence of a stable, but shallow in Fig. 1 show a trend from gas-dominated,
magma chamber. This constantly high- active plume degassing in an open-conduit set-
temperature background degassing is generally ting towards more water-dominated, hydrother-
no sign of unrest (e.g. Momotombo’s >700 °C, mal, passive degassing in a closed-conduit
Satsuma-Iwojima’s >900 °C and Kudryavy’s setting. These visual observations only give a
>700 °C; Menyailov et al. 1986; Shinohara et al. first glance of magmatic-hydrothermal activity,
1993; Taran et al. 1995), whereas the increase of and do not reflect the state of unrest of a volcano.
fumarolic temperatures from low (boiling point Throughout the life-time of magmatic-
of water at a given altitude, hence atmospheric hydrothermal systems (centuries to millenia)
pressure) to high (above boiling to magmatic volcanoes can evolve from gas-dominated to
temperature) can be a sign of resumed unrest water-dominated, and vice versa. The next sec-
(e.g. the 1980–1990s crisis at Vulcano, Italy, tions present what we know on the theoretical
Capasso et al. 1997). level, following the laws of chemistry. Despite
Magmatic-hydrothermal systems are aquifers these classic rules, it will become clear that the
inside a volcano or beneath a volcanic area, range of manifestations and variations in fluid
heated by a magma, at an unspecified depth. The signatures is wide, and volcano-dependent.
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 225

Fig. 1 Degassing manifestations at magmatic- thermal spring at the SE flank of El Chichón. f Thermal
hydrothermal systems. a Open-conduit degassing at flank spring at El Chichón. g The El Chichón crater lake.
Popocatépetl. b High-temperature fumarolic degassing at h Evaporative degassing from Laguna Caliente, Poás. The
Vulcano. c Boiling-temperature fumarolic degassing at cyan arrow points towards water-dominance; the yellow
the Arbol Quemado fracture in the Turrialba crater area. arrow towards more gas-dominance
d Boiling pools inside the El Chichón crater. e Bubbling

magma, the rising new melt and the surrounding

Magma Degassing from Bottom rocks. The second degassing process can be
to Top triggered by the mass removal from part of the
volcanic edifice. This superficial process can
Magma Degassing even trigger eruptions (e.g. 1980 Mt St. Helens).
Internal magma convection causes degassing of a
The degassing of a magma increases when the less dense, gas-rich magma batch. Once degas-
confining pressure in the magma decreases. sed, the now denser magma batch sinks (e.g.
Magma decompression can occur in four ways Stromboli, Aiuppa et al. 2009). The fourth pro-
(Fig. 2): (1) magma rise, (2) decompression by cess increases the gas/melt ratio in the magma,
uncovering a portion of the volcanic edifice, due to the loss of crystals from the cooling
(3) internal convection in a magma chamber, or magma, hence favoring degassing.
(4) secondary boiling upon cooling and conse- Based on gas geochemistry only it is hard to
quent crystallization. The first process is often rule out which process actually occurs at depth,
induced by the input of a deeper magma, into a and thus distinguish between magmatic and
shallower magma chamber. Magma rise towards non-magmatic unrest. Geophysical signals are
the lower pressure regime results from the needed. Following the definition of magmatic
buoyancy difference between the stagnant unrest by Rouwet et al. (2014a) (i.e. the
226 D. Rouwet et al.

recognition of a magma-on-the-move), only the these species gives qualitative insights into the
first process is initially consistent with the req- degassing state of a magma. A consecutive
uisite of magmatic unrest. increase with time in first CO2/SO2, then SO2/
Variations in magma degassing can be HCl, then HCl/HF ratios reflects the evolution in
detected both qualitatively and quantitatively. degassing state from a magma moving from
Detailed insights in the degassing state of a depth towards the surface. Especially an increase
magma can only be obtained if both are mea- in the CO2/SO2 is indicative of the state of
sured contemporaneously. Which gas species are unrest, pointing to an input of poorly degassed
released, how much and when? When a magma magma at great depths. A consecutive decrease
starts to degas, by any of the above processes, the in CO2/SO2 ratio, after the increase, is an indi-
less soluble species is released first (i.e. at higher cation of magma moving towards the surface.
confining pressure in the magma chamber). The The latter two ratios come into play when erup-
order in solubility of indicative magmatic gas tion of magma is imminent, or even ongoing: the
species is CO2 < SO2 < HCl < HF; the order of highly soluble species HCl and HF are released
release when a magma progressively degasses is from a highly degassed magma, a situation that
“CO2-first till HF-last” (Giggenbach 1987). reflects near-surface degassing (Aiuppa et al.
Hence, tracking variations in ratios between 2002). The arrival of HCl at the surface (e.g. in

Fig. 2 Sketch of the four

mechanisms that instigate
magma degassing (not to
scale)
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 227

fumaroles) generally indicates the presence of a At high-temperature magmatic conditions the

shallow magma batch (hundreds of meters, or following reaction applies (Giggenbach 1987;
less). Delmelle and Bernard 2015):
The “purest” proxy of the magmatic gas is
provided by direct sampling of a fumarole with a 3SO2 þ 7H2 ¼ H2 S þ 2S þ 6H2 O ð1Þ
near-magmatic temperature, followed by the
analysis of its chemical composition. During the This explains the presence of both SO2 and
past five-six decades fumarolic gases have been H2S in high-temperature fumaroles. The SO2/
extensively sampled and analyzed representing a H2S ratio is sensitive to temperature and the
wide range of temperatures and states of volcanic oxidation state (i.e. the role of H2 in Eq. 1
activity. Table 1 presents a compilation of reflects the oxidation vs reduction state,
chemical compositions of fumarolic gases for the Giggenbach 1987).
high- and low-temperature ranges. On the other hand, it is noted that the
Before dealing with the abundant water vapor low-temperature fumaroles lack SO2, instead, the
in fumaroles, we tackle the other gas species in dominant sulphur species is H2S (Table 1).
the “dry-gas phase”. Regardless of the fumarole Low-temperature hydrothermal conditions
temperature, CO2 is the most abundant gas spe- (T < 300 °C) favor the reduced S-species H2S
cies (Table 1). CO2 is often released at the sur- (Table 1), equilibrated by the rock phase, fol-
face across a wider spatial extent, as it degasses lowing the reaction:
from greater depth (Fig. 3). Old and deep magma
SO2 þ 6ðFeOÞrock þ H2 O ¼ H2 S þ 6ðFeO1:5 Þrock
bodies can continue to release CO2 for tens to
hundreds of millenia. ð2Þ
The second most abundant “dry-gas” species
are the sulphur species SO2 and H2S. SO2 is or simply through reduction of SO2 in the gas-
more soluble in magma than CO2, and will thus eous environment:
be released at lower pressure. As SO2 degasses at
SO2 þ H2 ¼ H2 S þ O2 ð3Þ
lower depth, the degassing tends to be more
centralized along the central conduit (open or To convert SO2 into H2S in a rock matrix
closed) of the volcano (Fig. 3). As such, SO2 is (FeO–FeO1.5-system), following reaction (2), a
often measured in volcanic plumes. An increase major constraint is “sufficient time”. Magmatic
of SO2 has often been interpreted as an indicator degassing (Eq. 1) is a faster process, while
of magma rise into the shallower environment. hydrothermal dynamics (Eq. 2) are slower. This

Table 1 Chemical composition of fumarolic gases (concentrations in micromol/mol), for high (magmatic, 280°–
1130 °C) and low (hydrothermal, 83°–160 °C) temperature conditions, expressed as the minimum and maximum
measured concentrations for 12 and 59 samples, respectively
T(° H2O CO2 SO2 H2S HCl HF H2 N2 CH4 CO
C)
High-T 280 311000 1200 320 4 275 21 3 40 0.1 0.2
minimum (#12)
High-T 1130 993000 672000 67800 21460 14200 2500 14900 1800 7.1 4600
maximum
(#12)
Low-T 83 638200 2655 0 0 0 0 16 32 0.4 0.011
minimum (#59)
Low-T 160 997200 355000 0 3700 0 0 1220 6800 5330 1.6
maximum
(#59)
228 D. Rouwet et al.

Fig. 3 Sketch of the passing “gas train” during magma degassing, resulting from the difference in solubility of the
various gas species (not to scale)

is why H2S is often called a “hydrothermal gas” at the Earth’s surface. As described before, a
and SO2 a “magmatic gas” (Table 1). magma degasses and heats the space between the
As noted in Table 1, the major gas species in magma and the surface, and will inevitably heat
fumaroles is often water, regardless of their and modify the volcanic aquifers. As the magma
temperature. Even high-temperature fumaroles heats the overlying aquifers, between the magma
are water-dominated. and the surface, from bottom to top, a gas-only,
Knowing that water boils at 100 °C (at vapor + gas zone, a vapor + liquid zone and a
atmospheric pressure, at sea level), in theory, the liquid-only zone can be found (Fig. 4). When a
temperature of a fumarole is buffered at 100 °C magma rises, or the heat input from the magma
and cannot rise until water is exhausted in the increases, the vapor + liquid zone or vapor + gas
underlying plumbing system. Moreover, the zone will be pushed upwards until intersecting
critical temperature of water is 374 °C (i.e. the the surface (e.g. in a volcano crater), manifested
temperature at which vapor and liquid water at the surface as boiling or bubbling pools and
cannot coexist anymore), imposing a second fumaroles (Fig. 4). If the distance between the
temperature buffer. This implies that (1) water magma and surface is larger, the thermal aquifer
directly originates from a high-temperature will intersect the surface and create thermal
magma under super-critical conditions (“an- springs (Fig. 4). Facing unrest, a liquid to vapor
desitic water” with T > 374 °C, Taran et al. transition at a surface manifestation reflects
1989), and/or (2) water is excessively present in heating of the hydrothermal system.
the fumarole plumbing system with respect to the When the above “gas train” (Fig. 3) consec-
gas, and will hardly ever exhaust. utively reaches the liquid-only zone, gas species
will be absorbed and react depending on their
specific chemical properties in water. The
When the Gas Hits the Water capacity to absorb magmatic gases in the liquid
phase is called “scrubbing” (Symonds et al.
The shallow subsurface environment of the dif- 2001). The CO2 that reaches the water from
ferent sections of the Earth’s crust hosts numer- greater depths during magma degassing (Fig. 3),
ous aquifers at various depths originating from will create CO2-dominated bubbling thermal
the infiltration and storage of meteoric water, or springs, and HCO3-rich slightly acidic springs
seawater in the case of low-lying islands; vol- (pH 5–7) (Fig. 4). The second least soluble gas
canic edifices are no different, being “small dots” species that hits the water is SO2 that will be
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 229

Fig. 4 Sketch of the gas that hits the water in an magma, dark grey). Red dotted lines are isotherms.
hypothetical “wet volcano” (not to scale). L = liquid only Cl-geothermal water are deep remnant waters, not of
zone (turquoise area), V + L = vapor + liquid zone, interest for geochemical monitoring of volcanoes for
(yellow area) V + G = vapor + gas zone (grey area), being “old and slow”
G = gas only zone (white area around the degassing

hydrolyzed as sulphuric acid (H2SO4) and dis- H2S is given in return (reaction 4 and 5,
sociate into HSO4− or SO42−, depending on the respectively).
pH of the water. The dissolved H+ creates the As SO2 degasses at shallower depth in a
high acidity or low pH, following pH = −log magma system than CO2, the resulting acid SO4-
[H+]. Volcanic environments are renowned for rich springs are found near the central degassing
being acidic; the acid is generated by the scrub- conduit (e.g. inside active craters). For SO2-
bing of acidic gases into the water. dominated magmatic systems a pH near 2 or less
The following SO2 disproportionation reac- is common; for H2S-dominated hydrothermal
tions occur (Kusakabe et al. 2000): systems a pH of 2–2.5 is the most acidic water
can get (i.e. hydrothermal or “steam heated”
3SO2 þ 3H2 O ¼ 2HSO4 þ S þ 2H þ ð4Þ waters).
The next gas species to be added to the “liquid
or cocktail” are HCl and HF (Fig. 3). HCl is highly
volatile, but also highly hydrophile. This con-
4SO2 þ 4H2 O ¼ 3HSO4 þ H2 S þ 3H þ ð5Þ tradictio in terminis means that when HClgas
reaches water it will be trapped in the liquid
Reaction (4) occurs under relatively oxidizing
phase as Cl− and H+. Generally, Cl− is consid-
conditions, low temperature and high total sul-
ered “conservative” in the liquid phase, and is
phur concentrations, whereas reaction (5) occurs
therefore often used as a tracer in the
under relatively reducing (i.e. oxygen poor)
hydrothermal environment (see Section “Rock
conditions, high temperature and low total sul-
Leaching Upon Weathering”). Conservative
phur concentrations. “To disproportionate SO2”
means that Cl− can be hardly lost from the
means that for each three or four moles of SO2
solution as Cl-salts are highly soluble in acidic
that hit the water, one mole of S°, or one mole of
230 D. Rouwet et al.

and hot environments, and HCl should not degas magmatic-hydrothermal plumbing system? What
from “high pH water” (>2). Sulphate minerals in is the time delay between the moment the gas hits
their turn can be lost from solutions by precipi- the water and the eventual release at the surface?
tation, demonstrating their non-conservative The kinetics (i.e. “speed”) of the gas migration
character (Rouwet and Ohba 2015). For extre- from a magma towards the surface are still poorly
mely acidic environments (pH 0 or <0) the constrained. If we can estimate the residence time
reaction of gas and water in the magmatic-hydrothermal
system, we are able to define a monitoring time
HCl ¼ H þ þ Cl ð6Þ window, and hence adopt an adequate monitor-
ing frequency. As explained, the acidity of the
moves to the left due to an H+ excess with feeding aquifer plays a role. For extreme acidic
respect to Cl− (i.e. HCl degassing). The same conditions, less gas scrubbing occurs and the
counts for HF. Bearing in mind the abundance of fumarolic system will react faster, and hence,
the acid SO2 in the magmatic-hydrothermal shorter monitoring time windows can be adop-
environment, providing a large part of the acid- ted. If the monitoring time window is longer than
ity, this “secondary HCl degassing from the liq- the window defined by the kinetics of gas
uid phase” is less unexpected than previously migration, details in degassing dynamics will be
thought. This implies that HCl can degas from a lost. On the contrary, the monitoring time win-
low-temperature aquifer, as long as the aquifer is dow should not be a lot shorter than the time
extremely acid (Capaccioni et al. 2016). More- window defined by the kinetics of gas migration,
over, as hot water releases vapor, this loss enri- if so, it will provide a too detailed view of what
ches the remnant liquid in solutes, including H+ the magmatic-hydrothermal system can maxi-
(i.e. salinity and acidity increase), leading to the mally provide.
fact that even SO2 tends to degas from the liquid,
instead of remaining in the water phase as SO42−
or HSO4−. The Other Liquid: Elemental Sulphur
Acid water (pH < 3.8) is completely trans-
parent for the omni-present CO2. This implies Whereas water melts at 0 °C and boils at 100 °C,
that CO2, one of the “deepest signals” available sulphur melts at 119 °C and boils at 444 °C
for a degassing volcano, will completely outgas (Oppenheimer and Stevenson 1989, and refer-
from bottom to top. In active craters, often ences therein). This physical fact on phase tran-
underlain by acidic thermal aquifers, the release sitions implies that in the hydrothermal
of CO2 thus behaves as though there is no water environment (boiling water) elemental sulphur is
present. This is a great advantage to monitoring solid, and that during the initial phase of transi-
volcanoes and tracking unrest, especially to tion towards a more magmatic, high-temperature
detect the onset of unrest. regime sulphur will become liquid. Molten sul-
In conclusion, the above insights demytholo- phur in the hydrothermal plumbing system can
gize two generally accepted facts: (1) high- be remobilized, clear vugs and vents and even-
temperature fumaroles cannot contain water tually be expelled as a liquid sulphur flow from
vapor (Table 1), and (2) low-temperature fumarole mouths, or “fill-and-freeze” pores in the
fumaroles cannot release acidic gases. shallower hydrothermal system. The first process
Within the scope of this book, tracking unrest opens up degassing pathways; the second pro-
using fluid geochemistry requires the introduc- cess can decrease rock porosity and permeability
tion of a time frame, or a monitoring time win- near the surface, thus sealing a hydrothermal
dow and frequency. Does the fumarole reflect the system. During the evolution from low-
exact moment of degassing, or is it rather an temperature (>119 °C, unrest) towards high-
average degassing for the longer previous temperature (occasionally magmatic unrest), the
period stored and steadily released from the viscosity of the liquid sulphur increases
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 231

2000-fold (>160 °C) to become an extremely Arbol Quemado fumaroles (González et al.
efficient sealer of a magmatic-hydrothermal sys- 2015). A third phreatic eruption episode involved
tem. Pressure build-up beneath seals in the 2010 and 2012 vents simultaneously (21 May
hydrothermal systems can lead to phreatic erup- 2013). During this 3.5-year long phreatic cycle
tions (Rouwet and Morrissey 2015; Rouwet et al. the SO2 flux from Turrialba’s plume was high:
2016). Monitoring fumarolic temperatures is thus from 2500 to 4300 t/d (Campion et al. 2012;
essential, and probably one the simplest methods Moussallam et al. 2014). CO2/SO2 molar ratios
to apply. in March 2013 were relatively low (2.6), hinting
at a CO2-depleted and SO2-rich magma (Mous-
sallam et al. 2014). From 2001 to 2013, Mous-
Tracking Hydrothermal Unrest sallam et al. (2014) suggest the progressive
and Related Hazards: Methods “drying-out” of the underlying hydrothermal
from Case-Studies system.
The first magmatic eruption at Turrialba since
From Quiescence to Unrest, to Phreatic the 1864–1866 phreatomagmatic activity occur-
Eruptions, to Magmatic Eruptions red during the night of 29 and 30 October 2014
(Mora-Amador et al. 2015). At the time of
Turrialba, Costa Rica (2001–2016) writing, the last magmatic eruptions took place in
A transition from the stable, passively degassing September 2016.
hydrothermal system of Turrialba volcano (Costa Despite the well-monitored and tracked evo-
Rica), to hydrothermal unrest, to phreatic erup- lution from volcanic quiescence to magmatic
tions, to magmatic eruptions, is a recent example eruption it remains unclear why some volcanoes
of an evolution from volcanic quiescence head- quickly evolve from quiescence to eruption,
ing towards eruption. while at Turrialba it took 14 years from quies-
In 2001, increased fumarolic activity (ap- cence to magmatic eruption, passing the com-
pearance of SO2 in late 2001) was paired to plete range of unrest manifestations during this
seismic swarms and ground deformation (Martini relatively long time span.
et al. 2010; Vaselli et al. 2009) (i.e. unrest). In
2007, the increased SO2/H2S molar ratio in Cotopaxi, Ecuador (2015–2016)
fumaroles (>100), combined with an increase in Another example of volcanic quiescence to
exhalation temperature up to 282 °C (in early magmatic unrest is the one of Cotopaxi volcano
2008, Martini et al. 2010; Vaselli et al. 2009), in 2015. As Cotopaxi is a dangerous volcano
point to more oxidized and magmatic conditions whose activity would potentially affect densely
(i.e. magmatic unrest). Clear plume degassing populated areas its monitoring network has been
resumed in early 2007 (Fig. 5a), and SO2 fluxes continuously improved since the late 70s. After
reached 740 t/d in January 2008 (Martini et al. 73 years of quiescence, the first sign of unrest at
2010). In late 2009, fumarolic degassing was Cotopaxi was a progressive increase in the
vigorous and extended into the Arbol Quemado amplitude of transient seismic events in April
fracture, newly formed in 2002. The first phreatic 2015. SO2 is measured at Cotopaxi by DOAS
eruption occurred on 5 January 2010, in the inner stations installed on the flanks of the volcano
crater wall of the actively degassing SW crater. since 2008. The permanently measured SO2
Strong jet-like degassing occurred afterwards emissions showed an increase on May 20 from
from this new vent (Fig. 5b), while diffuse almost non detectable up to *3000 t/d. The
fumarolic degassing diminished in the SW crater. fumaroles showed increased activity and a gas
A second phreatic eruption occurred on 12 Jan- plume from the crater was usually observed on
uary 2012, from a vent inside the Arbol Que- clear days. By early June SO2 emissions yielded
mado fracture. The day before, this eruption was up to 5000 t/d. On July 20 a green lake was
preceded by liquid sulphur flowing out of the observed filling the crater of the volcano,
232 D. Rouwet et al.

Fig. 5 Hydrothermal unrest at Turrialba, Poás and Irazú due to plume degassing and acid rain fall downwind.
volcanoes. a Vegetation die-back due to resumed plume d Efficient rock mass removal through Río Sucio, 30 km
degassing. b The 2010 phreatic eruption vent (picture by downstream Irazú volcano
S. Calabrese). c Lack of vegetation at Cerro Pelón (Poás)

nevertheless no significant changes in the SO2 24,000 t/d and decreased progressively until the
emission was observed, suggesting that the lake end of the activity in late November 2015.
was of high acidity and/or too small to be an Since June and more consistently since 14
efficient scrubber. The first phreato-magmatic August 2015, BrO was also detected in the plume
explosions occurred on August 14 and produced (Dinger et al. 2016). Airborne Multi-GAS mea-
ash and gas columns reaching up to 9 km above surements showed that the plume had a low CO2/
the crater. The satellite-borne instruments such as SO2 ratio, and that SO2 was >99% of total sulfur
OMI and OMPS reported 16,400 and 12,500 t/d (SO2 + H2S), indicating a shallow magmatic
of SO2 released to the atmosphere on that day origin for the gas. At the time of writing
(http://so2.gsfc.nasa.gov/pix/daily/0815/ecuador_ (September 2016), SO2 emissions decreased to
0815z.html). Continuous ash emissions followed background levels. SO2 permanent monitoring
the initial explosions producing a permanent gas proved to be a useful tool at Cotopaxi providing
and ash plume drifting westward. SO2 measured real time data contributing, together with other
in this permanent plume, by mobile-DOAS tra- geophysical methods, to better evaluate volcanic
verses or by the permanent stations, reached unrest scenarios.
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 233

Gas Impact and Acid Rain acidic magmatic-hydrothermal systems

(Ingebritsen et al. 2001; Taran and Peiffer 2009;
Persistent high-temperature fumarolic and plume Chiodini et al. 2014; Collard et al. 2014). Mea-
degassing impact volcano flanks down-wind suring the Cl-release from rivers draining thermal
(Fig. 5c). This can happen when it rains springs, and knowing the Cl-content and
through a plume, generating acid rain. Many Cl/solute ratios in thermal spring waters, the rock
volcanoes are subject to such long-term, mass removal rate can be estimated by:
non-eruptive hazard, making the surrounding
ground harsh living environments. Other volca- Qr  Cr ¼ Qs  Cs ð7Þ
noes suffer vegetation die-back, when activity
resumes after prolonged quiescence (Fig. 5a). where Qr and Qs are the discharges of rivers and
This was clearly visible during the increased thermal springs, respectively, in L/s; Cr and Cs are
activity at Turrialba volcano (Section “Turrialba, the concentrations of Cl of rivers and thermal
Costa Rica (2001–2016)”, Fig. 5a, b; González springs, respectively, in mg/L. Measuring the river
et al. 2015) and downwind Poás’ western flank discharge (Qr) and analyzing the river and spring
(Cerro Pelón, Fig. 5c). To assess such hazards, a waters for its Cl content (Cr and Cs, respectively)
meteorological station (wind direction, speed, air thus enables to estimate the spring discharge (Qs),
humidity and rainfall), and a DOAS device to which would otherwise be impossible to directly
measure SO2 fluxes are valuable tools. measure in the field (e.g. numerous spring dis-
charges, too irregular spring mouths). This method
combines gas-water-rock interaction and hydrol-
Rock Leaching upon Weathering ogy of magmatic-hydrothermal systems in order to
assess indirect hazard. Volcanoes with high
Absorption of magmatic gases into aquifers cre- rock mass removal rates are e.g. Irazú (Costa
ates acidic magmatic-hydrothermal systems that, Rica, Fig. 5d); extremely acidic magmatic-
sooner or later, will exit the volcano. Acidic hydrothermal rock removers are Kawah Ijen
water attacks the wall rock and becomes loaded (Java), Poás (Costa Rica), and Copahue
with solutes (Delmelle et al. 2015). If (1) mete- (Argentina-Chile). In the most extreme cases, the
oric recharge is high, (2) acid input is high, acidity and toxic metal load affects agricultural
(3) wall rock is fresh, and (4) (thermal) spring activities and human health.
water discharge is high, rock leaching capacity Moreover, through the same Cl-inventory
can reach thresholds to even mechanically approach, the geothermal potential (i.e. heat
destabilize volcano flanks. Enhanced chemical output) from springs can be estimated, by mul-
leaching for long periods favors physical rock tiplying the enthalpy of discharged spring waters,
removal, causing rock fall, landslides, or even often based on geothermometric temperatures of
flank and sector collapses of volcanic edifices. the deep system, with the spring discharge rates.
Even if magmatic degassing seems absent and a Such estimates were obtained for the active
volcano may be long dormant, hazards loom due magmatic-hydrothermal systems of El Chichón
to the scrubbing capacity of deeper aquifers and Tacaná (both in Chiapas, Mexico; Taran and
(Delmelle et al. 2015, and references therein). Peiffer 2009; Collard et al. 2014), and Domuyo
The best suited method to quantify and track (Argentina; Chiodini et al. 2014), and originally
rock mass removal from a volcanic edifice is by of the Cascades Volcanic Range by Ingebritsen
monitoring the discharge rates from thermal et al. (2001). Understanding the state of unrest on
springs, and their dissolved solutes and solids. the long term of a specific volcano is needed to
The “Cl-inventory” uses Cl as the conservative rule out if the volcano would be a feasible target
tracer. As mentioned earlier, this is true for less for geothermal exploitation, or not.
234 D. Rouwet et al.

Volcanic Lakes plumes, Multi-GAS to measure ratios between

gas species, e.g. CO2/SO2).
Acid Peak-Activity Lakes in a State Considering acidic lakes as “open-air”
of Unrest fumaroles (and fumaroles as “buried” acidic
Volcanic lakes are the intersection of the crater lakes) has changed the monitoring time frame,
surface and the underlying aquifer (Fig. 1g, h). which might lead to better chances to forecast
Hence, they become “windows” into the depths phreatic eruptions (Rouwet et al. 2014a, b;
of magmatic-hydrothermal systems. While Rouwet and Morrissey 2015).
fumaroles directly lose their signal from depth to
the atmosphere as a “snapshot” (but see Sec- Volcanic Lake Response to External
tion “When the Gas Hits the Water”), volcanic Triggers in the Absence of Magmatic
lakes preserve a past gas marker for a certain Unrest
period. The time we can track back by studying The Boiling Lake, in Dominica of the Lesser
the water chemistry depends on the duration the Antilles (a VUELCO target volcano), is a
water resides in the lake (i.e. residence time). The high-temperature volcanic crater lake, that is
residence time (RT) is estimated by dividing the believed to be formed as a result of phreatic or
lake volume (V in m3) by the input or output rate phreatomagmatic activity (Lindsay et al. 2005). It
(Q in m3/s) of fluids, assuming steady-state is approximately 50 m  60 m in size and ca.
conditions (Varekamp 2003; Rouwet et al. 12–15 m deep, with an estimated volume of
2014b): *1.2  104 m3 when filled (Fournier et al.
2009). Over the last 150 years, temperatures
RT ¼ V=Q ð8Þ taken at the edge of the lake have generally
ranged between 80 and 90 °C, and the pH from 4
Small lakes with high rates of fluid flushing to 6 (Joseph et al. 2011), however, the lake
offer a better time-resolution than large lakes experiences periods of instability where water
with slow rates of fluid in- and output. The level, temperature and state of hydrothermal
monitoring frequency (e.g. lake water sampling) activity fluctuate suddenly. The long term sta-
should be tuned to this residence time; a higher bility of the Boiling Lake is attributed to the
monitoring frequency will oversample the lake hypothesis that the lake is suspended above the
chemistry, while a lower monitoring frequency water table by the buoyancy of rising steam
will lead to the loss of information the lake bubbles from the underlying hydrothermal sys-
potentially provides (Rouwet et al. 2014b). tem (Fournier et al. 2009). Perturbations of this
Volcanic lakes are excellent gas scrubbers. interaction as a result of landslides into the lake
Nevertheless, recent studies quantify the gas or regional seismicity is attributed to the cause of
release from the lake surface of the most acidic the sudden periods of instability (i.e. lake drai-
lakes (e.g. Aso, Copahue, Poás, Kawah Ijen; nage and refill cycles), rather than changes in
Shinohara et al. 2015; Tamburello et al. 2015; underlying magmatic activity. This is mainly
Capaccioni et al. 2016; de Moor et al. 2016; because no coincident anomalous hydrothermal
Gunawan et al. 2016). Under the most extreme activity was observed in the Valley of Desolation
pH conditions (<0) HF, HCl and even SO2 can geothermal area, which is located nearby the
degas freely from the lake. This means that acidic Boiling Lake.
lakes are more sensitive than thought before, as The most recent episode of instability occur-
acid gas flashes through the water body with only red during December 2004 to April 2005
minor scrubbing. Monitoring frequency can thus (Fig. 6), when the lake water level dropped by
increase and, hence, spectroscopic or electro- *8–10 m, and the temperature at the water’s
chemical sensor tools become extremely useful edge decreased to <30 °C. Water acidity went
(DOAS to measure SO2 fluxes from volcanic from the usual acidic pH of 4–6 to neutral, while
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 235

Fig. 6 The refill of an empty

Boiling Lake during a period
of unusual activity where
water levels and geothermal
activity were rapidly
fluctuating. a An empty
Boiling Lake on 7 April 2005.
b A full Boiling Lake on 13
April 2005. Pictures by
Arlington James, Forestry
Officer, Forestry and Wildlife
Division, Dominica (used
with permission)

Cl concentration dropped from the typical 2000– returned to its normal steady-state of activity
6000 to 29–50 mg/L, and SO4 concentration (Joseph et al. 2011). This episode is reported to
dropped from 1500–4000 to 100–270 mg/L, have been triggered as a result of the extensional
indicating a drastic decrease of hydrothermal strain induced by a regional Mw 6.3 earthquake
fluid input into the lake (Joseph et al. 2011). that occurred on November 21, 2004 offshore
Additionally, total dissolved solid content Les Saintes (Guadeloupe) (Feuillet et al. 2011),
decreased from 13,400 to 4500 mg/L, suggesting which may have contributed to diminished water
strong dilution by fresh water. Measurements of inflow.
temperature, pH and chemical composition taken It should be noted, however, that a phreatic
in August 2006 indicate that the lake had explosion and gas release occurred at an “empty”
236 D. Rouwet et al.

Table 2 Geochemical signals and what they indicate with respect to volcanic unrest
Type of unrest Geochemical signal Indication
Unrest CO2 flux above background Changes in deep degassing dynamics
Increase in T of hot springs and/or fumaroles Increased heat input
Changes in H2O/CO2 ratios in fumaroles Changes in water/gas ratio
Appearance of new fumaroles and/or hot springs Aerial extension of activity
Magmatic unrest Appearance of acidic gases (SO2, HCl, HF) Changes in mid- to shallow magma
dynamics
T fumarole >119 °C Remobilisation of sulphur
SO2 flux > X t/d SO2 flux above background,
volcano-dependent
Increase in CO2/SO2 ratio Arrival of an undegassed magma at depth
Extreme increase in T fumaroles (>300 °C) Towards magmatic T
Magmatic Decreasing CO2/SO2 ratios after increase More superficial magma degassing
eruption Increase in Cl, Br, F concentrations in hot Input of highly soluble acidic gases
springs/pools
Decrease in H2O/CO2 and/or H2S/SO2 and/or More gas with a more magmatic signature
SO2/HCl ratios
Hydrothermal New fumaroles Aerial extension of activity
unrest Anomalous glacier defrosting Sudden removal of water mass… lahars
Water to vapour transition Pushing vapour front from below
Changes in hydrothermal features Variations or aerial extension of activity
Increase in B and/or NH4 in waters Input of vapour
Increase in CH4/CO2 in fumaroles A more hydrothermal signature in
fumaroles
Variations in phreatic level in aquifers Pushing vapour front from below
Hydrothermal 120 °C < T fumarole <200 °C Self-sealing by a change in S viscosity
eruption Extension of alteration areas or fumarolic fields Aerial extension of activity
Appearance of muddy pools Clearing bugs and vents, unplugging
Boiling/bubbling of pools that previously didn’t Rising vapour front and/or extra heating
and degassing

Boiling Lake on 10 December 1901 that resulted the subsurface environment. Over the last 10–
in the deaths of two individuals (Elliot 1938; Bell 15 years more attention has been paid to remote
1946). This suggests that hazards related to vol- sensing of volcanic gas plumes (DOAS,
canic lakes such as the Boiling Lake, may occur Multi-GAS) with the obvious advantage of
without magmatic input. increased safety and frequency of data gathering.
Nevertheless, the relationship between the
fumarole and plume has yet to be better con-
Take-Home Ideas: Implications strained. The best proxy of a magmatic gas
for Geochemical Monitoring remains a direct sample of a high-temperature
fumarole, although such target fumaroles are
Over the past five-six decades, gas geochemistry often inaccessible, especially during eruptive
at magmatic-hydrothermal systems has mainly phases. Compromises between data fidelity and
focussed on chemical equilibria and kinetics in safety of the operators, and the frequency of data
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 237

gathering should be framed in terms of what we Capaccioni B, Rouwet D, Tassi F (2016) HCl degassing
want and maximally can unravel. The advantage from extremely acidic crater lakes: empirical results
from experimental determinations and implications for
of fluid geochemistry in volcano monitoring geochemical monitoring. In: Caudron C, Capaccioni B,
arrises from the fact that volatiles are mobile and Ohba T (eds) GSL special publications 437 geochem-
thus reach the surface often before physical istry and geophysics of volcanic lakes. doi:10.1144/
changes manifest. Timely recognition of unrest SP437.12
Capasso G, Favara R, Inguaggiato S (1997) Chemical
and especially hydrothermal unrest is often pos- features and isotopic composition of gaseous mani-
sible. Table 2 summarizes geochemical signals festations on Vulcano Island, Aeolian Islands, Italy: an
and how they relate to the various states of vol- interpretative model of fluid circulation. Geochim
canic unrest, useful for monitoring based on Cosmochim Acta 61(16):3425–3440
Chiodini G, Liccioli C, Vaselli O, Calabrese S, Tassi F,
deterministic research and probabilistic Caliro S, Caselli A, Agusto M, D’Alessandro W
modeling. (2014) The Domuyo volcanic system: an enormous
Future research should focus on better con- geothermal resource in Argentine Patagonia. J Vol-
straining degassing dynamics at the canol Geoth Res 272:71–77. doi:10.1016/j.jvolgeores.
2014.02.006
surface-atmosphere boundary. Recent studies Dinger F, Arellano S, Battaglia J, Bobrowski N, Galle B,
have demonstrated that extremely acidic aquifers Hernández S, Hidalgo S, Hörmann C, Lübcke P,
can “slow down” the signal released from Platt U, Ruíz M, Warnach S, Wagner T (2016)
gas-dominated magmatic-hydrothermal systems Variations of the BrO/SO2 molar ratios during the
2015 Cotopaxi eruption. EGU 2016-1001
(fumaroles), but “speed up” the signal released Collard N, Taran Y, Peiffer L, Campion R, Jacome
from water-dominated systems (e.g. acidic crater Paz MP (2014) Solute fluxes and geothermal potential
lakes). These findings have strong implications of Tacana volcano-hydrothermal system,
Mexico-Guatemala. J Volcanol Geoth Res 288:123–
for the time frame of magma degassing, and
131. doi:10.1016/j.jvolgeores.2014.10.012
hence for the monitoring frequency. De la Cruz-Reyna S, Tilling RI (2008) Scientific and
public responses to the ongoing volcanic crisis at
Popocatépetl Volcano, Mexico: importance of an
effective hazards-warning system. J Volcanol Geoth
References Res 170:121–134
Delmelle P, Bernard A (2015) The remarkable chemistry
Aiuppa A, Federico C, Paonita G, Valenza M (2002) S, Cl of sulfur in hyper-acid crater lakes: a scientific tribute
and F degassing as an indicator of volcanic dynamics: to Bokuichiro Takano and Minoru Kusakabe. In:
the 2001 eruption of Mount Etna. Geophys Res Lett Rouwet D, Christenson B, Tassi F, Vandemeule-
29(11):1559. doi:10.1029/2002GL015032 brouck J (eds) Book chapter in volcanic lakes,
Aiuppa A, Federico C, Giudice G, Giuffrida G, Guida R, Springer, Heidelberg, pp 239–260. doi:10.1007/978-
Gurrieri S, Liuzzo M, Moretti R, Papale P (2009) The 3-642-36833-2_10
2007 eruption of Stromboli volcano: insights from Delmelle P, Henley RW, Opfergelt S, Detienne M (2015)
real-time measurements of the volcanic gas plume Summit acid crater lakes and flank instability in
CO2/SO2 ratio. J Volcanol Geoth Res 182(221):230. composite volcanoes. In: Rouwet D, Christenson B,
doi:10.1016/j.jvolgeores.2008.09.013 Tassi F, Vandemeulebrouck J (eds) Book chapter in
Baldoni E, Rouwet D, Mora-Amador R, Ramírez C, volcanic lakes. Springer, Heidelberg, pp 289–306.
González-Ilama G, Lucchi F, Capaccioni B, doi:10.1007/978-3-642-36833-2_12
Tranne CA, Pecoraino G (submitted) Hydrogeochem- de Moor JM, Aiuppa A, Pacheco J, Avard G, Kern C,
ical model of the Irazú-Turrialba volcanic complex Liuzzo M, Martinez M, Giudice G, Fischer TP (2016)
(Costa Rica) and implications for hazard assessment Short-period volcanic gas precursors to phreatic
and volcanic surveillance. In: Caudron C, Capac- eruptions: insights from Poás Volcano, Costa Rica.
cioni B, Ohba T (eds) GSL special volume geochem- Earth Planet Sci Lett 442:218–227. doi:10.1016/j.epsl.
istry and geophysics of volcanic lakes 2016.02.056
Bell H (1946) Glimpses of a governor’s life—Dominica. Elliot EC (1938) Boiling lake—the 1900 story. In: Broken
Samson Low, London Atoms. Unpublished report presented to the Govern-
Campion R, Martínez-Cruz M, Lecocq T, Caudron C, ment of Dominica
Pacheco J, Pianrdi G, Hermans C, Carn S, Bernard A Feuillet N, Beauducel F, Jacques E, Tapponnier P,
(2012) Space- and ground-based measurements of Delouis B, Bazin S, Vallée M, King GCP (2011)
sulphur dioxide emissions from Turrialba Volcano The Mw = 6.3, November 21, 2004, Les Saintes
(Costa Rica). Bull Volcanol. doi:10.1007/s00445-012- earthquake (Guadeloupe): Tectonic setting, slip model
0631-z
238 D. Rouwet et al.

and static stress changes. J Geophys Res 116(B10):1– Moussallam Y, Peters N, Ramirez C, Oppenheimer C,
25. doi:10.1029/2011JB008310 Aiuppa A, Giudice G (2014) Characterisation of the
Fournier N, Withal F, Moreau-Fournier M, Bardou L magmatic signature in gas emissions from Turrialba
(2009) The Boiling Lake of Dominica, West Indies: Volcano, Costa Rica. Solid Earth 5:1341–1350.
high temperature volcanic crater lake dynamics. doi:10.5194/se-5-1341-2014
J Geophys Res 114:B02203. doi:10.1029/ Oppenheimer C, Stevenson D (1989) Liquid sulphur lakes
2008JB005773 at Poás volcano. Nature 342:790–793
Giggenbach WF (1987) Redox processes governing the Phillipson G, Sobradelo R, Gottsmann J (2013) Global
chemistry of fumarolic gas discharges from White volcanic unrest in the 21st century: an analysis of the
Island, New Zealand. Appl Geochem 2:143–161 first decade. J Volcanol Geoth Res 264:183–196
González G, Mora-Amador R, Ramirez C, Rouwet D, Rouwet D, Morrissey M (2015) Mechanisms of crater
Alpizar Y, Picado C, Mora R (2015) Actividad lake eruptions: physical and numerical modeling. In:
historica y analysis de la amenaza del volcano Rouwet D, Christenson B, Tassi F, Vandemeule-
Turrialba, Costa Rica. Rev Geol Am Centr 52:129– brouck J (eds) Book chapter volcanic lakes. Springer,
149. doi:10.15517/rgac.v0i52.19033 Heidelberg, pp 73–92. doi:10.1007/978-3-642-36833-
Gunawan H et al (2016) New insights into Kawah Ijen’s 2_3
volcanic system from the wet volcano workshop Rouwet D, Ohba T (2015) Iotope fractionation and HCl
experiment. In: Caudron C, Capaccioni B, Ohba T partitioning during evaporative degassing from active
(eds) GSL special volume geochemistry and geo- crater lakes. In: Rouwet D, Christenson B, Tassi F,
physics of volcanic lakes Vandemeulebrouck J (eds) Book chapter volcanic
Hall M, Mothes P (2008) The rhyolitic-andesitic eruptive lakes. Springer, Heidelberg, pp 179–200. doi:10.1007/
history of Cotopaxi volcano, Ecuador. Bull Volcanol 978-3-642-36833-2_7
70:675–702. doi:10.1007/s00445-007-0161-2 Rouwet D, Sandri L, Marzocchi W, Gottsmann J, Selva J,
Ingebritsen SE, Galloway DL, Colvard EM, Sorey ML, Tonini R, Papale P (2014a) Recognizing and tracking
Mariner RH (2001) Time variation of hydrothermal hazards related to non-magmatic unrest: a review.
discharge at selected sites in the western United States: J Appl Volcanol 3:17. doi:10.1186/s13617-014-0017-
implications for monitoring. J Volcanol Geoth Res 3
111:1–23 Rouwet D, Tassi F, Mora-Amador R, Sandri L, Chiarini V
Joseph EP, Fournier N, Lindsay J, Fischer T (2011) Gas (2014b) Past, present and future of volcanic lake
and water geochemistry of geothermal systems in monitoring. J Volcanol Geoth Res 272:78–97. doi:10.
Dominica, Lesser Antilles island arc. J Volcanol 1016/j.jvolgeores.2013.12.009
Geoth Res 206(1–2):1–14. doi:10.1016/j.volgeores. Rouwet D, Mora-Amador R, Ramírez-Umaña C, Gonzá-
2011.06.007 lez G, Inguaggiato S (2016) Dynamic fluid recycling
Kusakabe M, Komoda Y, Takano B, Abiko T (2000) at Laguna Caliente (Poás, Costa Rica) before and
Sulfur isotopic effects in the disproportionation reac- during the 2006-ongoing phreatic eruption cycle
tion of sulfur dioxide in hydrothermal fluids: implica- (2005–2010). In: Caudron C, Capaccioni B, Ohba T
tions for the d34S variations of dissolved bisulphate (eds) Geochemistry and geophysics of volcanic lakes.
and elemental sulfur from active crater lakes. J Vol- Geological Society London Special Publications.
canol Geoth Res 97:287–307 doi:10.1144/SP437.11
Lindsay J, Robertson R, Shepherd J, Ali S (2005) Shinohara H, Giggenbach WF, Kazahaya K, Heden-
Volcanic hazard atlas of the Lesser Antilles. In: 279. quist JW (1993) Geochemistry of volcanic gases and
St. Augustine: Seismic Research Unit, University of hot springs of Satsuma-Iwojima, Japan: following
the West Indies Matsuo. Geochem J 27:271–285
Martini F, Tassi F, Vaselli O, Del Potro R, Martínez M, Shinohara H, Yoshikawa S, Miyabuchi Y (2015) Degas-
Van der Laat R, Fernandez E (2010) Geophysical, sing activity of a volcanic crater lake: volcanic plume
geochemical and geodetical signals of reawakening at measurements at the Yudamari crater lake, Aso
Turrialba volcano (Costa Rica) after 150 years of volcano, Japan. In: Rouwet D, Christenson B,
quiescence. J Volcanol Geoth Res 198:416–432. Tassi F, Vandemeulebrouck J (eds) Book chapter
doi:10.1016/j.jvolgeores.2010.09.021 volcanic lakes. Springer, Heidelberg, pp 73–92.
Menyailov IA, Nikitina LP, Shapar VN, Pilipenko VP doi:10.1007/978-3-642-36833-2_8
(1986) Temperature increase and chemical changes of Symonds RB, Gerlach TM, Reed MH (2001) Magmatic
fumarolic gases at Momotombo volcano, Nicaragua, gas scrubbing: implications for volcano monitoring.
in 1982–1985: are these indicators of a possible J Volcanol Geoth Res 108:303–341
eruption? J Geophys Res 187:12199–12214 Tamburello G, Agusto M, Caselli A, Tassi F, Vaselli O,
Mora-Amador R, Ramírez-Umaña C, González G, Calabrese S, Rouwet D, Capaccioni B, Di Napoli R,
Rouwet D, Lucchi F, Forni F, Sulpizio R, Baldoni E, Cardellino C, Chiodini G, Bitetto M, Brusca L,
Alpízar-Segura Y, Tranne CA (2015) First documen- Bellomo S, Aiuppa A (2015) Intense magmatic
tation of the ongoing phreatic-strombolian eruptions degassing through te lake of Copahue volcano,
of Turrialba volcano (Costa Rica). IUGG General 2013–2014. J Geophys Res doi:10.1002/
Assembly, Prague 3746 2015JB012160
Fluid Geochemistry and Volcanic Unrest: Dissolving the Haze … 239

Taran YA, Hedenquist JW, Korzhinsky M, Tkachenko SI, Varekamp JC (2003) Lake contamination models for
Shmulovich KI (1995) Geochemistry of magmatic evolution towards steady state. J Limnol 62(1):67–72
gases from Kudryavy volcano, Iturup, Kuril Islands. Vaselli O, Tassi F, Duarte E, Fernández E, Poreda RJ,
Geochim Cosmochim Acta 59:1749–1761 Delgado Huertas A (2009) Evolution of fluid geo-
Taran YA, Peiffer L (2009) Hydrology, hydrochemistry and chemistry at the Turrialba volcano (Costa Rica) from
geothermal potential of El Chichón volcano- 1998 to 2008. Bull Volcanol. doi:10.1007/s00445-
hydrothermal system, Mexico. Geothermics 38:370–378 009-0332-4
Taran YA, Pokrovsky BG, Dubik YM (1989) Isotopic
composition and origin of water from andesitic
magmas. Dokl Acad Sci 304:440–443

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licensesby/4.0/ unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Geophysical Footprints of Cotopaxi’s
Unrest and Minor Eruptions in 2015:
An Opportunity to Test Scientific
and Community Preparedness

Patricia A. Mothes, Mario C. Ruiz, Edwin G. Viracucha,

Patricio A. Ramón, Stephen Hernández, Silvana Hidalgo,
Benjamin Bernard, Elizabeth H. Gaunt, Paul Jarrín,
Marco A. Yépez and Pedro A. Espín

Abstract
Cotopaxi volcano, Ecuador, experienced notable restlessness in 2015 that
was a major deviation from its normal background activity. Starting in
April and continuing through November 2015 strong seismic activity,
infrasound registry, hikes in SO2 degassing and flank deformation with
small displacements were some of the geophysical anomalies that were
registered. Obvious superficial changes, such as small hydromagmatic
eruptions, emission of vapor and ash columns, thermal hotspots around
the crater and in nearby orifices and exacerbated glacier melting were also
observed. Our contribution provides an overview of the 2015 Cotopaxi
unrest by presenting the patterns of geophysical data and the sequence of
events produced by the volcano. Cotopaxi’s last important VEI 4 eruption
was in 1877. Then it had devastating effects because of the transit of huge
lahars down 3 major drainages. Comparatively, the 2015 activity never
surpassed a magnitude VEI 2 and principally produced limited hydro-
magmatic explosions and semi-continuous low energy emissions and light
ashfalls. Given the potential of major destruction from a large Cotopaxi
eruption it is important to understand the geophysical fingerprints that
characterized the 2015 episode with an eye to identifying onset of future
restless periods. Overall, the monitoring activities, the data interpretation,
formulation of reasonable eruptive scenarios, and finally, the preparation
of a stream of constant information being relayed to concerned authorities
and the public, was a real test of the IGEPN’s capacity to deal with a
complicated eruption situation whose outcome was not apparent at the
beginning, but which concluded in a very small eruptive episode.

P.A. Mothes (&)
M.C. Ruiz
E.G. Viracucha

P.A. Ramón
S. Hernández
S. Hidalgo
B. Bernard

E.H. Gaunt
P. Jarrín
M.A. Yépez
P.A. Espín
Instituto Geofísico, Escuela Politécnica Nacional,
Quito, Ecuador
e-mail: [email protected]

Advs in Volcanology (2019) 241–270

DOI 10.1007/11157_2017_10
© The Author(s) 2017
Published Online: 19 July 2017
242 P.A. Mothes et al.

Resumen
En 2015 el volcán Cotopaxi, Ecuador experimento un notable cambio, que
fue una desviación importante de su actividad normal de base. A partir de
abril y hasta noviembre de 2015 fuerte actividad sísmica, registros de
infrasonido, incremento en la desgasificación de SO2 y pequeños cambios
en la deformación de los flancos fueron algunas de las anomalías
geofísicas registradas. Evidentes cambios superficiales también fueron
observados como pequeñas erupciones hidromagmaticas, emisión de
vapor, columnas de cenizas, puntos calientes alrededor del cráter y el
deshielo de los glaciares. Nuestra contribución proporciona una visión
general de las anomalías del Cotopaxi en el 2015, mediante la
presentación de patrones de los datos geofísicos y la secuencia de eventos
producidos por el volcán. La última erupción importante del Cotopaxi fue
un VEI 4 en 1877. Esta tuvo efectos devastadores debido al descenso de
enormes lahares por sus tres drenajes mayores. Comparativamente, la
actividad del año 2015 nunca superó una magnitud VEI 2, principalmente
produciendo explosiones hidromagmaticas, escasas emisiones y leves
caídas de ceniza. Debido a la potencial destrucción por una eventual
erupción grande del Cotopaxi es importante entender los registros
geofísicos que caracterizó el episodio de 2015 para poder identificar el
inicio de futuros períodos eruptivos. En general, las actividades de
vigilancia, la interpretación de datos, formulación de escenarios eruptivos
razonables y por último, la preparación de un flujo de información
constante que llegue a las autoridades interesadas y el público, fue una
verdadera prueba de la capacidad del IGEPN para hacer frente a una
situación de erupción cuyo resultado no era evidente al principio, pero que
finalizó como una erupción pequeña.

Keywords



Volcanic unrest Precursory geophysical patterns Precursory LPs and
VLPs
Volcano monitoring

Cotopaxi volcano-Ecuador

State of
preparedness

awaken for short term low-level activity, then

Introduction
resume repose until additional magma inputs
herald an episode of further unrest (Phillipson
Long dormant volcanoes that begin to awaken
et al. 2013). Eruptions that barely bring magma
may have start and stop activity that has to be
to the surface may be classified as “failed”, since
evaluated with respect to the volcano’s known
so little juvenile magma erupts (Moran et al.
past. A volcano’s past activity is known by study
2011).
of its stratigraphy, other physical evidence and
At Cotopaxi ample geological and historical
possibly historical chronicles (Tilling 1989).
information exists with regard to its past activity
Many uncertainties preclude knowing the final
(Hall and Mothes 2008; Garrison et al. 2011;
outcome of a restless volcano (Newhall 2000;
Pistolesi et al. 2012). Formulation of eruptive
Sparks and Aspinall 2004), since a volcano may
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 243

scenarios with respect to the 2015 unrest period and an altitude of 5897 m (Hall and Mothes
were based on our collective knowledge of the 2008). Its actual glacier cap of 10.49 km2 is
volcano’s geology and eruptive history and rapidly diminishing due to climatic change
published information as well as interpretation of Cáceres et al. (2016) (Fig. 1). The volcano’s last
the abundant geophysical data streams available important VEI = 4 eruption was on 26 June
through instrumental and observational networks 1877. Then it generated highly erosive pyro-
operated by the Instituto Geofísico of the Escuela clastic flows that melted glaciers and triggered
Politécnica Nacional (IGEPN)-Quito, Ecuador, voluminous lahars (*100 Mm3 per drainage).
the entity in charge of volcano and tectonic Each lahar traveled hundreds of kilometers down
monitoring in Ecuador. The combination of these several drainages enroute to the Amazon basin,
inputs allowed scientists at the IGEPN to trans- Pacific Ocean and to the Atlantic, respectively
mit a coherent image of the evolving unrest (Mothes et al. 2004; Mothes and Vallance 2015).
presented during 2015 and to indicate the most These past lahar routes now host sprawling
likely eruption/activity scenarios. Two earlier suburbia, important economic activities and vital
unrest periods are known: 1975–1976, when the infrastructure. Ecuador’s second most visited
IGEPN had limited seismic equipment operating national park (Parque Nacional Cotopaxi-PNC)
on the volcano and then in 2001–2002. Both is centered on the volcano and draws some
periods were comprised of increased fumarolic 200,000 tourists a year.
activity both inside and outside of the crater and The volcano’s five most important eruptive
a hike in seismicity for the 2001 period (Molina episodes during the historical period (since 1532)
et al. 2008). The 2015 unrest displayed important have been of andesitic composition and ranged
changes in seismic and deformation patterns, gas between VEI = 3 and 4 (Pistolesi et al. 2012).
output and superficial activity, compared to Nonetheless, the volcano is bi-modal and pro-
background, whose level was established since duces VEI = 5 rhyolitic eruptions about every
around 1990. In sum, Cotopaxi’s 2015 unrest 2000 years (Hall and Mothes 2008). The last
displayed a progressive crescendo of geophysical important rhyolitic eruption, the Peñas Blancas
signals, then minor hydromagmatic explosions, event, occurred about 2800 years BP (Mothes
followed by overall seismic energy decrease at et al. 2015a). The youngest andesitic eruptive
the end of 2015, which was accompanied by products contain intergrowths of plagioclase and
fewer superficial manifestations. Like many other pyroxene and four different populations of pla-
volcanoes that have displayed unrest, this recent gioclase crystals which indicate pervasive
episode did not culminate in a full-fledged magma mixing (Garrison et al. 2011).
eruption with large volumes of juvenile pyro- Given the high probability for the generation
clastics (Phillipson et al. 2013). Moran et al. of long-distance lahars, wide dispersal of pyro-
(2011) maintain that an eruption is “failed” when clastic fall, and the consequential negative
magma reaches but does not pass the “shallow impact on many economic activities and the
intrusion” stage, i.e., the magma gets close to, compromise of critical infrastructure should an
but does not reach the surface. In the actual case, eruptive period last for months to years, Cotopaxi
the amount of erupted material was minor, and is considered a “National” volcano, located in the
had a dense rock equivalent volume of center of Ecuador, near to Quito, the capital and
*0.5 Mm3 (Bernard et al. 2016). other populated areas. Even a short-lived VEI 4
eruption (VEI = Volcano Explosivity Index)
(Newhall and Self 1982) has the potential to
Cotopaxi Volcano gravely affect Ecuador’s overall productivity and
major transport lines. Lastly, the volcano is
Cotopaxi volcano, located in central Ecuador considered one of the most dangerous in Ecua-
atop the Eastern Cordillera, is a large, symmet- dor, given the possible exposure of >300,000
rical stratocone with a basal diameter of 18 km residents to primary lahars and ashfalls during
244 P.A. Mothes et al.

Fig. 1 Location of Cotopaxi volcano in Ecuador’s Inter–Andean Valley and major cities
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 245

future VEI = 3 or greater magnitude eruptions serious unrest of Cotopaxi to date—since 1942,
(Mothes et al. 2016a). when slight ash emissions and mild explosive
activity were then reported (Siebert et al. 2010).
Here we present the macro patterns of seismic,
Monitoring Cotopaxi gas, deformation and visual observation moni-
toring data associated with the awakening vol-
Cotopaxi has long been a producer of swarms of cano. The data are provided by the monitoring
LP’s, and Ruiz et al. (1998) hypothesized that networks operated by the IGEPN. We avoid
they were related to the interaction of the dwelling on details, as forthcoming contributions
hydrothermal system with heat ascending from dedicated to exploiting specific datasets are in
depth. During unrest in 2001–2002 thousands of preparation. We also provide comment on
long period, volcano-tectonic and distal selected actions in which IGEPN scientists par-
volcano-tectonic seismic events were registered ticipated to make the overall societal outcome
and magma was hypothesized to have ascended more favorable in case Cotopaxi produced a
to shallow levels in the center of the edifice major eruption. We impart with the philosophy,
(Molina et al. 2008). Its flanks also experienced stated in Marzocchi et al. (2012) that “sound
deformation from magma input estimated at scientific management of volcanic crises is the
*20 Mm3 from modeling of EDM data (Hickey primary tool to reduce significantly volcanic risk
et al. 2015). But, observable superficial mani- in the short-term”. We also maintain that a con-
festations were meager, no magma erupted and stant and rapid analysis of the monitoring data is
the volcano returned to a relatively quiet state key to giving forecasts that include reasonable
with only short-lived LP swarms (Lyons et al. scenarios. Some of the IGEPN actions were
2012) and sporadic VLP events being registered guided by experiences gained in the VUELCO
until April, 2015 (Márquez 2012; Arias et al. project, since one of the scenarios in the simu-
2015). lated eruptions was that the volcano would wake
The IGEPN has monitored Cotopaxi volcano up, be active then return to repose.
since 1986. Subsequently, over the years the The 14th of August, 2015 explosions and
monitoring coverage has become denser and subsequent emissions pushed the first evidence
more robust (Fig. 2) (Kumagai et al. 2007, 2009, of new magma to the surface, although in a
2010). Presently there are approximately 60 limited way (Gaunt et al. 2016). Documentation
telemetered geophysical sensors operating on its of the geophysical signals and observations that
flanks. Cotopaxi hazard maps have been pub- we registered through late 2015 leads to the
lished in several versions since 1976, with the depiction of what transpired—mainly of an
newest version published in late 2016 (Mothes intrusion, which stayed deep, although the signs
et al. 2016a). The IGEPN has carried out a of intrusions that stall at depth may be very
long-term program of community education for similar to those produced by intrusions that
areas that are at highest risk, although as noted finally do erupt (e.g. Moran et al. 2011).
by Christie et al. (2015), the attention over such a
vast area (2000 km2), was uneven.
Cotopaxi was a VUELCO target volcano from Synthesis of the Geophysical
2013 to 2015. As part of the VUELCO project, in Fingerprints of the Unrest
November 2014 a simulation exercise was car-
ried out with the purpose of presenting a timeline Having passed 13 years with low levels of
of potential unrest and expected events and to activity since cessation in 2002 of its last reac-
test the communication between scientists, tivation, in mid-April 2015 Cotopaxi began
decision-makers and the public (www.vuelco.net departing from background levels: higher seismic
). This present contribution is written in the spirit energy release, gas outputs and superficial man-
of holistically documenting this recent and most ifestations transpired. The height of activity was
246 P.A. Mothes et al.

Fig. 2 Map of the instrumental monitoring network around Cotopaxi volcano, April 2016
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 247

Daily RMS values of BREF seismic station-vertical component

(a) 01 January, 2015 -01 April, 2016
2500

2000

1500

1000

500

0
28/05/2015
01/01/2015

22/01/2015

12/02/2015

05/03/2015

26/03/2015

16/04/2015

07/05/2015

18/06/2015

09/07/2015

30/07/2015

20/08/2015

10/09/2015

01/10/2015

22/10/2015

12/11/2015

03/12/2015

24/12/2015

14/01/2016

04/02/2016

25/02/2016

17/03/2016
(b)

Fig. 3 a Plot showing RMS values of seismic data that August explosions/emissions and subsequent ash emis-
has been segmented into 1 min windows and smoothed sions and emission tremor. Accumulated seismic energy
with a 31-point median filter. Of note is the calm period values through the end of 2015 show a marked decline.
from January to April 2015, than a slight increase in b Number of daily seismic events versus accumulated
seismic energy release in May. The increase in June is seismic energy of these events. The acronyms for different
associated with greater emissions and strong tremor. seismic events are: LP Long Period; HB Hybrid; VT
A sharp decline in early August and later the notable Volcano-tectonic; TREM High Frequency Emission Tre-
increase in August and September represents the 14th of mor; EXP Explosion; VLPs Very Long Period

a series of 5 explosions/energetic emissions on values showed a decline except for brief hikes in
the 14th of August, which expulsed preexisting October and in November, when light ashfalls
conduit plug material, ash and gases, but whose occurred. By December 2015 nearly all moni-
size did not surpass VEI = 1. By late September toring parameters were down to background
2015, activity mostly died back and RSAM levels (Fig. 3a), except for a protracted
248 P.A. Mothes et al.

Fig. 4 Registry of VT and LP seismic events at Cotopaxi since May 1996 to 31 December, 2015. The 2001–2002
period was the only other period with a marked jump in seismic activity before the 2015 episode

volcano-tectonic (VT) seismic swarm that began 2012; Jousset et al. 2013; Maeda et al. 2015;
on 10 September, 2015 and continued through Kumagai et al. 2010; Arias et al. 2015). VLP
March 2016, albeit displaying low levels of events are believed to be generated by volume
seismic energy release (Fig. 3b). This swarm changes and movements of magmatic-
produced nearly 15,000 VT events. hydrothermal fluids (e.g., Chouet and Matoza
2013). Between June 2006 and October 2014,
106 confirmed VLP events were identified at
Geophysical Registry of Cotopaxi’s Cotopaxi (Márquez 2012; Arias 2015). In 2015
Restlessness in 2015 Cotopaxi, VLPs rarely passed 11 events/day
(Fig. 5), but commonly had magnitudes of 2–3.
From 2002 to April, 2015, seismic registry of The recent VLP events that were located under
mostly long period (LP) seismic events averaged the Cotopaxi’s edifice, occurred in sectors of the
around 10 events/day. In April 2015 the monthly volcano where VLP’s had been previously
tally was about 630 earthquakes, then rose to located by Molina et al. (2008) (Fig. 6). The
3000 events in May (Fig. 4), with a jump to greatest number of VLPs, of the 114 located
about 180 events/day registered on 23 May events, were registered during the third week of
(Fig. 4). July up to the explosions on the 14th of August.
Of significance also was the notable increase While most were between 1 and 2 magnitude,
in very long period seismic events (VLPs) some were greater than 2.5 (Fig. 5a). The rela-
recorded since late May 2015. VLPs are often tionship between the great number of LPs which
interpreted to signify magma movement (Zobin started the awakening process at Cotopaxi and
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 249

Fig. 5 a Occurrence of VLP events in 2015 at Cotopaxi. Black arrow represents 14 August hydromagmatic
explosions. b Accumulated seismic energy from VTs, LPs and VLPs in 2015–2016 at Cotopaxi

afterwards the stalling out of these events to be Most VLP events had frequencies between 0.1
followed immediately by the strong VLPs is and 1 Hz and had strong P and S waves, such as
another possible indicator of the precursory nat- the example given for 04 August, 2015 which
ure of this type of volcanic earthquake before the was located 3 km below summit on the NE flank
discrete eruptions on 14 August (Fig. 5b). of the volcano (Fig. 6).
250 P.A. Mothes et al.

Fig. 6 Waveform of VLP registered on 04 August, 2015 15h27 GMT is one of the largest events registered at
Cotopaxi (local magnitude 2.5) and was located 3 km beneath the crater on the NE flank

The locations of earthquakes (of all types) Overall, there was a marked increase of LP
from January to December, 2015 were at two events from April to late May, followed by high
levels: at depths of about 3–5 km below the frequency tremor episodes (Fig. 8) which lasted
crater (Fig. 7a) and at a deeper level of 7–15 km until the onset of high frequency tremor related
below the crater. Most events were aligned with to gas emissions and which became prominent
the conduit. However some distal VTs were from 04 June and lasted to the second week of
registered about 15 km due north of the volcano August (represented by black bars), and could
(Fig. 7b) and were interpreted as fault slips due have been related to the boiling of the volcano’s
to stress transfer from the volcano (White and hydrothermal system, and coincided with the
McCausland 2016). Distal VTs were also high water vapor and SO2 flux then emitting
important in the reactivation of Pinatubo volcano from the crater (Bernard et al. 2016). In Fig. 8,
(Harlow et al. 1996). the VLPs that were important especially in July
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 251

−0.55 LOCATION OF SEISMIC EVENTS COTOPAXI VOLCANO (2015−04−01 − 2015−06−30)

−0.55
Quality factors B
Minimun magnitude= 0.1 B
RMS maximun value= 0.5
Minimum number of phases used= 6
−0.58

−0.58
Coverage maximum value GAP=270
Velocity model= cotopaxi
−0.61

−0.61
−0.64

−0.64
Latitude (degrees)

−0.67
−0.67

pBB[, 2]
A A'

−0.70
−0.70

−0.73
−0.73

−0.76
−0.76

−0.79
−0.79

−0.82
−0.82

B' B'
−78.57 −78.54 −78.51 −78.48 −78.45 −78.42 −78.39 −78.36 −78.33 −6000 −3000 0 3000 6000 9000 12000 15000
−6000

Depth (meters)
LEGEND
A A'
Seismic station Cotopaxi Volcano network
−3000

Volcanic Tectonic Events (VT)

Profundidad (metros)
Long Period Events (LP)
Very Long Period Events (VLP)
Hybrid Events (HB)
Explosions (EXP)
0

Tremor (TREM)
Non clasified events

Relative frecuency
Today events located (14 July 2016)
3000
Depth (meters)

Earthquakes number located

707

1.0
Volcanic Tectonic eartquakes magnitude histogram
6000

0.8
9000

0.6
0.4
12000

0.2
15000

0.0
−78.57 −78.54 −78.51 −78.48 −78.45 −78.42 −78.39 −78.36 −78.33 −78.30 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2
Longitude (degree) Magnitude
2016−07−14 21:23:16

Fig. 7 Top-Locations of seismic events from a–01 April to 30 June, 2015 and Base—01 July to 31 August, 2015.
Most are aligned with the conduit, however the SE flank is favored for harboring event locations

to mid-August are masked by this tremor signal, GPS stations on the W and S flanks showed
but can be observed in Fig. 3 and their accu- horizontal displacements of almost 16 ± 0.5 mm
mulative energy levels are shown in Fig. 5. toward the W and SW. GPS stations on the NE and
The increase in SO2 gas emissions, rose to E flanks showed displacements to the N at a
3000 ton/day by the end of May with a clear SO2 reduced velocity (Fig. 10). The vertical compo-
signal progressively more notable through late nent registered a maximum uplift of
May into June (Fig. 9) (Hidalgo et al. 2016). For 15 ± 2.3 mm. The movement to the west could
example, on the 22nd–23rd of May odors of have been accentuated by the volcano’s morphol-
sulphur were very evident above the 5700 m ogy, as the W flank is poorly buttressed and sits
level on the volcano’s northern flank, as reported upon Inter-Andean Valley volcaniclastic fill. In
by Cotopaxi Park personnel. comparison the east and northern part of the cone
252 P.A. Mothes et al.

−0.55 LOCATION OF SEISMIC EVENTS COTOPAXI VOLCANO (2015−07−01 − 2015−09−30)

−0.55
Quality factors B
Minimun magnitude= 0.1 B
RMS maximun value= 0.5
Minimum number of phases used= 6
−0.58

−0.58
Coverage maximum value GAP=270
Velocity model= cotopaxi
−0.61

−0.61
−0.64

−0.64
Latitude (degrees)
−0.67

−0.67
pBB[, 2]
A A'
−0.70

−0.70
−0.73
−0.73

−0.76
−0.76

−0.79
−0.79

−0.82
−0.82

B' B'
−78.57 −78.54 −78.51 −78.48 −78.45 −78.42 −78.39 −78.36 −78.33 −6000 −3000 0 3000 6000 9000 12000 15000
−6000

Depth (meters)
LEGEND
A A'
Seismic station Cotopaxi Volcano network
−3000

Volcanic Tectonic Events (VT)

Profundidad (metros)
Long Period Events (LP)
Very Long Period Events (VLP)
Hybrid Events (HB)
Explosions (EXP)
Tremor (TREM)
0

Non clasified events

Relative frecuency
Today events located (14 July 2016)
3000
Depth (meters)

Earthquakes number located

831

1.0
Volcanic Tectonic eartquakes magnitude histogram
6000

0.8
9000

0.6
0.4
12000

0.2
15000

0.0
−78.57 −78.54 −78.51 −78.48 −78.45 −78.42 −78.39 −78.36 −78.33 −78.30 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0
Longitude (degree) Magnitude
2016−07−14 21:28:41

Fig. 7 (continued)

Fig. 8 a Comparative graph of LP seismic events began the second week of August, 2015, after the c
(orange bars) and high frequency tremor (black bars) hydromagmatic explosions on the 14th. b Seismogram
possibly related to the boiling of the hydrothermal system (11 June, 2015) of BREF station showing registry of
and gas movements from 01 April to 14 August, 2015. spasmodic tremor related to internal fluid movements in
Ash and gas emission-related tremor (pink bars) abruptly the upper part of the edifice

lies upon a thick lava package and basement annual and semi-annual components using all
crystalline metamorphic rock and may be more available data between 2008 and 2015. The tran-
resistant to lateral movement. Data processing sient displacements identified during the 2015
employed the program GAMIT/GLOBK (Herring unrest period are with respect to this model. In a
et al. 2015) and used a local reference frame with second step, we applied a common-mode filtering
respect to fixed South America (Nocquet et al. estimated from the average time series residuals for
2014. We also defined a long-term displacement sites *50 km away from the volcano. Short-term
model for each GPS site by estimating a trend and repeatabilities are of the order of 1–2 mm on the
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 253
254 P.A. Mothes et al.

SO2 Emissions of Cotopaxi- Jan. 2015- Jan 2016

20,000

15,000
SO2 Flux, ton/day

10,000

5,000

0
01 May 2015

01 Jan 2016
01 Jan 2015

01 Dec 2015
01 Sep 2015
01 Feb 2015

01 Mar 2015

01 Nov 2015
01 Oct 2015
01 Aug 2015
01 Apr 2015

01 Jun 2015

01 Jul 2015

Fig. 9 Registry of SO2 values for Cotopaxi, January NOAA and VAAC alerts. Graph included in Cotopaxi
2015 until 05 January 2016, with a significant increase in Special Report, No. 1, 2016: http://www.igepn.edu.ec/
SO2 observed since early May. Data was processed daily cotopaxi/informes-cotopaxi/coto-especiales/coto-e-2016/
using a single wind speed and direction obtained from the 14074-informe-especial-cotopaxi-n-01/file

horizontal components and 3 mm on the vertical tremor were both strong, a positive tilt signal
component, enabling us to extract the small GPS predominated.
signal observed during the unrest period.
Due to the westward movement on the GPS
station PSTO, which is 22 km W of the crater Hydromagmatic Explosions/Strong
(Fig. 10), we surmised that the source was Emissions of 14 August, 2015
deep. Subsequent modeling of the data suggested
a source of about 24 km deep located under the On the evening of 13 August, a swarm of VT and
SE flank with a volume of 42 ± 26 Mm3 LP seismic events was registered between 20h03
(Mothes et al. 2016b). Nonetheless, as mentioned (GMT) on the 13th to 08h55 (GMT) in the early
in Sect. 3.1, analysis of the erupted ash suggests hours of the 14th, antecedent of the explosion
that the magma source is shallow, as least for the events (Fig. 11). At 09h02, 09h07 and later at
initial small volume that was emitted. 15h25, 18h45 and 19h29 (UT) five small
Data from a tiltmeter (VC1G on Fig. 2) explosions/strong emissions were registered at
installed in a thick lava package and located Cotopaxi which served to unblock the conduit
6 km NE of the crater, showed a strong and led to ejection of degassed altered conduit
inflationary pattern that had started in April, 2015 plug material and scarce juvenile components.
on both axis. This tilt anomaly coincided with the Although infrasound from these explosions did
notable increase in seismicity (Mothes et al. not exceed 4 Pascals (Pa) at stations located
2016b). Generally, when LP seismicity and approximately 6 km from the vent, the first two
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 255

Fig. 10 GPS vectors for stations around Cotopaxi’s cone represent the comparison of GPS data collected from 01
and one to the west for October, 2015. Displacements are January to 01 October, 2015 at the 7 station
expressed with respect to the North Andean Silver and Cotopaxi CGPS network

explosions were heard by climbers in the Refuge totally evacuated by the explosions. Observers
on Cotopaxi’s north flank, where lapilli-size also reported that the fallout had a “wet aspect”
fallout reached the Cotopaxi Refuge. and many of the fragments were agglutinated
Two months earlier public and authorities had by a fine clay-size patina. The eruption is cat-
been forewarned in the special IGEPN reports egorized as hydromagmatic, since the rapid
(No. 3 and 4) that phreatic explosions would be a interaction with water caused overpressures
likely phenomenon in precursory eruptive activ- beneath the plug, raising lithostatic pressures
ity (http://www.igepn.edu.ec/cotopaxi/informes- that overcame the capacity of the altered con-
cotopaxi/coto-especiales/coto-e-2015/12990- duit plug rock. After these main vent-opening
informe-especial-cotopaxi-11-06-2015/file). events the presence of hydrothermally altered
With these explosions the eruption column material gradually waned and possible juvenile
at 15h25 rose to 9 km above the crater rim and material became more prevalent (Gaunt et al.
was clearly visible from the SW (Fig. 12a, b). 2016).
Infrasound values of the explosions were less The ash emissions from this first activity
than 10 Pa at station BNAS (5 km from the covered agricultural lands to the NW and W of
crater), but the seismic source amplitudes of the the volcano with a  1 mm thick dusting of
tremor associated with the first two explosions altered silt to sand lithic grit and crystals
were greater than those of most Cotopaxi LP (Fig. 13a) and caused poor visibility along major
events and also of some explosions registered highways that enter Quito from the south. This
at Tungurahua volcano (Kumagai et al. 2015). ash emission mantled over 500 km2 with more
The initial explosions had evidence of water than 80 gr/m2 and amounted to a volume of
involvement. In previous weeks a small lake 118,000 m3, keeping it within the range of a
was observed in the crater’s bottom; this was VEI = 1 (Bernard et al. 2016) (Fig. 13b).
256 P.A. Mothes et al.

Fig. 11 Registry of VT—LP seismic swarm which begin late 13 August, 2015 and was followed by the 5 small
explosions/strong emissions on 14 August, all indicated with red arrows. Seismograms are of the IGEPN’s monitoring
network

was that of the ash and gas plume cascading

Post 14 August, 2015: Open Conduit down the W flank, with only the initial pulse
Degassing and Ash Emissions rising to <1 km upon emitting from the crater
(Fig. 14).
Ashfalls were prevalent towards the N and NW Emission tremor of varying amplitudes
after 14 August into October and became scarce accompanied the ash emissions and permitted
in late November (Fig. 14). A common scene IGEPN monitoring scientists to forecast if the
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 257

Fig. 12 Left Cotopaxi’s 14 August, continuing emission produced a column that ascended 6–8 km above the
—view of the volcano from SW at 14h10UT. Photo E. summit. Photo by Santiago Tapia, at the Novacero
Pinajota, IGEPN. Right- The 15h25UT strong emission company grounds, 20 SW of the volcano

Fig. 13 a Ashfall from 01 September emissions accumulated in cultivated fields near El Chaupi town. b Ash fallout map
associated with the eruptive activity of Cotopaxi on the 14th–15th of August, 2015. Map projection WGS 84, coordinates in
UTMs. Values expressed in isomass of grams/m2. Source Bernard et al. (2016)

rate of ash falls was increasing during frequent 15th and 24th of August, when OMI measure-
foggy, overcast conditions (Bernard et al. 2016). ments gave readings of 6500 and 6600 ton/day,
On the 14th of August, after the explosions, SO2 respectively.
levels reached 16,400 ton/day as registered by In early September ash columns still rose to
the satellite sensor OMI (http://so2.gsfc.nasa. over three kilometers height above the vent and
gov/pix/daily/ixxxza/loopall3.php?yr=15&mo= carried fine ash particles to cities on the piedmont
08&dy=15&bn=ecuador) (Fig. 15). Subse- of the coastal plain, such as Santo Domingo de
quently, SO2 levels were particularly high on the los Colorados, located 120 km W of the volcano,
258 P.A. Mothes et al.

Fig. 14 Left Photo with view toward south taken from 2015, from the north side of Cotopaxi. A low gas and ash
Autopista Ruminahui (SE of Quito) on 20 August. Photo- column trending to the NW is observed. Photo P. Mothes,
C. Zapata- EPN. Right Photo taken on the 23rd of August, IGEPN

essentially situated under the red swath trending which implied low magma ascent rates and
W in Fig. 15. stiffening of magma in the upper part of the
Ashfall was still prevalent in mid-October, but column (Gaunt et al. 2016).
had all but terminated the third week of Ashes collected on the 20th of October, had a
November where it was seen W-NW of the high concentration of dense microcrystalline
volcano. VAAC Washington again reported material. Although there is evidence of few
suspended fine ash above Santo Domingo as well vesiculated clasts (diktytaxitic texture); about
as in Los Rios province to the SW. The ash 65% of the ash is considered possibly juvenile.
column generally rose to only 1 and 3 km above Gaunt et al. (2016) suggest that the origin of the
the summit and had a velocity between 6 and ash is the top of a degassed magma column which
10 m/s and lasted about one week. Fieldwork had ascended from about 3 km below the crater.
permitted the estimation of a mass and volume
total of 3.49  107 kg (22,100 m3) for this late,
waning period (Bernard et al. 2016). The total Seismicity
ashfall dense rock equivalent (DRE) volume for
the entire eruption was calculated in 0.5 Mm3 For most of the post explosive period after
(Bernard et al. 2016). mid-August, seismic hypocenters still remained
located at the two depths mentioned above
(Fig. 7). Most relevant was the sporadic occur-
Ash Componentry rence of VT events with magnitudes of 3 or
greater that occurred. Sometimes spasmodic tre-
Analysis of ash beneath both binocular and mor was registered and continued for hours, as
scanning electron microscope showed clearly for example, that registered on 02 September,
that there was an evolution in ash componentry 2015.
from the eruption’s beginning on 14 August and Starting on 10 September, a swarm of VT
later. The first ash from 14 August had more seismic events kicked in with a rate of approxi-
hydrothermal lithics (pyrite, scoria with vesicles mately 100 events/day and a daily registry of
filled with altered material and hydrothermal coeval small internal explosions which has
quartz). As the eruptions progressed we saw an associated infrasound signatures, (shown in
increase in more fresh magmatic components, green color in Fig. 16). This swarm lasted past
such as free crystals, glass particles with low the New Year, but the overall seismic energy
vesicularity and a high percentage of microlites, release was low (Figs. 3b and 5).
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 259

Fig. 15 Registry of SO2 (16,400 ton/day) from satellite as detected by the OMI sensor on the 14th of August, 2015

Visual Observations and Secondary small secondary lahars descending several

Effects streams and we estimated that many of them
were due to increased glacier melting.
Thermal images (August/2015) showed the Orthophotos made on August 18 and then
presence of new thermal anomalies (*15 °C) again on October 8, show a decrease of about
inside the crevices on the N side glaciers, at the 0.49 km2 of the area covered by glaciers. This
same time fumarolic gases were observed com- represents a very high rate of glacier melting, not
ing out from those fractures. The highest tem- explained exclusively by climate change
perature obtained was about 200 °C (Fig. 17) (Cáceres et al. 2016).
from gases ascending the crater. We estimate that small volumes of magma
On September 3, water emerging from the reached surface levels in the volcano conduits
basal fronts on the northern glaciers was clearly causing increased circulation of hot fluids inside
observed, and countless new crevices in the the edifice, apparently reaching the basal area of
majority of glacier ends and on the upper flanks the glaciers and producing increased melting. It
were evident. All this led to the conclusion that is necessary to further investigate the hazard due
an abnormal process was producing increased to instability in the melting glaciers and their
melting of the glaciers. Starting in mid Septem- eventual collapse that could lead to greater sec-
ber it was possible to observe the presence of ondary lahars. Numerical modeling by
260 P.A. Mothes et al.

Fig. 16 Registry of overall seismicity at Cotopaxi vol- finally with the advent and continuance of frequent VT
cano from 01 January 2015 to 01 April, 2016. Most seismic events with accompanying internal explosions
notable is the presence of LP earthquakes in the first that had no superficial manifestation, except infrasound
semester of 2015, later followed by emission tremor and registry

Hemmings et al. (2016) has shown the impor- and there were frequent small secondary lahars.
tance of hyrdrothermal perturbations at Cotopaxi Those lahars that have been especially associ-
in generating watery flows. ated with rain storms obtained the highest dis-
Incandescence was also occasionally observed charges—on the order of 10–30 m3/sec (D.
on cold still nights with a thermal camera or by Andrade-IGEPN, Pers, Comm, 2015). The
simple vision. These events were considered to Agualongo channel, on Cotopaxi’s W side was
have been caused by pulses of hot gases. frequently flooded by lahars and on three
The glacier around the crater, on the W and occasions they covered partially the main road
N flanks, became partially covered by ash. This giving access to the volcano.
coating of dark ash decreased the glacier
albedo and consequently increased the absorp-
tion of solar rays. Therefore expedite melting Interpretation and Model
of the glacier tongues increased, leaving obvi-
ous melt water channels issuing from the gla- As shown in Fig. 16, LP events gradually
ciers’ base. increased starting in April, 2015, beginning with
As a result of the afternoon melting by small magnitudes and low energy levels (Fig. 3).
insolation and perhaps also by higher tempera- We interpret the LPs to imply fluid movement
tures of the rock beneath the glaciers, runoff occurring at 10–12 km below the crater and then
increased, especially off the W flank glaciers up to shallower levels (Fig. 7). There were only
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 261

Fig. 17 Thermal image taken on 03 September, 2015 from fractures below the summit rim. Photo P. Ramón,
looking toward the SE sector of the upper cone. The IGEPN. Source http://www.igepn.edu.ec/cotopaxi/
infrared camera detected temperatures of 200 °C associ- informes-cotopaxi/coto-especiales/coto-e-2015/13529-
ated with emission from the crater and lower temperatures informe-especial-no-14/file

scarce VTs that occurred in concert with the LPs. implying that there may already have been
In late May VLPs were registered (Fig. 5) and magma ascent from a deeper depth to a shallower
were interpreted to be possible mass magma reservoir in order to show changes on the most
transport processes as reported by Arias et al. proximal tiltmeters. The GPS stations begin to
(2015). Often VLPs are identified as eruption show minor displacements at the same time,
precursors, e.g. at Redoubt (Power et al. 2012), particularly on the W-SW flank, where no strong
for example. In hindsight, vigorous VLPs were evidence was detected in seismicity, but to the
also registered in 2009–2010 at Cotopaxi and contrary seismicity was concentrated more on the
correlated with recoverable deformation patterns E-SE flanks (Fig. 7). In sum, we registered both
at borehole tilt stations (Mothes et al. 2010), but shallow and deep seismic activity. A leading
did not result in an extended seismic crisis or hypothesis for this pattern likely was the inter-
magmatic activity on the surface. action between fluids being released by a
Nonetheless, LP and VLP events registered in deep-seated source, say at 24 km depth as pro-
April–August, 2015 were apparently responding posed in our geodetic model. These fluids
to a slow ascent of a small magma slug and ascended and perturbed a preexisting
associated fluids and in April 2015 deformation shallow-seated source, which may be the mag-
recorded by tilt and GPS stations began almost matic remnant that drove the 2001–2002 unrest
synchronically with the jump in LP seismicity, reported by Hickey et al. (2015).
262 P.A. Mothes et al.

With more new magma in Cotopaxi’s system, result of the repeated formation and destruction
SO2 output became prevalent in mid-May, one of a shallow magmatic plug by brittle fragmen-
month after the hike in seismicity. Background tation through mechanical stresses and decom-
SO2 levels of <500 ton/day were surpassed and pression. Gas overpressure must have been
rose to over 3000 ton/day. The strong onset of accumulating beneath the conduit plug and may
bands of tremor about the 1st of June, were have contributed to the flank deformation, par-
conjectured to be related to continual fluid ticularly as registered by the tiltmeters. In the
movement within the edifice and perhaps to the succeeding post-explosion days SO2 output rose
boiling of the hydrothermal system and was a to 16,000–18,000 ton/day (Hidalgo et al. 2016), a
signal that more overall heat was circulating likely testimony to the accumulated gas that had
within the edifice. been trapped in the plumbing system.
Along with the rise in SO2 there was also a The expulsed material showed evidence of
trend in production of more VLP’s since 114 of strong hydrothermal alteration and there was
these events were registered between May and initially little evidence of juvenile components.
mid-August, 2015. Of great significance is that There was also a low pH (3.6–5.1) and high
the largest VLPs were registered in the last sulfate- SO4 concentrations (up to 13,000 mg/kg)
3 weeks before the hydromagmatic explosions. in the expulsed ashes of the 14th to 25th of
Afterwards too there were infrequent VLP’s August, as detected by leachate analysis
(Fig. 5). (P. Delmelle, Pers. Comm, 2015). Later, as
With the highest energy levels of the VLPs described above, the percentage of juvenile
being logged before the explosions, these events components increased through time, to the last
seem to have been one of the detonators of the erupted material collected in late November,
explosions (Fig. 5b). They seemed to herald that 2015.
magma/fluids were ascending. Of particular note After the explosions and strong emissions of
was the VT/LP swarm of the 13th–14th of the 14th of August, the conduit lost its retaining
August that began 12 h before the hydromag- plug and remained open and continual fluid
matic eruptions on the morning of 14 August movement was facilitated, although pulsatile
(Fig. 11). The swarm comprised of some 40 VTs superficial activity continued, and few shallow
and >50 LPs was the most energetic of any explosions occurred. After the explosions LP
seismic swarm registered at Cotopaxi since 2002 events were initially high (>200 events/day), but
and was a warning in hindsight which presaged dropped to <20 events/day by the first week of
the subsequent explosions/strong emissions some September, where they remain at this writing.
hours later. These seismic trends and the higher While the LPs diminished, to the contrary the
SO2 flux, would indicate that magma was VT events rose notably. Around the 1st of
working in the upper part of the system—at least September 15–20 LP events/day were registered.
in the 0 to 6 km level below the crater. By the third week of September these increased
The initial explosions had a phreatic compo- to >200 events/day (Fig. 16), and subsequently
nent since water was available in the small pond this value decreased to 50–100 events/day, but in
at the crater’s base, sub-glacial melting and from all totaled nearly 15,000 events. In the last
pore water within the hydrothermal system. 3 months of 2015 persistent VT daily activity
Nonetheless, Gaunt et al. (2016) argued that the was registered as well as waning tilt and GPS
most likely driving force of the initial explosions offsets. Even with the VT swarm overall seismic
was magmatic heat interacting with the energy levels decreased compared to the levels
hydrothermal system providing energy to trigger registered in May to August, which may be due
hydromagmatic eruptions at Cotopaxi. Textural to the overall successful degassing of the system,
evidence for this process was only preserved in but could also be explained by closing of the
the deposits of the initial eruptions, but not conduit by a degassed magma column, thus
subsequent ones. Later emissions were likely the impeding freer liberation of gases. Between
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 263

Fig. 18 Cartoons synthesizing the internal and superficial processes observed from April to December, 2015

October 2015 through April, 2016 internal through restrictive areas within the conduit.
explosions of deep providence were registered at While displaying only minimal infrasound and
a rate of 20–30 such events/month. These no detectable superficial vestiges, these explo-
explosions could be interpreted as gas passing sions may be occurring due to pressurization
264 P.A. Mothes et al.

Fig. 18 (continued)

deep within the system (e.g. Valentine et al. perturbed the system remained at depth. As a
2014). comparison, the 2001–2002 restless period had
A synthesis of the 2015 unrest at Cotopaxi, essentially been internal with only the weakest
the hypothesized driving forces and possible of superficial manifestations. For the 2015
paths taken are shown in the following schematic reactivation IG scientists took the warnings very
cartoons (Fig. 18). seriously and put all their collective experience
to the test to make interpretations of the moni-
toring data and to manage successfully the great
The Science-Society Interface expectations of the public and authorities during
the crisis period described above. Over-
Strong evidence points to a magmatic compo- all IGEPN volcanologists wrote 23 special
nent in Cotopaxi’s 2015 reactivation and that reports that were disseminated via multiple
this may be a precursor for future episodes, media (www.igepn.edu.ec). Given the low levels
given that the bulk of the new magma that of acceleration of seismic energy, small ground
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 265

displacements and visual observations, eruption Management at all levels and participation in
scenarios that the IGEPN formulated stayed guiding eruption simulations. Some of the dis-
within the realm of VEI 1–2 levels and clearly cussions with them were based on what had been
stated that the least likely scenarios was the reviewed in the VUELCO workshop-simulated
generation of a paroxysmal eruption in which eruption exercises carried out in late 2014 in
PDC’s, voluminous ashfalls and giant lahars Quito. The IGEPN also greatly benefitted by the
would be formed. strong collaboration and presence of members of
With each convincing sign that Cotopaxi was the USGS/USAID Volcano Disaster Assistance
displaying stronger activity, IG scientists were Team who led informed discussions on the trends
proactive in improving and strengthening all of the geophysical precursors and also helped to
monitoring systems while simultaneously help- reinforce the lahar-detection network. Personal
ing to prepare the populations and authorities for from Chalmers University of Technology (Swe-
what a major eruption of Cotopaxi could mean. den), JICA (Japan), IRD (France) and NASA
Work by Christie et al. (2015) (an earlier (USA), DEMEX-EPN with help in SEM,
VUELCO contribution) had shown that residents LMU-Germany with grain size x-ray diffraction
of the Chillos valley, to the east of Quito, were and UCL (Belgium) with leachates, also collab-
particularly ill-prepared to confront lahar haz- orated during the crisis.
ards due to their recency of living in that valley. As the eruption process waned, it was obvious
On the other hand, residents of the Latacunga that high-risk populations were tired of being
valley had a clearer memory of lahar hazards, constantly alert and didn’t want to be perpetually
since many of their distant relatives had lived attentive to volcanic processes that could threaten
through Cotopaxi eruptions and their collective their livelihoods and families. This issue will
memory is better preserved. Nonetheless, social have to be acknowledged and dealt with in future
media, both beneficial and alarming, steered reactivations.
perceptions and actions of residents. The area of
influence by the volcano, especially with respect
with lahars, includes four important provinces, Conclusions
several counties and Quito’s jurisdiction. It was
particularly difficult to meet the demands and Seismic activity and its evolution in event types,
expectations and provide the personal attention energy release, shallowing depths and locations,
of monitoring scientists to the authorities in each elevated degassing and ash emissions and flank
of these different municipalities as well as to deformation typified the restlessness of Cotopaxi
meet with other community groups and respond during 2015. The important accumulative energy
to their uncertainties. There were also the con- release first of LPs than followed by registry of
stant attacks on social media of several particu- an important suite of large VLP’s was a signif-
larly meanly-intentioned individuals who icant geophysical pattern indicating fluid move-
constantly tried to steer attention away from the ment, followed then by a more convincing
IGEPNs scientific work by saying that it was transfer up conduit of small slugs of magma and
operating with poor instrumentation or that the gases to beneath the conduit plug. In the late
monitoring work at IGEPN was “carried out by hours of the 13th of August this plug fractured
amateurs”. and ruptured, evidenced by the vigorous swarm
All told, IGEPN scientists provided abundant of VTs and LPs before the hydromagmatic
custom guidance to local and national officials explosions that occurred early on the 14th of
and residents with regards to volcano hazards August 2015. Minor ground deformation, the
and the proposed scenarios. A total of about 125 small, limited explosions in August and subse-
talks were given by IGEPN personal during the quent ash emission suggest that the ascended
unrest period. Additionally, there was broad magma volume was small, and indeed as cal-
coordination with Ecuador’s Secretary for Risk culated by Bernard et al. (2016), was only about
266 P.A. Mothes et al.

0.5 Mm3 DRE. This value is far inferior to the Montserrat (Sparks and Young 2002) to suggest
possible volume of 42 ± 26 Mm3, which is that the next round of Cotopaxi eruptions could
hypothesized to be at depth based on modeling last more than just several months. In the case of
of the observed GPS displacements (Mothes Tungurahua, activity started gradually in 1999
et al. 2016b). A second energetic magmatic and displayed oscillating low-level behavior over
pulse did not arise, and certainly not one with a the years to finally generate a rapid-onset VEI 3
sufficient volume to produce a VEI 3 or 4 eruption in 2006 (Hall et al. 2013). Such long
eruption, which was one of the least likely sce- waits test the population’s resilience, but is also a
narios, but nonetheless dreaded by the society time for monitoring scientists to become
and scientists. acquainted with the volcano’s eruption style.
Following the 14 August explosions and During a reactivation period of a long dormant
subsequent ash emissions we did not observe a volcano there are many uncertainties and this
new phase of outward GPS displacement trends demands stringent work and continual mindful-
in the deformation data, which could have ness by monitoring scientists and frequent
implied a new magma input to cause another ongoing interactive and personal communication
phase of deformation. The post-explosion VT with local communities and authorities.
seismic swarm which lasted 5 months was During the 2015 unrest period at Cotopaxi,
indicative of persistent internal perturbation but people living in high-risk zones (i.e. Latacunga
did not transpire in a new phase of deformation, and Valle de los Chillos) were swayed by spec-
thus we assumed that we were dealing with a ulation, rumors and lies concerning the status of
small magma volume. The magma that tipped off the volcano. Some people also tended to
the 2015 unrest may have been a remnant of that weigh-in toward imprecise information posted on
which provoked the 2001–2002 episode and was Facebook or Twitter and heed pseudo volcanol-
reported by Hickey et al. (2015). This residual ogists and detractors, rather than rely on infor-
magma could have been disturbed by the mation from official channels. It was not
ascending heat and fluids from the new magma uncommon to receive telephone calls from hys-
input at depth (*24 km) whose source was terical residents in either of these population
possibly under the SE flank, and which provoked centers inquiring if a Cotopaxi eruption was im-
the recorded ground deformation and the LP and minent? All told the IGEPN put out 3 reports
VLP seismicity. each day about the volcano’s activity and more
The volcano was benevolent and had awak- than 24 special reports, all which are available on
ened to only a VEI 2 level. No major damage the IG website (www.igepn.edu.ec). The number
was imparted upon the population or on liveli- of followers on the IGEPN’s Facebook page
hoods, except for temporary local economic grew to >1 million. To help stem the flow of bad
depression, increased anxiety of the population, information at the community level Ecuador’s
mild crop losses and premature selling of live- 911 system, in coordination with IGEPN per-
stock due to fears of future losses. Overall, the sonal, formed a pan-volcano vigía network
volcano’s manifestations served as a warning to comprised of volunteer observers who report via
everyone to keep attentive of Cotopaxi’s capacity radio several times a day about their visual and
to cause destruction and possible severe ruin by audible observations of Cotopaxi or the rivers
lahar transit down major drainages which are that are borne on it. This network with 55 vol-
heavily populated and host important strategic unteers, is in many ways a replica of the suc-
infrastructure. cessful community-based vigía system that has
An eruption process can last months to dec- functioned at Tungurahua volcano since 2001
ades, and we need only to look at Tungurahua, an (Stone et al. 2014; Mothes et al. 2015b). The
andesitic stratocone also in Ecuador’s Cordillera information provided by the vigía volunteers
Real, with ongoing eruptions for 17 years compliments the ongoing geophysical monitor-
(Mothes et al. 2015a, b), or Soufrière Hills on ing and also serves to strengthen their capacities
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 267

as community leaders and guides during volcano In all likelihood little or no evidence of the
crisis (Espín Bedón et al. 2016). 2015 restless period will be preserved in the
A hypothesis for a future trend in activity geological record. We know from written chron-
weighs heavily towards hydromagmatic to Vul- icles (1534–1877) that Cotopaxi often had weeks
canian explosions which may have a rapid onset, to months of ramping up before unleashing VEI 3
similar to the 14th of August episode, then or 4 eruptions, i.e. there were probably several
evolve to sub-Plinian to Plinian eruptions of poorly preserved 2015-sized like events, and
VEI 3–4 magnitude, if enough magma has therefore unrest has been poorly documented. In
accumulated at a relatively shallow depth (maybe this recent case, scientists had the benefit of
0–7 km below the crater) as shown in Fig. 7 for observing and analyzing the geophysical moni-
the upper level seismicity, and can make it to the toring output during the entire episode and
surface before degassing. Vulcanian eruptions knowing what level of activity the volcano was at.
have been prominent in the volcano’s historical But, monitoring scientists, just like the citizens of
activity (Gaunt et al. 2015). Unraveling the story Ecuador, experienced the anxiety of pondering
will be difficult. what could be the volcano’s next steps, i.e., the
As the volcano is well-monitored possible rapid intrusion of a new batch of
24 h/day/365, we anticipate that the IGEPN volatile-rich magma or returning to calm. Fortu-
will provide early warnings to the public and nately, in this time around, the first scenario did
officials before onset of important eruptive not transpire. However, this training opportunity
activity. This 2015 “dry run” allowed for diver- that we experienced could prove invaluable for
sification and hardening of Cotopaxi’s monitor- when the next scenario is played out.
ing network, frequent preparation and reappraisal
of eruption scenarios and for the creation of a Acknowledgements This work was supported by a grant
from European Commission FP 7 program
society-wide discussion of the possible conse- (ENV.2011.1.3.3-1; grant no: 282759; VUELCO). We
quences of a large Cotopaxi eruption. Some of would like to thank the IGEPN staff for keeping all
these steps were facilitated by previous work in a monitoring operations at Cotopaxi volcano optimally
VUELCO workshop. Essentially, attending to functioning. Members of the IGEPN’s Volcanology group
gave most of the talks and explanations of eruptions
the 2015 activity was an opportunity to test the scenarios to authorities, community groups and wherever
level of preparedness of the scientists and of the was necessary during the crisis. We acknowledge the
Ecuadorian society. All IGEPN scientists strived VUELCO project for opportunities to share opinions
hard to be ready to “call it right”, had the occa- about future eruptive crisis scenarios. Also, we are
grateful for the support of the VDAP/USAID team during
sion arisen and a large eruption was in prepara- the crisis. The EPN, SENESCYT and SENPLADES
tion. Since so little new magma erupted and there provided funding for much of the instrumentation and
was no detectable subsequent shallow magmatic daily operations. Support by JICA instrumentation and
recharge, we consider the eruption as extremely guidance is also noted. The Ecuadorian military provided
overflights of Cotopaxi, while IGM provided ortophotos.
small, and that the residual magma is in repose The LMI project was instrumental in carrying out near
until a future time. Overall, the crisis was an real time petrologic monitoring. Thanks to Viviana Val-
important opportunity for learning about Coto- verde for preparation of several figures. Recognition is
paxi’s restlessness, with particular recognition of given to the expressions of interest by STREVA project
members on the societal implications of eruptive activity
the increase in the VLP events and their energy at Cotopaxi.
levels just weeks before the mid-August explo-
sions and the synchronous but progressive
ground deformation signals, albeit small, that References
coincided with the increased seismicity. These
two patterns more than any other geophysical Arias G (2015) Estudio de las Señales Sísmicas de Muy
signals, announced the ascent of Cotopaxi’s Largo Periodo del volcán Cotopaxi. Proyecto de
magma, although finally only a small quantity Titulación, Facultad de Ciencias, Escuela Politécnica
breached the surface. Nacional
268 P.A. Mothes et al.

Arias G, Molina I, Ruiz M, Hernandez S, Alvarado A, of Mount Pinatubo, Philippines. University of Wash-
Plain M, Mothes P, M Yépez, Hidalgo S, Barrington, ington Press, Seattle, pp 285–306
C (2015) very long period seismicity accompanying Hemmings B, Whitaker F, Gottsmann J, Hawes MC
increasing shallower activity at Cotopaxi Volcano. In (2016) Non-eruptive ice-melt driven by internal heat at
Conference Paper No. S51D-2724, AGU Fall Meet- glaciated stratovolcanoes. Submitted to J Vol
ing, San Francisco, 14–18 December Geotherm Res
Bernard B, Battaglia J, Proaño A, Hidalgo S, Vásconez F, Herring TA, King RW, Floyd MA, McClusky SC (2015)
Hernandez S, Ruiz M (2016) Relationship between Introduction to GAMIT/GLOBK, Release 10.6.
volcanic ash fallouts and seismic tremor: quantitative Department of Earth, Atmospheric, and Planetary
assessment of the 2015 eruptive period at Cotopaxi Sciences, Massachusetts Institute of Technology, PSF:
volcano, Ecuador. Bull Volcanol. doi:10.1007/ http://www-gpsg.mit.edu/*simon/gtgk/Intro_GG.pdf,
s00445-016-1077-5 50p
Cáceres B (2016) Dramatic reduction of Cotopaxi glaciers Hickey J, Gottsmann J, Mothes P (2015) Estimating
during the last volcano awakening 2015–2016. Con- volcanic deformation source parameters with a finite
ference: AGU Fall Meeting, San Francisco-Moscone element inversion: The 2001–2002 unrest at Cotopaxi
Center. doi:10.13140/RG.2.2.17483.80168 volcano, Ecuador. Journal of Geophysical Research:
Chouet BA, Matoza RS (2013) A multi-decadal view of Solid Earth 120(3):1473–1486
seismic methods for detecting precursors of magma Hidalgo S, Bernard B, Battaglia B, Gaunt E, et al.
movement and eruption. J Volcanol Geoth Res (2016) Cotopaxi volcano’s unrest and eruptive activity
252:108–175 in 2015: mild awakening after 73 years of quiescence.
Christie R, Cooke O, Gottsmann J (2015) Fearing the Geophys Res Abstracts Vol. 18, EGU2016-5043-1,
knock on the door: critical security studies insights 2016 EGU General Assembly, Viena-Austria
into limited cooperation with disaster management Jousset P, Budi-Santoso A, Jolly AD, Boichu M, Dwiy-
regimes. J Appl Volcanol 4:19. doi:10.1186/s13617- ono S, Sumarti S, Hidayati S, Thierry P (2013) Signs
015-0037-7 of magma ascent in LP and VLP seismic events and
Espín Bedón P, Mothes P, Montesdeoca M, Ruiz M, link to degassing: an example from the 2010 explosive
Alvarado A, Córdova M, Santamaria S (2016) Com- eruption at Merapi volcano, Indonesia. J Volcanol
munication network of “Vigias” implemented at Geoth Res 261:171–192. doi:10.1016/j.jvolgeores.
Cotopaxi Volcano, Ecuador, Abstract, Cities on 2013.03.014
Volcanoes-9, Puerto Varas, Chile, November, 2016 Kumagai H, Yepes H, Vaca M, Caceres V, Nagai T,
Garrison JM, Davidson JP, Hall M, Mothes P (2011) Yokoe K, Vasconez F (2007) Enhancing
Geochemistry and petrology of the most recent volcano-monitoring capabilities in Ecuador. Eos 88
deposits from Cotopaxi Volcano, Northern Volcanic (23):245–252
Zone, Ecuador. J Petrol. doi:10.1093/petrology/ Kumagai H, Palacios P, Maeda T, Castillo DB, Nakano M
egr023 (2009) Seismic tracking of lahars using tremor signals.
Gaunt HE, Mothes P, Chadderton A, Lavallee Y (2015) J Volcanol Geoth Res 183(1):112–121
Physical characteristics of conduit plug rocks during Kumagai H, Nakano M, Maeda T, Yepes H, Palacios P,
Vulcanian eruptions at Cotopaxi and Tungurahua Ruiz M, Yamashima T (2010) Broadband seismic
volcanoes, Ecuador. Abstract, IUGG meeting, monitoring of active volcanoes using deterministic
Prague-Czech Republic, June 2015. and stochastic approaches. J Geophys Res: Solid Earth
Gaunt EH, Bernard B, Hidalgo S, Proaño A, Wright H, 115 (B8)
Mothes P, Criollo E, Kueppers U (2016) Juvenile Kumagai H, Mothes P, Ruiz M, Maeda Y (2015) An
magma recognition and eruptive dynamics inferred approach to source characterization of tremor signals
from the analysis of ash time series: the 2015 associated with eruptions and lahars. Earth, Planets,
reawakening of Cotopaxi volcano 328 October 2016. and Space. 67:178, 67:178, doi:10.1186/s40623-015-
doi:10.1016/j.jvolgeores.2016.10.013 0349-1
Hall M, Mothes P (2008) The rhyolitic–andesitic eruptive Lyons J, Segovia M, Ruiz M (2012) Reporte del
history of Cotopaxi volcano Ecuador. Bull Volcanol Enjambre Sísmico en la Zona de Volcán de Coto-
70(6):675–702 paxi—Mayo 2012. Internal report. Not published.
Hall ML, Steele A., Mothes P, Ruiz M (2013) Pyroclastic IGEPN, Quito.
density currents (PDC) of the 16–17 August 2006 Maeda Y, Kumagai H, Lacson RJ, Figueroa MS,
eruptions of Tungurahua volcano, Ecuador: geophys- Yamashina T, Ohkura T, Baloloy AV (2015) A
ical registry and characteristics. J Volcanol Geotherm phreatic explosion model inferred from a very long
Res 265(2013):78–93. https://doi.org/10.1016/j. period seismic event at Mayon Volcano, Philippines.
jvolgeores.2013.08.011 J Geophys Res Solid Earth 120:226–242. doi:10.1002/
Harlow DH, Power JA, Laguerta EP, Ambubuyog G, 2014JB011440
White RA, Hoblitt RP (1996) Precursory seismicity Márquez M (2012) Caracterización de los sismos de muy
and forecasting of the June 15, 1991, eruption of largo periodo en el volcán Cotopaxi y sus implica-
Mount Pinatubo. Fire and Mud: Eruptions and Lahars ciones. Tesis de Grado, Carrera de Geofísica, Facultad
Geophysical Footprints of Cotopaxi’s Unrest and Minor Eruptions … 269

de Ciencias Astronómicas y Geofísicas, Universidad Bibcode:1982JGR….87.1231 N. doi:10.1029/

Nacional de La Plata. Argentina JC087iC02p01231
Marzocchi W, Newhall CG, Woo G (2012) The scientific Nocquet JM, Villegas-Lanza JC, Chlieh M, Mothes PA,
management of volcanic crises. J Volcanol Geoth Res Rolandone F, Jarrin P, Martin X (2014) Motion of
247–248:181–189 continental slivers and creeping subduction in the
Molina I, Kumagai H, García-Aristizábal A, Nakano M, northern Andes. Nat Geosci 7(4):287–291
Mothes P (2008) Source process of very-long-period Phillipson G, Sobredelo R, Gottsmann J (2013) Global
events accompanying long-period signals at Cotopaxi volcanic unrest in the 21st century: an analysis of the
Volcano, Ecuador. J Volcanol Geoth Res 176(1):119– first decade. J Volcanol Geother Res 264:183–196.
133 doi:10.1016/j.jvolgeores.2013.08.004
Moran SC, Newhall C, Roman DC (2011) Failed Pistolesi M, Rosi M, Cioni R, Cashman KV, Rossotti A,
magmatic eruptions: Late-stage cessation of magma Aguilera E (2012) Physical volcanology of the post–
ascent. Bull Volc 73(2):115–122. doi:10.1007/ twelfth-century activity at Cotopaxi volcano, Ecuador:
s00445-010-0444-x behavior of an andesitic central volcano. Geol Soc Am
Mothes P, Hall ML, Espín, PA, Vasconez F, Sierra D, Bull 123(5–6):1193–1215
Córdova M, Santamaría S, Andrade D (2016a) Mapa Power JA, Stihler SD, Chouet BA, Haney MM, Ket-
Regional de las Amenazas del Volcán Cotopaxi- Norte ner DM (2012) Seismic observations of Redoubt
and Sur, scale 1:50,000, Instituto Geofísico and Volcano, Alaska—1989–2010 and a conceptual model
Instituto Geografico Militar, Quito http://www.igepn. of the Redoubt magmatic system. J Volcanol
edu.ec/mapas/mapas-volcan-cotopaxi.html Geotherm Res 259:31–44
Mothes PA, Nocquet JM, Jarrín Tamayo P, Morales Ruiz M, Guillier B, Chatelain JL, Yepes H, Hall M,
Rivera AM, Lundgren PR, Yépez YM, Viracucha EG, Ramon P (1998) Possible causes for the seismic
Gaunt EH (2016b) Geodetic signature associated with activity observed in Cotopaxi volcano Ecuador. Geo-
unrest at Cotopaxi in 2015–2016: modeling of GPS phys Res Lett 25(13):2305–2308
data, a deep magma source, synchronous seismic Siebert L, Simkin T, Kimberly P (2010) Volcanoes of the
swarms and petrologic constraints. Abstract, Cities on World. Univ. of California Press and Smithsonian
Volcanoes-9, Puerto Varas, Chile, November, 2016 Institute. 551 p
Mothes P, Hall ML, Andrade D, Yepes H, Pierson TC, Sparks RSJ, Aspinall WP (2004) Volcanic activity:
Ruiz AG, Samaniego P (2004) Character, stratigraphy frontiers and challenges in forecasting, prediction
and magnitude of historical lahars of Cotopaxi and risk assessment. The State of the Planet: Frontiers
volcano (Ecuador). Acta Vulcanol 16(1/2):1000–1023 and Challenges in Geophysics, Geophysical Mono-
Mothes P, Lisowski M, Ruiz M, Ruiz G, Palacios P graph 150, IUGG Volume 19
(2010) Borehole Tiltmeter and CGPS Response to Sparks RSJ, Young SR (2002) The eruption of Soufrière
VLP Seismic Events under Cotopaxi Volcano. Hills Volcano, Montserrat (1995–1999): overview of
Ecuador, Fall AGU, San Francisco Abstract scientific results Geological Society, London, Mem-
Mothes PA, Vallance JW, Hall ML, Garrison JM (2015a) oirs 21:45–69. doi:10.1144/GSL.MEM.2002.021.01.
Cotopaxi’s most recent rhyolitic eruptions, 2800 y 03
BP. In Conference Abstract, 26th IUGG General Stone J, Barclay J, Simmons P, Cole PD, Loughlin SC,
Assembly, Prague Ramón P, Mothes P (2014) Risk reduction through
Mothes PA, Yepes HA, Hall ML, Ramón PA, Steele AL, community-based monitoring: the vigías of Tungu-
Ruiz MC (2015b) The scientific–community interface rahua Ecuador. J Appl Volcanol Soc Volcanoes 3:11.
over the fifteen-year eruptive episode of Tungurahua doi:10.1186/s13617-014-0011-9
Volcano Ecuador. J Appl Volcanol 4(1):1–15 Tilling RI (1989) Volcanic hazards and their mitigation:
Mothes PA, Vallance JW (2015) Lahars at Cotopaxi and progress and problems. U.S. Geological Survey,
Tungurahua Volcanoes, Ecuador: highlights from Menlo Park, California, p p33
stratigraphy and observational records and related Valentine G, Graettinger AH, Sonder I (2014) Explosion
downstream hazards. In Volcanic Hazards, Risks and depths for phreatomagmatic eruptions. Geoph Res Lett
Disasters. doi:10.1016/B978-0-12-396453-3.00006-X, doi:10.1002/2014GL060096
chapter 6 White R, McCausland W (2016) Volcano-tectonic earth-
Newhall CG (2000) Volcano Warnings. In: Sigurdsson quakes: A new tool for estimating intrusive volumes
et al. (Eds) Encyclopedia of Volcanoes. San Diego, and forecasting eruptions. J Volcanol Geoth Res
CA, pp 1185–1198 309:139–155
Newhall, Christopher G, Self Stephen (1982) The Zobin, VM (2012) Long-Period and Very Long-Period
Volcanic Explosivity Index (VEI): An Estimate of Seismic Signals at Volcanoes. In: Zobin (Ed), VM
Explosive Magnitude for Historical Volcanism (PDF). Introduction to Volcanic Seismology (2nd ed).
J Geophys Res. 87 (C2): 1231–1238. pp 327–354
270 P.A. Mothes et al.

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter
the Creative Commons Attribution 4.0 International are included in the chapter’s Creative Commons license,
License (http://creativecommons.org/licenses/by/4.0/), unless indicated otherwise in a credit line to the material.
which permits use, sharing, adaptation, distribution and If material is not included in the chapter’s Creative
reproduction in any medium or format, as long as you Commons license and your intended use is not permitted
give appropriate credit to the original author(s) and the by statutory regulation or exceeds the permitted use, you
source, provide a link to the Creative Commons license will need to obtain permission directly from the copyright
and indicate if changes were made. holder.
 Volcanic Unrest Simulation Exercises:
Checklists and Guidance Notes

R. J. Bretton, S. Ciolli, C. Cristiani, J. Gottsmann,

R. Christie and W. Aspinall

Abstract both realistic and prudent to adopt three work-

When a volcano emerges from dormancy into a ing assumptions. The complex legal and
phase of unrest, the civil protection authorities administrative infrastructures of risk gover-
charged with managing societal risks have the nance will be largely untested and possibly
unenviable responsibility of making difficult inadequate. Many volcano observatory scien-
decisions balancing numerous competing soci- tists, and probably even more risk managers
etal, political and economic considerations. and at-risk individuals/communities, will have
A volcano that is threatening to erupt requires inadequate recent experience of the challenges
sound risk assessments incorporating trusted of hazard communication during a period of
hazard assessments that are timely, relevant and unrest. And lastly, the scientists may also have
comprehensible. Foreseeable challenges arise inadequate practical experience of the needs
when the inevitable uncertainties of hazard and management capacities of the
assessment and communication meet societal risk-mitigation decision makers with whom
and political demands for certitude. In some they must communicate. “Practice doesn’t
regions that host volcanic hazards, it would be make perfect. Practice reduces the imperfec-
tion.” (Beta 2011). If this statement is correct,
volcanic unrest simulation exercises (VUSE)
have a vital role to play within the complex
processes of volcanic risk governance. Consis-
tent with the broad approach of the Sendai
R. J. Bretton (&)
J. Gottsmann
W. Aspinall Framework for Risk Reduction 2015–30, this
School of Earth Sciences, University of Bristol,
chapter argues that practical knowledge of
Bristol, UK
e-mail: [email protected] VUSE can and should be analysed and recorded
so that key lessons can be shared for the widest
R. J. Bretton
J. Gottsmann
R. Christie
W. Aspinall possible benefit. This chapter investigates five
The Cabot Institute, University of Bristol, recent simulation exercises and presents six
Bristol, UK complementary checklists based upon data,
S. Ciolli
C. Cristiani insights and practice pointers derived from
Dipartimento della Protezione Civile (DPC), those exercises. The use of checklists, sup-
Rome, Italy
ported by guidance notes, is commended as a
R. Christie pragmatic way to create, test and develop
School of Sociology, Politics and International
acceptable standards of governance practice.
Studies, University of Bristol, Bristol, UK

Advs in Volcanology (2019) 271–298

DOI 10.1007/11157_2018_34
© The Author(s) 2018
Published Online: 14 July 2018
272 R. J. Bretton et al.

It is argued here that well planned and executed Framework and to share knowledge derived from
simulation exercises are capable of informing several recent simulation exercises with a wide
and motivating a wide range of risk governance audience.
stakeholders. They can identify process and
individual shortcomings that can be mitigated.
Simulation exercises can and should play a vital
role in reducing volcanic risks. 1.2 The Managerial and Scrutiny
Dimensions of Risk

Keywords 1.2.1 Managerial Dimension

Volcanic unrest Risk governance
When a volcano emerges from dormancy into a



Simulations Exercises Training
phase of unrest, the civil authorities in charge of
Communication managing volcanic risks have to make challeng-
ing decisions (Fiske 1984; Dohaney et al. 2015).
Decisions balancing safety and cost typically
must made with limited information (Sparks et al.
2012a, b; Marzocchi et al. 2012; Jenkins et al.
2012) and in real time and under uncertainty, in a
1 Introduction context of intense pressure (Marrero et al. 2015,
2). While the primary objective is to minimise the
1.1 Simulation Exercises loss and damage from any volcanic event, the
and the Sendai Framework socio-economic losses resulting from false alarms
and evacuations must also be considered” (Woo
This chapter is about reconstructions of the 2008; Hincks et al. 2014, 2; Donovan and
evolution of past, and realistic simulations of Oppenheimer 2014).
hypothetical future, volcanic unrest events. They Poorly handled unrest periods cause social,
will be referred to as Volcanic Unrest Simulation economic and political problems, even without
Exercises (VUSE). A simulation is “a learning an eruption. Ill-considered responses may facili-
experience that occurs within an imaginary or tate the release of inappropriate advice and
virtual system or world” and involves role-play, emergency declarations, and may lead to
which has been defined as “the importance and unwarranted media speculation and the prema-
interactivity of roles in pre-defined scenarios” ture cessation of economic activity and commu-
(van Ments 1999; Errington 1997, 2011; Doha- nity services (Johnston et al. 2002, 228).
ney et al. 2015). A fuller working description is Recent crises, including the 2010 Icelandic
set out in Sect. 3.1. Eyjafjallajökull eruption, have highlighted the
The Sendai Framework prioritises mitigation difficulty of co-ordinating and synthesising sci-
of risks before response and recovery. It also entific inputs from many different disciplines and
recognises the importance of “building the institutions, and translating these into useful
knowledge of, inter alia, government officials… policy advice at very short notice (Harris et al.
through sharing experiences, lessons learned, 2012; Dohaney et al. 2015). Effective commu-
good practices, and training and education on nication, collaboration and cooperation are nec-
disaster risk reduction” (UN/ISDR 2015, 15). essary between many expert and technical
Mutual learning, dialogues and cooperation advisors, emergency management agencies and
between risk governance stakeholders are lifeline organisations (Doyle et al. 2015).
encouraged. The Sendai Framework also identi- Jordan et al. (2011) have emphasised the need
fies a need for quality standards. for scientists to have a clear role, a clear and
The principal purpose of this chapter is to authoritative voice and effective communication
accept the challenge of the goals of the Sendai skills. The Organisation for Economic
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 273

Co-operation and Development (OECD) (2015) of the Sendai Framework; and (4) the require-
has identified the need for scientific advisers to ments of the IFRC/UNDP legal checklist.
have: (1) permanent authoritative structures; (2) a
central contact point; (3) clear reporting proce-
dures; (4) a pre-defined public communication 1.3 Academic Support for Training
strategy; and, when necessary, (5) ways to and Simulation Exercises
coordinate their actions internationally.
Although civil protection “authorities may The experience and levels of expertise of obser-
have theoretical knowledge of volcanos, few have vatory scientists are critical to making accurate
any practical experience of eruptions” (Solana forecasts and training is important (McGuire and
et al. 2008, 312). Furthermore, the timescales of Kilburn 1997). In the context of volcanic risks,
periods of volcanic unrest, especially bigger ones, Doyle et al. (2015) undertook a review of the
do not correlate well with the tenures of political literature on emergency management team
and senior management appointments (Donovan response, decision making, mental models and
and Oppenheimer 2012; Mothes et al. 2015). situation awareness and exercising. They argue
The term ‘standard equivocality’ relates to the that science agencies and science advisory
“absence of commonly recognised standards groups must embark on a suite of training
capable of guiding, measuring and evolving activities to enhance their response during a
acceptable practice” (Hood 1986; Bretton 2014; disaster. These should include exercise and
Bretton et al. 2015; Rothstein 2002). It is sug- simulation programmes within their own organ-
gested here that there are no readily accessible isations rather than participation solely as exter-
standards regarding how hazard communication nal players in emergency management activities.
should be conducted during a period of volcanic Structures, resources and time must be provided
unrest. for these programmes.
It is further argued by Doyle et al. (2015) that
1.2.2 Scrutiny Dimension training will enhance the future response capa-
The International Federation of Red Cross and bilities of both scientists and risk-mitigation
Red Crescent Societies (IFRC) and United agencies in several ways.
Nations Development Programme (2015)
(IFRC/UNDP) legal checklist requires national
laws to establish and promote the training of 2 Methodology
public officials and relevant professionals.
There is an emerging international law duty This chapter reviews five simulation exercises
upon sovereign states to have substantive regu- (four of which were conducted as part of the
latory systems to ensure that risks from natural VUELCO project), which for ease of reference
hazards are mitigated so that they do not are summarised in Table 1. It investigates the
endanger human lives. States must inform at-risk ways in which these exercises were planned and
communities of the potential of unmitigated risks undertaken. By design, it does not address in any
and establish sufficient co-ordination and coop- detail: (1) the features of the various and varying
eration between administrative authorities. This volcanic settings in which the exercises were
is an onerous duty and it is argued here that it conducted; or (2) the nature, scope or analysis of
would be prudent to assume that the general duty the monitoring data that were painstakingly cre-
includes several more specific subsidiary duties. ated for the volcanic hazard scenarios that
One would be to consider the merits of simula- underpinned them.
tion exercises as a way of satisfying: (1) the This chapter draws from ethnographic obser-
specific requirements of national laws; (2) the vational data (recorded in hand written field
general expectations of international law; (3) the notes) collected during five simulation exercises
education, training and knowledge-sharing goals from the perspectives of overt non-participant
 274

Table 1 Brief details of the analysed five volcanic unrest simulation exercises
Name Hosting organisations Volcanic hazard scenarios and interpretation of observables
country/when
Colima VUELCO Volcan de Colima—evidence of moderate effusive activity alternating with moderate explosive events; escalating unrest, dome
Mexico (Colima) growth, some explosions, a 2 km + eruptive column, ash-fall, possible evolutions included: effusive dome growth alone, a large
17–23/11/2012 explosive eruption after dome collapse, partial flank collapse, a mixed ‘Merapi type’ event and cessation of dome growth;
culmination in risk sufficient to require the evacuation of a single small town—La Becerrera
Campi VUELCO Campi Flegrei—conflicting evidence of either a rise of magma from about 7 km or an increase in shallow hydrothermal activity;
Flegrei Italy (Rome) further evolution involving a possible 3 km sill or continued shallow fluid migration; further rapid evolution raised the
11–12/02/2014 possibility of phreatic explosions and small volume magmatic eruptions in the eastern sector; further dramatic evolution made
explosions and eruptions likely within days or weeks in the Bagnoli-Solfatara area and small offshore eruptions were not
excluded; culmination in eruption between Bagnoli and Monte Spina and a sustained eruptive column, ash fall and pyroclastic
flows south and east
Mount Presidencia del Gobierno Teide-Pico Viejo—evidence of rumbling noises, small debris avalanches and no seismic/deformation changes; further evolution
Teide de Canarias and Instituto with more rumbling noises, small swarm of seismic events; further evolution with swarm of small high-frequency VT, an
Geográfico Nacional explosive event, new thermal events and ash emissions with LFs and tremors; further evolution with pronounced SO2 decrease
(IGN), Ministerio de interpreted as near-surface sealing of rising magma with high probability of eruption; culmination in eruption
Fomento, Gobierno de
España
Tenerife, Spain (Santa
Cruz de Tenerife)
25/04/2014
Cotopaxi VUELCO Cotopaxi—evidence of anomalous seismic activity with returns to background levels and no physical-chemical changes;
Ecuador (Quito) possible local tectonic activity, fluctuating gas emissions, some deformation, further evolution with many long-period events,
13/11/2014 increasing released energy, sub-vertical fault in NE flank, increasing deformation, high gas emissions, increased fumarolic
activity, increased thaw in upper parts of the edifice; further evolution to sporadic vulcanian explosions, eruptive columns
<5/8 km plus tephra; further evolution with escalating seismic activity, inflation and degassing changes; culmination in eruption
Dominica VUELCO Southern volcanic region including Trois Piton—evidence of elevated seismicity, landslides and small hydrothermal changes;
Dominica (Roseau) further evolution with many VT events and landslides, a return to hydrothermal background levels plus Boiling Lake drainage;
14–15/05/2015 further evolution with more VT events and hybrids followed by VT decrease with hybrids dominating, continuing geochemical
background levels, no recent deformation and contraction in some areas, indications of a deeper source than in the early unrest
phases, possible pressurisation of deep seated magmatic system or hydrothermal activity; culminated in the consideration of a
lowering of the Alert level
R. J. Bretton et al.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 275

observers and, to a very limited but recorded infrastructures, policies, procedures and people is
extent, of participant-observers. Four exercises tested.
relate to VUELCO volcano unrest simulations in The VUELCO exercises involved the provi-
Mexico, Italy, Ecuador and Dominica. Detailed sion of raw or partially analysed monitoring data
analysis of the documents prepared before, dur- and information to geo-scientists. Their role was
ing and after these exercises was undertaken. The to undertake a scientific analysis of the data with
challenges of avoiding researcher bias and the purpose of passing on characterisations of
unintended observer effects were recognised. likely volcanic hazard scenarios, by means of
Additional observational and documentary hazard communications, to risk mitigation
data were acquired by the lead author during the decision-makers (civil protection authorities and
Tenerife (Spain) exercise as an invited external the representatives of at-risk individuals) and the
non-participant observer commissioned to pre- mass media. The civil protection authorities,
pare a post-exercise evaluation at the request of having considered not only societal issues but
the Presidencia del Gobierno de Canarias. No also political, economic and other values, made
ethical agreement was signed before or during and communicated risk mitigation decisions. In
that exercise. Data from that exercise and extracts some exercises, representatives of at-risk com-
from the post-exercise report are included by munities, emergency services and relevant gov-
kind written permission of the Presidencia del ernment entities played active roles in testing
Gobierno de Canarias. emergency, law and order, rescue, medical and
Within the main body of this chapter, no evacuation procedures.
attempt is made to provide complete details of
the five exercises listed in Table 1. Essential
background information and additional reading 3.2 Checklists and Guidance Notes
sources are to be found in additional files A–E
which, for ease of reference, have deposited at 3.2.1 Checklists
the ‘Collaborative volcano research and risk The checklists within this chapter are presented
mitigation’ (VHUB) website under reference in response to the perceived importance of such
“Volcanic Unrest Simulation Exercises: check- documents as sources of good practice in the
lists and guidance notes—Additional files A-E”. eyes of recent commentators (e.g. Gawande
The additional reading sources are also listed at 2010; Newhall 2010; IAVCEI 2015). For
the end of the references for this chapter. example, within the ‘Safe Surgery Saves Lives’
project, which started in 2007, the World Health
Organisation (WHO) introduced a Surgical
3 Background Safety Checklist developed by Dr. Atul Gawande
based, inter alia, upon the success of pre-flight
3.1 VUELCO Themes and Goals checklists in enhancing safety within the aviation
industry. “A systematic review and meta-analysis
As VUSE are purpose-driven learning activities, of the effect of the WHO checklist” strongly
each will have tailored goals and an overall suggests a related “reduction in post-operative
design based upon the needs of its participants. complications and mortality” (Bergs et al. 2014,
Consistent with VUELCO’s stated goals, the 150; Treadwell et al. 2014).
principal purpose of each VUELCO exercise was Drawing upon the research of Gawande
to present a realistic simulation of the evolution (2010), this chapter’s checklists aim to incorpo-
of past and hypothetical future volcanic unrest rate several critical features that are output and
events associated with a host volcano or wider outcome-focussed. As far as reasonably practi-
volcanic setting. In the manner of a film set, a cable, the authors’ aim has been to ensure that
simulated volcanic event is the dynamic back- each checklist is: (1) concise and preferably short
drop against which a selection of risk governance as well; (2) simple, precise and unambiguous;
276 R. J. Bretton et al.

(3) targeted by addressing only evidence-based identification of key laws, policies, procedures
priorities that are considered either critical or and people.
significant to risk governance; and The ‘Volcano team’ has the very difficult
(4) non-prescriptive and non-comprehensive. obligation to craft and deliver monitoring data
Checklists may encourage rational, systematic consistent with realistic volcanic hazard scenar-
(i.e. both consistent and complete), routine and ios. The data must be suitable to meet the hazard
transparent practices whilst recognising the analysis challenges confronted by the experts
importance of, and encouraging careful attention within the ‘Scientific Advisory Committee’
to, a wide variety of constraints and expectations. (SAC). It is noted here that, in some situations,
However, it is not envisaged that they should be the roles of monitoring and analysis may be
used as a regulatory device, an enforceable legal undertaken by the same team or with substantial
requirement, or part of a blame-avoidance strat- personnel overlaps. This chapter retains the
egy (Hood 2011; OECD 2015). separation to cover circumstances where there
The checklists presented in this chapter are distinct group remits, functions and
address the general planning, funding and exe- responsibilities.
cution of purposeful simulation exercises. The The SAC receives and handles the data pro-
exercises also highlighted the many challenges of vided utilising, as appropriate, a variety of soft
hazard communication and the difficulties that skills (including those of analysis, deliberation
may result if the needs of other stakeholders are and communication), tools (including expert
not identified and responded to. elicitation and probabilistic models) and estab-
This chapter presents six checklists namely: lished procedures and protocols.
(1) Planning; (2) Logistics; (3) The Volcano The ‘Risk Managers - Civil Protection
team; (4) The Scientific Advisory Committee Authorities’ (CPA) make risk mitigation deci-
(SAC); (5) The Risk Managers - Civil Protection sions based in part upon scientific communica-
Authorities (CPA); and (6) The Observers/ tions received from the SAC or, where the
Auditors. The first two relate to exercise ‘pha- existence of a SAC is not foreseen within the
ses’ and the other four relate to ‘roles’. It is not relevant governance system, the Volcano team.
suggested that these checklists will be relevant to An important part of the CPA role is usually the
all simulation exercises, however, they will pro- challenging duty, not only to advise individuals
vide an indication of some of the critical phases and entities driven by political values, but also to
and major roles that must be considered. interact with members of the public and respond
The checklists inevitably reflect the fact that to mass and social media demands. The press and
VUELCO’s exercises placed particular emphasis social media “can play an important role in the
on the scientific analysis of monitoring data, the dissemination of information, true or false” and
communication of hazard characterisations to social media can rapidly ferment public, anxiety,
risk managers, and interactions between scien- distrust or dissent (OECD 2015, 37).
tists, risk managers, the media and the general Simulation exercises are learning/training
public. It is accepted that future exercises may exercises and, accordingly, the importance of
have other goals for which more or different ‘Observers’ should not be underestimated. If
checklists may be helpful. properly briefed, observers and auditors will not
The ‘Planning’ checklist encourages the only contribute their own candid views about all
identification of goals and objectives, and the aspects of the exercise but also encourage other
need for clear leadership, careful design and participants to be reflective about their own
realistic financing. contributions and actions.
The ‘Logistics’ checklist is probably the most
important. The actual task of arranging the 3.2.2 Guidance Notes
‘planned’ exercise may have many benefits in its To support and supplement the checklists, it is
own right as it will require the careful suggested here that guidance notes should be
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 277

issued from time to time to provide a dynamic to be in charge of co-ordination and leadership of
and helpful knowledge and innovation resource. the various preparatory activities”.
Their aim should be: (1) to gather together, After one VUELCO exercise it was suggested
record and share the accumulated experience of that at least “a full-time scientist” should be
other practitioners in relevant fields of expertise; dedicated to preparing any exercise that focusses
and (2) to suggest ways to find optimal solutions on scientific analysis.
to the most critical issues.
It is readily accepted that the observations Purpose and Goals
within the guidance notes are inevitably subjec- What is the overall purpose of the VUSE?
tive. They are not intended to be in any respect What are the short and long term goals of
either comprehensive or prescriptive. They are the exercise and its players?
presented as options to be considered along with A VUSE is a learning activity and the prin-
other issues that will be found within best prac- cipal reasons for it must be identified and stated.
tice guidelines such as those identified in Doyle It should respond to the perceived core needs of
et al. (2015). the participating stakeholders and, like any
learning activity, be carefully planned with stated
assumptions, aims, objectives and themes.
4 Checklists (in bold italics) Before the exercise, each and every partici-
and Guidance Notes (in normal pant should know what they will learn from the
font) exercise. Thought should also be given to how
and by whom the success of the exercise will be
4.1 Planning evaluated and what will be done by whom to
build upon the exercise.
Leadership
The assumptions, aims, objectives and themes
Who (individual and entity) has overall
should be set out in the pre-exercise briefing note
responsibility for planning and delivering the
referred to in Sect. 4.2 of this checklist.
exercise?
The choice of exercise leaders should be dic-
tated by the planned goals and activities of the Scope
exercise. If a significant involvement of civil Which parts of the risk governance process
protection authorities is planned (for example, are going to be tested?
where risk mitigation decisions are to be made A VUSE should be focussed, purpose-driven
and/or related mitigation measures are to be tested) and planned accordingly.
those authorities should probably take the lead. No VUSE can realistically attempt to replicate
The VUELCO exercises have suggested that all aspects and phases of, and all stakeholders
overall responsibility for a VUSE should be interested in, a societal risk governance regime. It
assigned to just one person within the risk gov- is possible that more regular exercises, which
ernance stakeholder (e.g. the civil protection concentrate upon very carefully defined aspects
authority) likely to gain most from it. That per- of the system (e.g. communicating in real time
son, who will need suitable and sufficient support with the general public and the mass media), may
from a working team or steering group, should be more cost-effective and beneficial.
have the gravitas, personality, authority, experi- Consideration might be given to those issues
ence and resources (both human and financial) and functions that require the most
needed to plan and handle a complex high profile inter-stakeholder planning, cooperation and col-
project that will inevitably attract political, soci- laboration and accordingly excellent means of
etal and media attention. communication. A VUSE can and should be a
The MIAVITA Handbook (2012, 118) sug- learning exercise in communication between
gest that it is good practice for “a steering group many key players such as:
278 R. J. Bretton et al.

• Geoscientists and other geoscientists. identified eleven functions—”Public Information

• Geoscientists and risk governance advisers Management, Governance, National Financial
(e.g. weather forecasters, aviation and marine System, Logistics and Other Support
space managers, communication specialists Co-ordination, International Assistance and
etc.). Liaison, Rescue, Health, Welfare, Building
• Geoscientists and other stakeholders (e.g. Safety, Restoration of Access and Restoration of
individuals within civil protection authorities, Lifelines” (NZ/MCDEM 2008).
interested and affected members of the public, Another possible example comes from the
representatives of community, religious, pub- “Metodo Augustus”, the organisation of emer-
lic utility and commercial interests, and repre- gency management, used at the Italian Depart-
sentatives of international, national, regional, ment of Civil Protection. That organisation
municipal and other levels of government). provided several support functions including
• Local geoscientists (e.g. volcano observatory technical-scientific, health and veterinary assis-
staff) and visiting scientists (e.g. academics tance, mass-media and information, volunteers,
contributing to a scientific advisory commit- means and materials, transportations and mobil-
tee) and visiting researchers. ity, telecommunications, essential services,
• Hazard analysts (e.g. local and visiting geo- damage assessment, operational structures, local
scientists) and experts in risk assessment and administrations, dangerous materials, people
management. assistance, cultural heritage and coordination
• All scientific analysts/risk decision makers and (Galanti et al. 2006).
(1) the general public and (2) the mass media. VUELCO chose a science-orientated focus for
its four exercises concentrating upon the inter-
In 1999 several of these interactions were faces between:
addressed by IAVCEI’s Subcommittee for Crisis
Protocols, which issued a report entitled “Pro- • Hazard monitoring and hazard assessment:
fessional conduct of scientists during volcanic • Long-term monitoring
crises” (Newhall et al. 1999). In the context of • The host volcano’s main precursors of
the governance of the risks of volcanic hazards, volcanic unrest.
this paper represents a rare, if not unique, • The host volcano’s short-term monitoring
example of an attempt to offer authoritative resources (e.g. equipment, employed staff
practice guidance based on past events and an and volunteers etc.), the nature, adequacy
extensive literature review. It was concerned and timing of the data output, and their
principally with personal and institutional inter- capacity to respond to changing demands.
actions during a volcanic crisis. • Hazard assessment and risk assessment
Consideration should also be given to whole • As the period of unrest evolves: (1) real
or part of an exercise being in ‘degraded mode’. time characterisations of the host vol-
The MIAVITA Handbook (2012, 118) suggests cano’s possible and most likely hazards
that exercises should be planned to test system and their temporal, physical and spatial
level capabilities of response when some parts of parameters; and (2) other advice e.g. the
the system are not fully operative. By this means, merits/safety of further/different
depending upon which aspects are being tested, short-term monitoring.
allowance can be made for, inter alia, holidays, • The communication of scientific analysis
the malfunctioning or destruction of monitoring (with its inherent assumptions, limitations,
or telecommunication equipment, blocked escape complexities and uncertainties) to a variety of
or rescue routes, and weather conditions. stakeholders each possibly having different
By way of illustration of needs that an exer- requirements and expectations and related
cise might seek to target, Exercise Capital Quake communication challenges.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 279

If an exercise focusses predominantly upon Care must be exercised if one of the hazard
scientific issues, it may be very difficult to scenarios will result in the need for the evacua-
engage participants from civil protections agen- tion of representatives of at least one vulnerable
cies and more ‘risk’ related functions. community. Of course, this will have to be
A VUSE can be a good opportunity to test the organised and announced well in advance of the
processes by which scientific analysis is inte- start of the exercise. This may affect (i.e. make
grated into risk assessment agencies, such as civil rather implausible) other parts of the exercise
protection authorities, that do not have embedded which may depend upon realistic and continuing
scientists. uncertainty as to the future evolution of the initial
What is the proposed active participant scope? hazard.
In a VUSE, if time and resources permit, a What is the proposed passive participant (e.g.
wide range of stakeholders can be represented and observers and monitors) scope?
actively involved including, but not limited to: In the VUELCO exercises, very valuable
contributions were made by monitors and
• the host volcano as the source of scientific and observers. They can be from the participating
visual data, which are provided sequentially in organisations or entirely ‘independent’. Monitors
discrete pre-planned “phase” briefings/reports and observers (particularly with experience of
(hereinafter called the Volcano team); previous exercises and the practices of countries
• host volcano observatory scientists (the VOS); other than the host country) can play a critical
• local and external scientists (collectively called role in providing both hot and cold feedback
the scientific advisory committee or SAC); whilst also gaining invaluable personal experi-
• risk assessors and managers possibly from a ence to assist in the planning of future exercises.
range of national, federal, state, regional and VUSE are invariably observed by local and
municipal tiers of government (hereinafter visiting students and early career academics.
called civil protection authorities or CPA); With careful planning, it may be possible for less
• the media; experienced scientists and civil protection
• interested and affected members of the public. authority decision makers to be involved directly
(as happened in the Campi Flegrei VUSE) or
In a more sophisticated exercise, it may be indirectly in the ‘shadow’ SAC and CPA teams.
worthwhile ensuring that a “dissenting/minority Reference should be made to the separate
view” and/or “maverick” scientists are repre- Observers/Auditors checklist in Sect. 4.6.
sented to test robustly relevant democratic and What is the proposed geographical scope?
communication processes. The geographic scope of a VUSE should be
The MIAVITA Handbook (2012, 114–116) considered carefully. A balance must be achieved
details the benefits of informing, sharing and between the spatial parameters that produce a
training involving: (1) national, regional and risk exposure area that is realistic in the context
local authorities; (2) scientists; (3) volunteers; of the host volcano and those that create an area
(4) the media; (5) pupils and students; and (6) the that is too large in terms of the duration of the
public. exercise, data production and/or the roles of the
The VUELCO Colima exercise incorporated active participants.
the planned evacuation of an at-risk community. What is the proposed administrative scope?
A VUSE can usefully focus upon the real time National laws tend to identify, authorise and
implementation of the risk mitigation strategies fund risk governance bodies (e.g. government
that result from the dynamic characterisation of departments and agencies, public corporations)
an evolving unrest. However, the time and and public officials (e.g. individuals such as
resources required to identify and engage multi- governors, mayors, prefects and village heads)
ple stakeholders, including members of the within a coherent legal and administrative
public, should not be underestimated. framework—a risk governance infrastructure.
280 R. J. Bretton et al.

These laws often use and build upon existing What type of exercise will be undertaken and
entities with existing administrative frameworks why?
that have multi-level national, regional, district, There are many types of management exer-
municipal etc. political divisions and subdivi- cises including: (1) full-scale; (2) reduced;
sions (Bretton et al. 2015). (3) orientation; (4) drill; (5) table-top/discussion;
The MIAVITA Handbook (2012, 118) sug- and (6) functional (see Doyle et al. 2015).
gests that it is good practice for an exercise to “be Exercises can be also announced or unan-
based on the regulations and laws of the [host] nounced depending upon their objectives. It has
country”. been suggested that the use of unannounced
The administrative scope of a VUSE should exercises is necessary to verify the real strength of
be considered carefully and it is likely this scope systems and levels of preparedness, when people
will be related to those parts of the risk gover- do not expect them (MIAVITA 2012, 116–117)
nance process selected for review and testing. It
may be very difficult to involve actively all levels How frequently should exercises be held?
of governance and therefore a decision will have The MIAVITA Handbook (2012, 116) notes
to be made as which levels will participate and that exercises are fundamental for testing existing
which will merely observe. procedures, plans, and preparedness and for
It is likely that a good starting point will be a maintaining the attention of stakeholders “on the
complete flow diagram of the existing societal spot”. It also argues that exercises should be
risk governance infrastructure for the host vol- scheduled frequently with the frequency
canic region. This can then be annotated to depending upon several factors including: (1) the
indicate which national, regional, local bodies behaviour of the host volcano; (2) the
and individuals will participate and their social-economic context; (3) the levels of risk
respective roles (both active and passive) in the perception; and (4) the democratic trend.
exercise. Finance
For the SAC and CPA at least, consider set- What are the planned budgets for all phases
ting out clearly its respective: of the exercise including:
• legal status;
• legal remit and reporting lines; and • Before?
• the rights, responsibilities and liabilities of its Consider planning, field trips, printed field
members. guides, data and scenario creation, documenta-
tion etc.
Duration and type
How long will the exercise last and why? • During?
This is a very important issue with funding, Consider venue, catering, data/scenario com-
resources, logistical and many other munication etc.
implications.
VUSE vary greatly in length as shown by the • After?
data in Table 1. Much will depend upon the Consider feedback (hot and cold) and follow
critical decisions that must be made about scop- up meetings, reports, presentations, interviews,
ing—process, participants, geographical area and actions.
administrative levels.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 281

Who is funding: of analytical/diagnostic confidence, dealing with

communication network failures, using social
• The exercise? media etc.), it might make sense to mention these
• The participants, including invited guests, in advance to encourage pre-exercise thought and
experts and invited observers/auditors (in- preparation.
cluding travel, accommodation, field trips,
other expenses etc.)? • Models, methods and protocols to be
used/made available for use
Consider whether there should be an under-
writer of last resort for any unplanned deficiencies. The VUELCO exercises have proved that
VUSE can be used successfully to test in simu-
Do the funders have any demands, expectations lated real time the merits of probabilistic methods
etc. as to any aspects of the exercise? and tools such as BET_EF (Bayesian Event Tree
If yes, these must be understood by the for Eruption Forecasting), HASSET (Hazard
organiser, planned and budgeted for. Assessment Event Tree) and QVAST (Quanti-
fying Volcanic Susceptibility) (Sandri et al.
2009; Sobradelo and Bartolini 2014; Bartolini
4.2 Logistics et al. 2013).
In advance of the exercise, there must be
How long will the planning stage take? communication between the modellers and the
It would be easy to underestimate the time that Volcano team. This should ensure that all
will be needed for, and the complexity of, the appropriate monitoring data (perhaps deliberately
planning stage upon which the whole success of not all the data) is available for all relevant
the exercise will depend. It may be prudent to phases of the exercise scenario including periods
speak to the organisers of other VUSE to discuss between phases.
this and other financial/practical issues. Reference Feedback from the VUELCO exercises sug-
could also be made to Dohaney et al. (2015). gests that input from the modellers during each
The MIAVITA Handbook (2012, 118) notes step of an exercise may be welcomed by other
that planning an announced exercise takes time participants. Testing simultaneous integration
and states “six to twelve months are necessary to was a specific VUELCO goal. However, if this is
prepare a full-scale exercise. If the promoted to be done, it must be carefully planned and
exercises are repeated on a fixed schedule, three sufficient time allocated within the timetable.
can be sufficient”. We consider these to be the It is important to state in advance of the
minimum planning periods. exercise, for example in the briefing documents,
Will every participant/observer receive in what role the models, tools etc. will play during
advance a full briefing note covering: the exercise. Consideration should be given to
whether the results will be available in real time
• Purpose to the participating SACs and CPAs. If yes, how
Refer to the Sect. 4.1 regarding the overall and when will the results be communicated? If
purpose of the training exercise. no, when will feed back regarding the utility of
• Scope the items tested be communicated?
Refer to the Sect. 4.1 regarding scoping. VUSE can also be used to give SAC’s the
Issues covered may include processes, partici- opportunity to use formal expert elicitation
pants, geographical, administrative etc. (EE) methods. EE may provide formality and
• Themes direction to the deliberation of monitoring data,
If some important themes have been identified assist the framing of likely hazard scenarios, and
for particular attention (e.g. expressions (nu- facilitate the drafting of expressions of temporal
merical and narrative) of likelihood, expressions certainty and analytical/diagnostic confidence.
282 R. J. Bretton et al.

EE were conducted in the VUELCO exercises in necessary to ask representatives of several dif-
Colima and Campi Flegrei. ferent organisations to work together. By way of
The Somma Vesuvio MESIMEX exercise illustration, this chapter refers to four teams
undertaken in 2006 was used to conduct a before (Volcano, SAC, CPA and Observers) but others
and after volcanic risk perception survey (Ricci can and should be used as dictated by the specific
et al. 2013). goals of the planned exercise.
Consideration should also be given to the use
• Players (participants, observers etc.):
of a technical team to support the SAC’s delib-
– by name and organisation
erations in respect to issues such as real time
– by role in the exercise
GPS mapping, model trials and expert elicita-
tions. The time needed for these complex matters
It is very likely that a large number of people
must be considered particularly during short
will be involved from a wide range of organisa-
exercises. In VUELCO’s Campi Flegrei simula-
tions and, perhaps, several countries. It would be
tion, a technical team composed of civil protec-
prudent to assume that nobody will be known to
tion personnel operated continuously with the
all of the other participants. Consider the use of
aim of supporting the SAC’s simulated
name badges stating Name, Organisation, and
‘real-time’ deliberations. The team’s results were
Exercise Role. As found during the Campi Fle-
only presented at the end of the exercise because
grei exercise, colour coding for teams and
they lagged behind the ‘real time’ evolution of
active/passive roles might also be helpful. The
the phases of the exercise’s hazard scenarios.
scheme adopted should be explained carefully to
save time during the start of exercise when • Venue (full address, plan, transport and
valuable time can be lost easily. contact details)
By way of illustration, more than 100 partic- – Sub-venues for all aspects of the exercise
ipants from over 10 European and Latin Ameri- (including any evening events) with a plan
can countries attended the VUELCO Campi – Layout for the plenary sessions
Flegrei exercise. The respective figures for the – Layout for any breakout sessions
VUELCO Cotopaxi exercise were about 50
participants from 13 countries. These critical issues should not be over-
After the Dominica exercise it was suggested looked. The exercise organiser should decide in
that, at the very start of the exercise, introductions advance, in respect for each phase of the exer-
would have been helpful “to establish who each cise, which ‘team’ will sit where and why,
person was and what their role was…It was probably by reference to which other participants
challenging accepting information from a person they will have to communicate with. For the sake
whom we did not know or have any clue as to of intended realism, consider whether they
their background or capabilities. Also the scientist should be near or far apart and by what means
[s] must be aware of who the[y] are working with they will communicate.
in order to frame their advice appropriately”. Consider whether, in a real emergency, par-
For numerous purposes (before, during and ticipants would speak to each other face to face.
after the exercise) a comprehensive, up-to-date Consider whether it would be more realistic to
and accurate email list containing all participants separate some teams physically in order to force
will be necessary. Inadequate email lists caused the use of video conferences, emails and phone
difficulties during several VUELCO exercises. calls in team interactions. At the planning stage it
will be necessary to allow more time for this
• Teams nuance, which may highlight the challenges of
Consider how many teams will be required long-distance communication, data sharing, data
and how and when their members will be allo- analysis, collaboration, consensus building and
cated. For the sake of simplicity, it may be decision-making.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 283

Feedback from the VUELCO exercises – Greetings, Introductions, Reviews, Thank

specifically mentioned the problems associated you’s etc.
with having too many people in the same room – Presentations to or by representatives of
and failing to plan in advance who should be central, local government, etc.
where and near to whom.
It is worth considering that each refreshment Predicting accurate timings for these two
break will take longer than expected unless components is notoriously difficult and careful
refreshments are either provided in all rooms or planning in advance and strict control on the day
available at all times. Will working lunches will be necessary. Consider carefully whether
simulate the difficulties of a tense period of these are really necessary. Consider whether it
emerging unrest? For a short period of intense would be better for relevant entities to be given
training, is a degree of inconvenience and/or formal active or passive roles within the exercise
discomfort in fact to be encouraged? itself and be appropriately involved in that way.

• Timetable – Dissemination of new data

– Dates for each part of the exercise and
timings for each day addressing: Consider when, by whom, to whom and how.
– Registration (including perhaps the handing During the VUELCO exercises, difficulties arose
out of name badges in accordance with the due to format incompatibilities and internet
stated scheme) access issues.
This will need very careful planning and
resourcing to avoid a late start on Day 1. – Deliberation of new data
– Start, finish
Start and finish promptly. The time necessary for deliberation may
– Food/drink breaks (if needed!) depend upon the amount/ format of the data, the
need to communicate with other participants,
Refer to the above guidance about the need expectations about output deliverables (e.g.
for these in an exercise that attempts to simulate written reports) etc. As a general rule, allow
real time challenges. plenty of time within the context of the realism
that the exercise seeks to simulate.
– Evening events, banquets etc.
– Dissemination of ‘injects’
Consider how long it will take participants to
return to their accommodation before evening Consideration should be given to the timing and
events. Give clear advice about the dress code, if drafting of ‘inputs’ (for example, requests for local,
any, and other cultural expectations (e.g. the use national and international media interviews or
of cameras or the need for modest clothes and briefings with concerned public officials, unwel-
head coverings). come social media exchanges, etc.), which will be
distributed to the participants at planned stages
– Venue tours during the exercise. Participants might be asked to
think about how they would respond and who
It is likely that many of the participants will would be delegated to meet these urgent requests.
expect/request a tour if the exercise takes place in
a major risk management/communication centre. – Outputs from the SAC and other teams
Allow plenty of time for this and consider mak-
ing these visits part of the pre-exercise field trip As a general rule, allow plenty of time for the
to avoid wasting time when all of the participants dissemination of outputs within the context of the
(and probably the media) are present. realism that the exercise seeks to simulate.
284 R. J. Bretton et al.

Will information about current CPA mitiga- reflect the governance infrastructures, roles and
tion actions, risk alert levels etc. be provided at duties required by:
any time during the exercise?
– the national laws of the host country as
If yes, why, what, when and to whom?
encouraged by the UN Sendai Framework
Feedback from the VUELCO exercises sug-
and the IFRC legal checklist;
gests that scientists benefit greatly from
– any regional emergency management or
improved knowledge of civil protection authority
response arrangements, such as those in the
systems and in particular better appreciation of
Caribbean involving the Caribbean Disaster
those hazard parameters that are most relevant to
Emergency Agency (CDEMA), which are set
risk mitigation decisions.
out in inter-country agreements, memoranda
• Factual background and related resources of understanding and protocols; and
– any applicable international law standards.
Consider whether a field trip is necessary
before Day 1 of the exercise so that all of the
• Information Technology (IT) including
participants can gain a basic grasp of the topog-
disseminating/sharing data*
raphy, geography, geology (including major
structures, faults, tectonic plates, existing volcanic This critical issue should not be underesti-
and other hazards, and aquifers), geo-history (a mated based upon difficulties encountered during
time-line of major events is often helpful), VUELCO exercises.
infrastructures and other essential information. It is inevitable that most of the participants
All the VUELCO exercises incorporated will wish to have easy access to the internet.
well-planned and informative pre-exercise field Ensure that the Wi-Fi network has sufficient
trips that were supported by printed, carefully capacity and range to allow easy access.
researched field guides with relevant histories, Consideration should be given to the format
maps, reading lists etc. requirements of the computer models and proba-
A field guide can be supplemented by more bilistic tools that will be tested during the exercise.
detailed essential information about the moni- • Language arrangements*
toring history, thresholds etc. Feedback from
VUELCO exercises suggests that the briefing Consider the dominant language for the
pack should include maps showing the positions exercise and its documents (particularly the
of all relevant observatories, monitoring equip- pre-exercise briefing pack) and what arrange-
ment and stations, GPS positions, cameras, ments can be made, if any, for
sample sources etc. translations/translators.
In addition to a field trip, it may be helpful to Feedback from the VUELCO exercises sug-
have a day/half-day of short presentations about gests that some participants (e.g. locally based
the host volcano with careful oblique references risk managers and members of the emergency
to features that will be relevant during the services) may become bored and disengaged if
exercise. they cannot understand fully what is going on
and contribute.
• Real/assumed legal/regulatory framework • Other equipment*
and related duty holders
Consider whether it will be necessary, for the
Each participant must have a clear and com- purposes of training, publicity or planning future
prehensive understanding of their role, the roles exercises, to use other equipment including:
of all other participants and all planned lines and
methods of communication. To ensure a higher – Display screens, smart boards and white
degree of realism, these roles should accurately boards
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 285

– Fixed and roving microphones • Checklists for the main aspects of the
A shortage of roving microphones caused exercise?
irritation at several VUELCO exercises.
– Visual and audio recording equipment Day 1 (Start of exercise)
– A projector for PowerPoint presentations Ensure that registration does not delay a
Feedback sessions are enhanced by such prompt start, which will thereby set the standard
presentations, if time allows for their for all other timings during the exercise.
preparation. Who will lead the exercise and ensure that the
– Laser pointers timetable is adhered to strictly?
– Video conference links Consider giving this role to a person (proba-
bly supported by an assistant) who does not have
• Presentations of data, results etc.* any other role in the exercise and therefore can
How, why and when will the data, results etc. move easily from room to room and deal
be presented? promptly with any difficulties that arise.
Consider imposing a strict timetable to avoid Is there a need for a short introduction? If yes,
timing problems. why, who will give it and how long will it last?
Is there a need for a short end of day sum-
• Requirements regarding pre-Day 1 reading, mary? If yes, why, who will give it and how long
preparation, queries etc. will it last?
Consider stating clearly that there will no Day 2 etc. (Continuation of exercise)
lengthy introductions or briefings at the start of Is there a need for a short introduction? If yes,
Day 1 because the participants are required and why, who will give it and how long will it last?
expected: (1) to read in advance the pre-exercise Is there a need for a short end of day sum-
briefing pack; and (2) to raise any queries before mary? If yes, why, who will give it and how long
Day 1 with the exercise organiser. will it last?
• Procedures for daily and end of exercise Day (End of exercise)
feedback. Is there a need for a short introduction? If yes,
why, who will give it and how long will it last?
If daily feedback or announcements are nec- Is there a need for a short end of exercise
essary, consider in advance what issues are likely summary? If yes, why, who will give it and how
to be covered, who will give them and allow long will it last?
adequate time within the timetable.

*Note 4.3 The Volcano Team

It is accepted that during a real emergency Membership

these facilities may not be readily available. Who will lead the team and why?
Accordingly, any difficulties encountered during Who will be the other Volcano members?
an exercise may provide helpful insights into the VUELCO’s exercises indicated that consid-
challenges that would be encountered during a erable time, exceptional skills and infinite
fast evolving crisis. patience are required to write any coherent and
Will role/team leaders receive in advance? challenging hazard scenario.
Members of the Volcano team must: (1) be
• A full pre-exercise briefing note about the selected and appointed at a very early stage in the
overall legal/regulatory infrastructure for planning process; and (2) be aware of, under-
the exercise? stand and accept the needs and goals of the
exercise.
286 R. J. Bretton et al.

The overall hazard scenario and its planned What data will be provided?
phases must be tailored to fit within the agreed The monitoring data must be suitable and
timetable for the exercise. The Volcano team will sufficient to test the whole range of geo-scientific
need to work very closely with the exercise lea- disciplines represented in the SAC. Feedback
der and the steering group as soon as it is formed. from the VUELCO exercises identified the
Where the Volcano team is the host volcano’s temptation to favour the overprovision of seismic
actual monitoring team, a chance to test the data at the expense of adequate geochemical,
monitoring team in ‘real time’ action will be lost. geodetic, petrological and other data.
Although this might reflect real situations or
Preparation even the architecture of existing monitoring
How long will be needed to prepare: networks, careful consideration should be given
• the scenario/s? to the consequences of having experts from some
• all of the briefing documents? disciplines having insufficient data to occupy
• the pre-exercise field trips? them and therefore becoming disengaged from
the exercise.
The exercise It may also be necessary to provide in the
What will be the time scale in months/years briefing pack related historical data and back-
covered by the exercise? ground information (such as historical thresh-
VUELCO’s exercise scenarios varied greatly olds) to allow the data used during an exercise to
in length—Colima (2 months), Campi Flegrei be considered in context.
(7 years), Cotopaxi (5 years) and Dominica The data must remain coherent, albeit delib-
(2 years). erately unclear and uncertain, over the full
The period of time covered by the simulation, duration of the scenario including periods of
the number of phases and the length of the inactivity. Feedback after the Dominica exercise
intervals between the phases, must take into praised the quality of the data and the design of
account the needs and goals of the exercise. For the scenario. The following expressions illustrate
example, consideration must be given to the time some of the features of a successful scenario—
that will be required for data to be delivered, very realistic unsure signals creating a state of
entered and assimilated, and for probabilistic limbo, non-linearity, phreatic evidence that was
models to be run. not a precursor of magmatic activity, complex
How many exercise phases will be needed to data and realistic missing data.
cover the period of years chosen? During some phases of the exercise scenario it
Each phase will involve, and must be allo- is possible that the data will be incomplete (some
cated sufficient time for, the delivery, entry, critical data may be held back deliberately to
analysis and modelling of the monitoring data await a request for more data) and/or difficult to
and the drafting and delivery of any expected analyse because it may be a stated objective of
outputs (e.g. reports, press releases etc.). the exercise to test the SAC’s ability to handle
VUELCO’s exercise scenarios had between 3 uncertainty, disagreement, monitoring inadequa-
and 6 phases - Colima (2 months in 4 phases cies etc. and/or to request data from additional or
over 10 h spread over 5 days), Campi Flegrei different monitoring.
(7 years in 4 phases over 3 days), Cotopaxi If data is provided for the periods between
(5 years in 6 phases in 1 day) and Dominica scenario phases, it is likely that the longer the
(2 years in 3 phases over 2 days). inter-phase period the more complete the data
As already stated, the precise specifications will be. Time will have to be planned for data
for the phases must be agreed with the exercise delivery and analysis. A large amount of data
leader taking into account the needs of the may also create very great difficulties for the
modellers and other technical teams. utility of probabilistic tools such as BET.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 287

The Volcano team should communicate with During the Mount Teide exercise, consider-
the exercise leader and modellers well before the able attention was paid to secondary hazards. If
exercise to avoid unnecessary surprises regarding secondary hazards are to be included, what data
the format and quantum of the monitoring data need to be provided before, at and after the start
and the timing of their release. In the Colima of the exercise?
exercise, three BET nodes were passed before Will plans be needed showing geographical,
Phase 1 of the exercise scenario and this degra- geological or societal features?
ded the utility of the BET trial. Will the exercise use fixed/dynamic data
How will the years between the periods that about weather, ground conditions etc.?
are covered by the data be described? If yes, what data need to be provided before,
This difficult issue should not be overlooked. at and after the start of the exercise? Will plans
In the absence of data, what will the participants be needed showing geographical, geological or
assume in absence of guidance? An absence of societal features?
data may also create very great difficulties for the Relevant weather/ground water data will be
utility of probabilistic tools such as BET. essential if ash dispersion and fall-out simula-
Consider issuing, along with the monitoring data tions are to be included within the exercise.
for the next phase, a brief summary of the ‘missing When, how and to whom will the monitoring
periods’ in terms of volcanic activity, precursor data be disseminated:
evidence and/or the risk mitigation decisions and
actions of civil protection authorities. • Before the start of the exercise?
Will the exercise scenarios be (1) entirely • At the start of the exercise?
fictitious; (2) based on suitably disguised real • During the exercise?
events; or (3) a mix of the two?
During the VUELCO exercises feedback This is a critical issue and important lessons
suggested that: were learned during the VUELCO exercises.
During a VUSE, reliance on a limited Wi-Fi link
• it may be very difficult to create realistic may be very problematical.
scenarios that are entirely fictitious because In the Dominica exercise, monitoring infor-
the data, under the very close expert analysis mation was provided within separate
that they will receive, may appear to be ‘phase-specific’ documents in password-protected
inconsistent and improbable; pdf format and emailed before the start of exercise
• a coherent and plausible chronology for the to all of the participants. At the beginning of each
scenario is essential; scenario phase, the relevant password was
• after a period of sudden emerging unrest, a released. This solution worked very well.
period of reducing unrest or quiescence may During the Cotopaxi exercise, Wi-Fi and other
form the basis of a challenging scenario as communication difficulties were identified by the
proved during the Dominica exercise; civil protection authorities and, as a result, they
• in order to simulate reality, the Volcano team were better placed to improve their existing
should provide monitoring data to the SAC systems and resources.
without any form of analysis or interpretation, Consideration should be given to the format
however the provision of tables comparing requirements of the computer models and prob-
data sets may be helpful; abilistic tools that will be tested during the
• the amount, relevance and format of the exercise.
monitoring data provided should be consid- Feedback from the VUELCO exercises sug-
ered carefully. gests that when data are provided they should
not duplicate data that have already been
Will the exercise introduce secondary hazards provided.
(e.g. forest fires, contaminated aquifers, etc.)?
288 R. J. Bretton et al.

Will the SAC be able to ask for further risks by the inclusion of disclaimers and exclu-
monitoring data? sion statements in hazard assessments. A VUSE
If yes, when and how can it be requested? If may serve as, at least, a prompt for committees to
yes, when and how will it be provided? seek legal advice from a competent local lawyer.
Will any player other than the SAC be able During the Dominica exercise, a member of the
to ask for monitoring data? Scientific Advisory Committee sought legal
If yes, by whom, why, when and how can it advice from the lead author about: (1) the wording
be requested? If yes, when and how will it be of a liability disclaimer for the benefit of visiting
provided? scientists; and (2) the way in which ‘risk-related’
Will anything other than monitoring data be advice could be passed to civil protection
provided? authorities without liability for subsequent risk
Consider whether the SAC and CPA should mitigation decisions. In respect of the first, it is
be given, or have access to when they request it, likely that a visiting scientist will have little or no
fixed/dynamic data about weather, ground con- control in respect of either the absence or ade-
ditions etc.? quacy of monitoring data and the selection and
If yes, why, what, when and to whom? competence of other committee members. Based
upon this advice, a disclaimer was included in the
Phase 1 hazard assessment report and the Phase 2
4.4 The Scientific Advisory and 3 reports were more carefully worded.
Committee (SAC)
Membership
Is a SAC necessary in every exercise? Who will be leading the SAC and why?
Careful consideration should be given to this The SAC will require a leader who should be
important question. It is appreciated that advice identified in advance.
to civil protection authorities may only be pro- Feedback from the VUELCO exercises sug-
vided by the head of the local volcano observa- gests the role of Chairperson of the SAC may be
tory, who is not, and would never be supported critical to the success of an exercise.
by a SAC. Consider whether a deputy Chairperson
Feedback from VUELCO exercises suggests should be identified in advance of the exercise.
that it makes good sense for exercises to simulate In several of the VUELCO exercises, it
existing risk governance structures as far as became evident that the Chairperson should not
reasonably practicable. It follows that a SAC be given, or retain, prime responsibility for the
team should be included in an exercise only: initial drafting of the hazard assessment report.
(1) where one exists, or could exist, in a period of Who will be the other SAC members?
evolving unrest or (2) where the civil protection Consider whether it is desirable to have a range
authorities wish to test the utility of one. of experts to simulate the issues that might arise
If an exercise SAC is simulating the actions of during a real period of emerging unrest. Consider
a real SAC, the brief to the SAC and the briefing the inclusion of: (1) local volcano observatory
note for the exercise should set out clearly the scientists; (2) other host country scientists; (3) host
SAC’s constitution, mandate, responsibilities and country academics (4) non-national observatory
powers. scientists; (5) non-national academics; and
(6) young researchers.
Legal status and duties The involvement of non-local and foreign
A VUSE can be a good opportunity to con- scientists may represent an opportunity to share
sider the legal status of scientific committees and knowledge and to have access to a wider range of
their full-time, part-time and seconded members. opinions uninfluenced by local social contexts,
Consideration should be given to the taking of non-scientific values or entrenched scientific
appropriate steps to avoid or reduce managerial assumptions or preconceptions.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 289

Consideration should be given to the need for Will the SAC get anything other than mon-
rules regarding the nature and extent of the itoring data?
interactions between local and non-local/foreign Consider whether the SAC should be given,
scientists. or have access to on request, fixed/dynamic data
During the Campi Flegrei exercise a number about weather, ground conditions etc.?
of young researchers were co-opted in rotation If yes, what data need to be provided before,
onto the SAC. This worked well and gave them a at and after the start of the exercise? Will plans
unique learning experience. be needed showing geographical, geological or
Which geoscience disciplines (e.g. geo- societal features?
physics, geochemistry, geodetics, geo-history,
and petrology) will be covered within the SAC? Deliberations
If the monitoring data resources of the host Will the SAC’s meetings/deliberations follow
volcano permit, ensure that the Volcano team standardised operating procedures (SOPs)
provides sufficient monitoring data to keep each and/or use a standardised reporting template?
discipline engaged and, if possible, challenged Feedback following the Dominica exercise
throughout the exercise. advocated the use of SOPs and the keeping of a
Will any geoscience disciplines not be cov- log recording and timing, inter alia, main data
ered? Why? inputs and analytical decisions, assumptions,
Consider whether this is a realistic represen- outputs and communications.
tation of what might happen in a real period of Careful consideration should be given to the
emerging unrest. SAC adopting working assumptions that:
Will the SAC have local members who are
usually based near or at the host volcano? • All hazard communications, including all
Consider whether the briefing pack should hazard assessments, in their format, content
refer to the IAVCEI protocol mentioned above. and delivery must be focussed upon: (1) fa-
Will the SAC have international members cilitating informed risk mitigation decisions
who are not based in the host country? that may not only be difficult but also based
Consider whether the briefing pack should upon many sources of knowledge and social,
refer to Newhall et al. (1999) mentioned above. political, economic and other influences; and
Will the SAC be realistic in size, too big or (2) responding to the identified needs and
too small? expectations of their makers.
Will any attempt be made to deal with the • Before acting as hazard communicators,
issue of ‘maverick’ scientists? hazard analysts must (by means of active and
If yes, how? If no, why? careful two-way dialogue) canvass, note and
A VUSE may present a good opportunity to respond to the needs and expectations of the
consider and test procedures for dealing very risk decision makers and the legitimate fore-
difficult issues such as this. seeable demands of mass and social media
and allow sufficient time to do this.
Inputs • The utility of hazard communications (i.e. the
When and in what format will the SAC get outputs of hazard analysis) must be judged by
the monitoring data? empirical evidence of the sentiments and
This is a critical issue. During a VUSE, reli- actions of risk decision-makers (i.e. the out-
ance on a limited Wi-Fi link may be very comes of hazard analysis within wider risk
problematical. mitigation processes).
Will the SAC be able to ask for further • No assumptions will be made as to what risk
monitoring data? decision-makers know about: (1) the science
If yes, when and how can it be requested? If of volcanic hazards and in particular its
yes, when and how will it be provided? complexities, uncertainties and limitations;
290 R. J. Bretton et al.

(2) the role of scientists and the many tem- envisaged their results will be integrated within
poral, financial, legal and other constraints the overall analysis of relevant hazard scenarios.
under which they operate particularly during Consider including in the pre-exercise briefing
periods of emerging volcanic unrest; and pack an additional file containing essential
(3) the role, benefits and limitations of background information about expert elicitations
long-term and short-term monitoring. and modelling.
• No assumptions will be made about what risk Will the SAC have to liaise with any
decision-makers either need or want and, non-SAC players?
accordingly, the widest possible range of If yes, when and how will this be done?
reasonably practicable options for not only During the Dominica exercise, sub-team
hazard communication but also hazard anal- leaders of the SAC were asked to attend early
ysis should be offered and discussed. parts of the CPA’s deliberations. In the
• No assumptions will be made that risk post-exercise feedback, it was suggested that all of
decision-makers know what general and/or the SAC scientists would have benefitted greatly
bespoke analysis, information, data, advice from hearing the CPA’s deliberations which were
and guidance scientists may be able to pro- often very critical of the scientific analysis upon
vide if asked and given adequate resources. which risk decisions had to be made.

Will any guidance be given as to how the Outputs—Hazard communications

data should be considered (e.g. expert elicita- What outputs (e.g. oral and written reports,
tion, probabilistic tools etc.)? written minutes etc.) will be required from the
If an expert elicitation is a possibility or is SAC?
actively encouraged/required, timetable enough The quality of the dialogue between risk
time for this as it is likely that a briefing will be management stakeholders, and the provision of
necessary before the elicitation starts to ensure information and advice, depends upon a mutual
that everyone is fully aware of what will be understanding by those stakeholders of their
involved. Do not underestimate the amount of respective needs, responsibilities, functions
time that a worthwhile elicitation will take. demands and roles, and their capacity to antici-
Consider providing essential information in pate other stakeholders’ demands and decision
the form of an elicitation toolkit in advance. Who needs (Salas et al. 1994; Ronan et al. 2008;
will prepare this toolkit? Lipshitz et al. 2001; Paton and Jackson 2002;
Can the SAC take into consideration any Doyle and Johnston 2011a, b; Doyle et al. 2014).
societal factors such as high societal exposure in “Without knowing the concerns” and under-
a particular area? If yes, what ‘exposure’ and standings “of the targeted audience, communi-
‘vulnerability’ data will be provided, when and cation will not succeed” (Renn 2008, 147).
how? Do not underestimate the amount of time Careful consideration should be given to this
that this will take. critical issue. Based upon the VUELCO exer-
Consider how and when the results of prob- cises, it may be very unwise to provide any
abilistic models will be delivered to the SAC and guidance whatsoever as to how the outputs of the
thereafter considered. Do not underestimate the hazard analysis must be communicated to the
amount of time that this will take. CPA and other exercise participants. This may
Feedback after VUELCO’s Dominica exer- encourage the type of two-way dialogue referred
cise showed that scientists involved in exercises to under the above heading “Deliberations”. By
need to be briefed about: (1) the claimed benefits providing guidance of any sort, a unique learning
offered by the probabilistic models being used; opportunity may be lost.
(2) their limitations; (3) when and in what form Consider briefing the observers in advance
their results will be presented; and (4) how it is about this important issue, so that they can make
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 291

an early invention if unwarranted assumptions Consideration might be given to a require-

are made about what risk decision makers know, ment within both the briefing pack and the
need or want. timetable that representatives of the SAC discuss
Will the SAC be required to communicate with the CPA:
with teams representing:
• the timing, format, content etc. of hazard
• The CPA?
reports, communication channels and proto-
• The public?
cols (the preparation of a template was sug-
• The media?
gested after the Campi Flegrei exercise),
• Representatives of local/central government?
and/or
If yes, when and where? • the arrangements for a possible mock joint
A VUSE is a good opportunity to practise press conference later in the exercise. A joint
hazard communications and to consider the press conference might involve another
needs and expectation of risk decisions and their press/media team (possibly made up of indi-
makers. Reference should be made to Doyle et al. viduals from the CPA’s public relations
(2015). department) and/or invited observers from
Consider how, when and by whom the needs local and national press organisations.
and expectations of risk decision-makers will be
sought. Membership
Will the CPA represent a real/fictional entity?
Who will be leading the CPA and why?
4.5 The Risk Managers—Civil Who will be the other CPA members?
Protection Authorities
Inputs
(CPA)
What inputs will the CPA get?
Will the CPA get monitoring data?
Role
If yes, why?
It is critical for the role of the CPA to be
considered very carefully during the planning Deliberations
stage. Will any guidance be given as to how the
Feedback from the VUELCO exercises sug- CPA inputs should be considered?
gests that the CPA’s role must be: Will the CPA have to liaise with any
• clearly defined and integrated within the non-CPA players?
design of the hazard scenario; If yes, when and how will this be done?
• sufficiently interesting and demanding to keep
Outputs
the participants engaged; and
What outputs will be required from the CPA?
• known to all other participants.
Will the CPA be required to communicate
The geoscientists will be engaged predomi- with teams representing:
nantly with the prompt analysis of monitoring
• The SAC?
data and this will lead subsequently to the
• The public?
delivery of hazard assessments in some form to
• The media?
the CPA. Depending upon the goals of the
• Representatives of local/central government?
exercise, it is likely that, whilst this is being
done, the CPA team must have something
A simulated press conference, which involves
worthwhile to do.
both scientists and CPA officials within the same
292 R. J. Bretton et al.

panel, can serve a number of useful purposes. In 4.6 Observers/Auditors

a potentially hostile environment, scientists will
have to communicate effectively complex Is it necessary to have observers and/or
geo-scientific concepts and explain issues such as auditors?
uncertainty and analytical/diagnostic confidence. In the VUELCO exercises a very valuable
They may also have to avoid being drawn into contribution was made by monitors and obser-
sensitive areas involving societal issues and risk vers. They can be from the participating organi-
decisions. Risk decision makers will have to sations or entirely ‘independent’. Monitors and
demonstrate an understanding of complex hazard observers (particularly with experience of previ-
scenarios and to explain what risk mitigation ous exercises and the practices of countries other
decisions they have taken and the reasons for than the host country) can play a critical role in
them. providing both hot and cold feedback whilst also
Open and convincing displays of collabora- gaining invaluable personal experience to assist
tion and agreement between different authorities in the planning of future exercises.
and key individuals during real crises can build VUSE are invariably observed by local and
trust with representatives of the media and visiting students and early-career geoscience aca-
general public. Panel question and answers demics. With careful planning, it may be possible
sessions involving key personnel, who will have for less experienced scientists and civil protection
‘communication’ roles, can and should be authority decision makers to be involved directly
practised during VUSE because it is likely that (as happened in the Campi Flegrei VUSE) or
related training and resource needs will be indirectly in ‘shadow’ SAC and CPA teams. The
identified. participation of young researchers in this kind of
The VUELCO Dominica exercise included exercise represents a great chance to train future
three very realistic but successful simulated press generations of advisory scientists.
conferences. What role will be played by nominated
observers and auditors?
Feedback Consider whether it would be worthwhile
The processes of hazard analysis and hazard differentiating between: (1) observers who might
communication should be outcome focussed and have responsibility for capturing and presenting
driven by the expectations and needs of risk immediate unstructured feedback; and (2) audi-
decisions and risk decision-makers. tors who would compile a more formal report
Consideration should be given to measuring possibly in a format and using audit criteria
the extent to which the exercise and its partici- agreed by the main participants in advance.
pants addressed: (1) the needs and expectations Would something be gained by having an
of the CPA; (2) the steps, if any, the SAC took to ‘independent’ observer/auditor who is not part
identify those needs and expectations; and (3) the of any organisation taking part in the exercise?
extent to which the SAC satisfied those needs; Independent observers played a very useful
and identified actions that might lead to role in the VUELCO exercises in Campi Flegrei,
improvements. Cotopaxi and Dominica and the author was an
The VUELCO exercises identified the value invited external auditor of the Tenerife exercise.
of civil protection authorities having an internal During the Campi Flegrei exercise, it was
technical-scientific structure: (1) to support the recorded that undertaking the role of observer is
SAC in technical analysis (mapping, elicitation, particularly helpful for senior civil protection
models); and: (2) to improve interactions with the authority managers who are thinking of planning
SAC. an exercise in their own locality.
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 293

Who will be nominated and/or invited to be – information (e.g. geographical informa-

observers/auditors? tion systems ‘GIS’ digital bases, overlays,
This will depend largely upon the roles which metadata, geological maps,
have been chosen and whether some of the remote-sensing images, background
observers/auditors should be ‘independent’. levels of unrest indicators, such as seismic
energy release rates, normal rates on
Format inflation/deflation, aqueous and geo-
At the end of the exercise will there be a chemistry and fluxes).
‘hot’ feedback session? • post-exercise actions to review and revise
The MIAVITA Handbook (2012, 118, 179) existing procedures in particular those for any
recommends an “on the spot debriefing straight future exercises.
after the exercises with the participants and
another one approximately one month later with Will the observers give real-time feedback
the exercise organisers, evaluators, observers, during the exercise?
and a representative of the general public”. Consideration should be given to this difficult
If yes, why, who will lead it and how long issue. A balance must be achieved between, on
will it last? the one hand, unhelpful and intrusive interven-
When and how will the results be recorded, tions and, on the other, allowing the exercise to
promulgated and acted upon? proceed in ways that may waste valuable time
Will there be ‘cold’ feedback meetings, sur- and/or miss or diminish valuable learning
veys or questionnaires? opportunities.
If yes, why and who will be responsible for During the Dominica exercise, the three ‘in-
their design and implementation? dependent’ observers gave brief feedback to the
When and how will the results be recorded, SAC at the end of Phases 1 and 2 of the exercise.
promulgated and acted upon? They also gave spontaneous suggestions to the
Will a more formal audit of the exercise be CPA. The participants indicated that they found
undertaken? this helpful and there was clear evidence that
If yes, consider whether it might address suggestions for change were understood and put
issues such as the identification of the need for into practice immediately.
better or additional:
Preparation
• structures of risk governance Will instructions be given to the
• information sources and resources observers/auditors in advance?
• long-term monitoring resources Consider giving the observers/auditors a clear
• consideration of the role of tectonics and brief before the start of the exercise.
resulting faults and features Experience from the VUELCO exercises
• geo-history data suggests that providing observers with a list of
• training more specific considerations, to include within
• financial, personnel and other resources their more general observations, may result in
• documented procedures more focussed outputs.
• communication plans and equipment Will instructions be given to the participants
• media response plans in advance of the exercise to facilitate richer
• easily accessible database catalogues, which feedback?
may be important and needed urgently during Consider asking each participant and/or each
a crisis, of: team (1) to list their main roles, responsibilities
– existing data (e.g. seismic, geodetic, gas, and tasks and (2) to keep a written list of needs
water geochemistry and geospatial) and that would help them accomplish their tasks
common archiving procedures for newly covering for example better/different information,
acquired datasets; and training, guidance, software, communication
294 R. J. Bretton et al.

equipment, other equipment, personnel, and so • The ‘logistics’ phase will ensure that the
on. These lists can then be used during the hot and execution of the exercise is not undermined
cold feedback sessions. by avoidable technical, communication and
Will the participants be told in advance other related difficulties. The importance of a
about the planned arrangements for observing/ comprehensive and comprehensible
auditing? pre-exercise briefing note and exercise plan
If yes, when? Consider whether the pre-exercise cannot be overstated.
briefing pack is the best place to set out the planned • A pre-VUSE field trip will serve numerous
arrangements and time table and what might be purposes including those of introducing rele-
required from the participants in this regard. vant geographical, geological, cultural and
governance histories and providing an
opportunity for relationships to develop
5 Discussion between the exercise participants.
• VUSE require and depend upon numerous
The checklists presented in this chapter are acts of communication between a wide range
dedicated to the organisers and participants of the of stakeholders with a variety of knowledge,
five exercises that this chapter has investigated. expertise, experience, needs and expectations
The checklists record and build upon the suc- within many types of formal/informal
cesses, experiences, occasional oversights, mis- relationships/associations (see Doyle et al.
judgements and mistakes of a few in order to 2015).
provide readily accessible sources of knowledge, • VUSE are not judged in terms of ‘success’ or
learning and inspiration for others. ‘failure’ but rather in terms of whether rele-
This chapter has identified a number of vant knowledge has been generated, recorded,
themes: considered and utilised.
• Whenever possible, VUSE simulate:
• VUSE are purpose-driven learning/training – ‘existing’ risk governance arrangements,
activities. Vital knowledge is acquired not or those being considered for the future,
only during but also before and after each with real policies, processes and people
exercise (see Doyle et al. 2015). rather than contrived unrealistic gover-
• VUSE must have clear recorded goals based nance scenarios and false role-play; and
upon the needs and expectations of their – volcanic hazard scenarios based upon, and
participants. significant within, the context of the host
• VUSE are complex and require clear and firm volcano region, which may experience
leadership, and very careful planning, funding other relevant or related natural hazards
and execution. This finding is supported by such as tectonic and weather-related
the work of Dohaney et al. (2015), which hazards.
contains very helpful guidance on design and • VUSE provide unique opportunities:
evaluation methods. – to address many known challenges
• During the ‘planning phase’, the critical including those of inter-scientist delibera-
issues of leadership, purpose, scope, duration, tion, scientific uncertainty, analytical/
type and financing should be considered. diagnostic confidence, unorthodox/
• The mere planning of a VUSE will provide maverick science sources, hazard com-
invaluable knowledge of the wider legal and munication, and mass media and social
administrative infrastructures in which it is networking relations and communications
framed. (see Doyle et al. 2015);
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 295

– to test new/prototype risk arrangements volcanic risk mitigation? This question can be
and tools, models and protocols for answered only if they are used and relevant
analysis/communication. evidence is generated within future post-exercise
• Great care should be taken to ensure that empirical studies.
comprehensive feedback and goal-related The Sendai Framework sets ambitious goals
knowledge can be captured during and after for quality standards, learning, knowledge
each exercise and disseminated as widely as exchange, education and training. The paramount
possible. goal of this chapter is to ensure that the pooled
experiences of the few, who have had the
advantages and privilege of being exercise par-
ticipants, will be accessible to the widest possible
6 Conclusions audience, to encourage future exercises and
thereby to improve the governance of volcanic
“Practice doesn’t make perfect. Practice reduces risks.
the imperfection”. None of the five exercises
considered in this chapter was completely per-
fect. They involved not only well-intentioned yet Disclaimer
imperfect policies and procedures capable of
improvement but also dedicated yet not infallible The content of this paper reflects the authors’
people keen to seek further knowledge, training views and not necessarily the opinion of the
and experience. Careful analysis of these exer- organisations to which they belong. The authors
cises shows that, as suggested by Doyle et al. and their organisations are not liable for any use
(2015), with very careful planning, execution and that may be made of the information contained
review, worrisome periods of emerging volcanic therein.
unrest and the dynamics of real-time hazard
assessments and risk decisions can be simulated
for a wide variety of worthwhile purposes. References
However this chapter does not attempt to dic-
tate how future exercises must be organised. Based Bartolini S, Cappello A, Martin J, Del Negro C (2013)
upon a review of the published records of four QVAST: a new quantum GIS plug for estimating
volcanic susceptibility. Nat Hazards Earth Syst Sci 13,
exercises and the authors’ personal experiences of 3013-3-34
the five further exercises listed in Table 1, it is Bergs J, Hellings J, Cleemput I, Zurel Ő, De Troyer V,
believed that these lists will assist the organisers of Van Hiel M, Demeere J-L, Claeys D, Vandijck D
future exercises to meet the specific challenges of (2014) Systematic review and meta-analysis of the
effect of the World Health Organization surgical safety
the volcanic hazards they face and the societal, checklist on postoperative complications. BJS
political, economic and legal contexts in which 101:150–158. https://doi.org/10.1002/bjs.9381
difficult and timely risk decisions have to be made. Beta T (2011) Master of Stupidity. http://www.goodreads.
The authors hope that the checklists will be com/book/show/8542012-master-of-stupidity
Bretton RJ (2014) The role of science in the rule of law.
considered and used, and, above all, improved. http://expertsinuncertainty.net
Candid feedback from future exercises will Bretton RJ, Gottsmann J, Aspinall WP, Christie R (2015)
ensure that the guidance notes evolve and will be Implications of legal scrutiny processes (including the
supplemented by the addition of further detailed L’Aquila trial and other recent court cases) for future
volcanic risk governance. J Appl Volcanol 4:18.
exercise studies. Will these checklists facilitate https://doi.org/10.1186/s13617-015-0034-x
more effective future exercises and improve
296 R. J. Bretton et al.

Dohaney J, Brogt E, Kennedy B, Wilson TM, Lindsay JM Hood C (1986) Administrative analysis: an introduction
(2015) Training in crisis communication and volcanic to rules, enforcement and organisations. Wheatsheaf
unrest forecasting: design and evaluation of an Books, Sussex
authentic role-play simulation. J Appl Volcanol 4:12. Hood C (2011) The blame game—Spin, bureaucracy and
https://doi.org/10.1186/s13617-015-0030-1 self-preservation in government. Princeton University
Donovan A, Oppenheimer C (2012) Governing the Press, Princeton, New Jersey, USA and Woodstock,
lithosphere: insights from Eyjafjallajökull concerning Oxfordshire, UK
the role of scientists in supporting decision-making on IAVCEI (2015) News no. 1–3. http://www.iavcei.org/
active volcanoes. J Geophys Res 117:B03214. https:// International Federation of Red Cross and Red Crescent
doi.org/10.1029/2011JB0090802012 Societies (IFRC) and United Nations Development
Donovan A, Oppenheimer C (2014) Science, policy and Programme (2015) The checklist on law and disaster
place in volcanic disasters: insights from Montserrat. risk reduction, Pilot version, March 2015. IFRC,
Environ Sci Policy 38, 150–161 Geneva, Switzerland
Doyle EEH, Johnston DM (2011) Science advice for Jenkins S, Spence R, Baxter P, Komorowski J-C, Sara
critical decision-making. In: Paton D, Violanti J Barsotti S, Esposti-Ongaro T, Neri A (2012) Prepa-
(eds) Working in high risk environments: developing ration of disaster risk scenarios at La Soufriere,
sustained resilience. Charles C Thomas, Springfield, Guadeloupe poster at volcanic and magmatic studies
Illinois, USA, pp 69–92 group annual meeting 2013
Doyle EEH, Johnston DM, McClure J, Paton M (2011) The Johnston D, Scott B, Houghton B, Paton D, Dowrick D,
communication of uncertain scientific advice during Villamor P, Savage J (2002) Social and economic
natural hazard events. N Z J Psychol 40(4):39–50 consequences of historic caldera unrest at the Taupo
Doyle EEH, McClure J, Johnston DM, Paton D (2014) volcano, New Zealand and the management of future
Communicating likelihoods and probabilities in fore- episodes of unrest. Bull NZ Soc Earthquake Eng 35
casts of volcanic eruptions. J Volcanol Geoth Res 272 (4):215–229
(2014):1–15 Jordan TH, Chen Y.-T, Gasparini P, Madariaga R, Main I,
Doyle EEH, Paton D, Johnston DM (2015) Enhancing Marzocchi W, Papadopoulos G, Sobolev G,
scientific response in a crisis: evidence-based Yamaoka K, Zschau J (2011) Operational earthquake
approaches from emergency management in New forecasting: state of knowledge and guidelines for
Zealand. J Appl Volcanol 4:1. https://doi.org/10.1186/ implementation, findings and recommendations of the
s.13617-014-0020-8 international commission on earthquake forecasting
Errington EP (1997) Role-play. Higher education research for civil protection, submitted to the Department of
and development society of Australasia, Jamison Civil Protection, Rome, Italy. Ann Geophys 54, 4.
Centre, ACT https://doi.org/10.4401/ag-5350
Errington EP (2011) Mission possible: using near-world Lipshitz R, Klein G, Orasanu J, Salas E (2001) Focus
scenarios to prepare graduates for the professions. Int J article: taking stock of naturalistic decision making.
Teach Learn High Educ 23:84–91 J Behav Decis Mak 14(5):331–352
Fiske R (1984) Volcanologists, Journalists, and the Marrero JM, García A, Llinares Á, Berrocoso M, Ortiz R
Concerned Local Public: a tale of two crises in the (2015) Legal framework and scientific responsibilities
Eastern Caribbean. In: Explosive volcanism: incep- during volcanic crises: the case of the El Hierro
tion, evolution and hazards. Geophys. Study Commit- eruption (2011–2014). J Appl Volcanol 4:13. https://
tee, National Research Council, National Academy doi.org/10.1186/s13617-015-0028-8
Press, Washington, pp 170–176 Marzocchi W, Newhall C, Woo G (2012) The scientific
Galanti E, Goretti A, Foster B, Di Pasquali G (2006) management of volcanic crises. J Volcanol Geoth Res.
Geotech Geol Earthquake Eng 369–384 https://doi.org/10.1016/j.volgeorgres.2012.08.016
Gawande A (2010) The checklist manifesto—how to get McGuire WJ, Kilburn CRJ (1997) Forecasting volcanic
things right. Profile Books Ltd, London events: some contemporary issues Geo. Rundsch
Harris AJL, Gurioli L, Hughes EE, Lagreulet S (2012) 86:439–445
Impact of Eyjafjallajökull ash cloud: a newspaper MIAVITA (Mitigate and Assess risk from Volcanic
perspective. J Gepphys Res 117:35. https://doi.org/10. Impact on Terrain and human Activities) (2012)
1029/2011JB008735 Handbook for volcanic risk management: prevention,
Hincks TK, Jean-Christophe Komorowski J-C, crisis management, Resilience, MIAVITA team,
Sparks SR, Aspinall WP (2014) Retrospective analysis Orleans. http://miavita.brgm.fr/default.aspx
of uncertain eruption precursors at La Soufrière Mothes PA, Yepes HA, Hall ML, Ramón PA, Steel AL,
volcano, Guadeloupe, 1975–77: volcanic hazard Ruiz MC (2015) The scientific-community interface
assessment using a Bayesian Belief Network over the fifteen-year eruptive episode of Tungurahua
approach. J Appl Volcanol 2014(3):3. https://doi.org/ Volcano, Ecuador. J Appl Volcanol 4:9. https://doi.
10.1186/2191-5040-3-3 org/10.1186/s13617-015-0025-y
Volcanic Unrest Simulation Exercises: Checklists and Guidance … 297

Newhall CG (2010) A checklist for volcanic risk mitiga- Solana MC, Kilburn CRJ, Rolandi G (2008) Communicat-
tion. Cities on volcanoes 6 presentation, earth obser- ing eruption and hazard forecasts on Vesuvius, Southern
vatory of Singapore Italy. J Volcanol Geoth Res 189(2010):308–314
Newhall C, Aramaki S, Barberi F, Blong R, Calavache M, Sparks RSJ, Aspinall WP, Auker M, Crosweller S,
Cheminee J-L, Punongbayan R, Siebe C, Simkin T, Hincks T, Mahony S, Nadim F, Polley J, Syre E
Sparks RSJ, Tjetjep W (1999) International associa- (2012a) Mapping and characterising volcanic risk,
tion of volcanology and chemistry (IAVCEI) subcom- magmatic rifting and active volcanism conference,
mittee for crisis protocols - professional conduct of Addis Ababa, 12 Jan 2012
scientists during volcanic crises. Bull Volcanol Sparks RSJ, Aspinall WP, Auker M, Crosweller S,
60:323–334 Hincks T, Mahony S, Nadim F, Polley J, Syre E
NZ/MCDEM (2008) Exercise Ruaumoko 2008: final (2012b) Volcano hazard and exposure in global
exercise report. Ministry of Civil Defence & Civil facility for disaster reduction and recovery priority
Management, Wellington, NZ countries and risk management measures, GVM 1st
OECD (2015), Scientific advice for policy making: the workshop 30 Apr 2012, University of Bristol and NGI
role and responsibility of expert bodies and individual Norway for the World Bank
scientists. OECD science, technology and industry Treadwell JR, Lucas S, Tsou AY (2014) Surgical
policy papers, no. 21, OECD Publishing, Paris. http:// checklists: a systematic review of impacts and imple-
dx.doi.org/10.1787/5js33l1jcpwb-en mentation. BMJ Qual Saf 23:299–318. https://doi.org/
Paton D, Jackson D (2002) Developing disaster manage- 10.1136/bmjqs-2012-001797
ment capability: an assessment centre approach. UN/ISDR (2015) Sendai framework for disaster risk
Disaster Prev Manage 11(2):115–122 reduction 2015–30
Renn O (2008) Risk governance – coping with uncer- van Ments M (1999) The effective use of role-play:
tainty in a complex world. Earthscan, London, New practical techniques for improving learning. Konan
York Page Ltd, London
Ricci T, Nave R, Barberi F (2013) Vesuvio civil Woo G (2008) Probabilistic criteria for volcano evacua-
protection exercise MESIMEX: survey on volcanic tion decisions. Nat Hazards 45:87–97. https://doi.org/
risk perception. Ann Geophys 56(4):s0452. https://doi. 10.1007/s11069-007-9171-9
org/10.4401/ag-6458
Ronan KR, Johnston DM, Daly M, Fairley R (2008)
School children’s risk perception and preparedness: a
hazard education survey. Aust J Disaster Trauma Stud
1. ISSN: 1174-4707. Available at: http://www.massey. Further reading sources on VHUB
ac.uk.nz/*trauma/issues/201-1/ronan.htm website
Rothstein H (2002) Neglected risk regulation: the insti-
tutional attenuation problem, centre for analysis of risk
and regulation. London School of Economics and VUELCO resources
Political Science, London Colima Exercise Plan
Salas E, Stout RJ, Cannon-Bowers JA (1994) The role of Colima Debriefing Report
shared mental models in developing shared situational Ciolli S, De la Cruz-Reyna S (2013) Colima volcano exercise
awareness. In: Gilson RD, Garland DJ, Koonce JM debriefing report. https://vhub.org/resources/2477
(eds) Situational awareness in complex systems: Campi Flegrei Simulation Plan
proceedings of a cahfa conference. Embry-Riddle Campi Flegrei Simulation Debriefing Report
Aeronautical University Press, Daytona Beach, FL, Ciolli S, Cristiani C, Golisciani GA (2014) VUELCO—
pp 298–304 Campi Flegrei Caldera unrest scientific simulation—
Sandri L, Guidoboni E, Marzocchi W, Selva J (2009) debriefing report. https://vhub.org/resources/3640
Bayesian event tree for eruption forecasting (BET_EF) Cotopaxi Debriefing Report
at Vesuvius, Italy: a retrospective forward application Dominica Exercise Plan
to the 1631 eruption Bull. Volcanol 71:729–745 Dominica Exercise Debriefing Report
Sobradelo R, Bartolini S (2014) Marti J (2014) HASSET: Ciolli S, Robertson REA (2015) VUELCO—Dominica
a probability tool to evaluate future volcanic scenarios exercise debriefing report. https://vhub.org/resources/
using Bayesian inference. Bull Volc 76:770 3822
298 R. J. Bretton et al.

Open Access This chapter is licensed under the terms of The images or other third party material in this chapter are
the Creative Commons Attribution 4.0 International included in the chapter’s Creative Commons license, unless
License (http://creativecommons.org/licenses/by/4.0/), indicated otherwise in a credit line to the material. If
which permits use, sharing, adaptation, distribution and material is not included in the chapter’s Creative Commons
reproduction in any medium or format, as long as you license and your intended use is not permitted by statutory
give appropriate credit to the original author(s) and the regulation or exceeds the permitted use, you will need to
source, provide a link to the Creative Commons license obtain permission directly from the copyright holder.
and indicate if changes were made.
 Appendix
Volcanic Unrest: Terminology and Definitions
Servando De la Cruz-Reyna, Ana Teresa Mendoza Rosas,
Joachim Gottsmann

these definitions on volcanic unrest risk commu-

Motivation and Background nication world-wide. However, we expect this
compilation to instigate a dialogue and a shared
This document provides a set of definitions of sci-
platform of relevant terminologies between the
entific and legal terminologies used in the context of
different actors and stakeholders of volcanic risk
volcanic unrest. The underpinning research behind
governance [see also (Bretton et al. 2017)].
this document was performed at the Universidad
Analysis of past volcanic disasters shows that
Nacional Autónoma de México—Geophysics
one of their causes were communication failures
Institute (Instituto de Geofísica, IGEF-UNAM) in
amongst stakeholders (Fiske 1984; Hall 1990; Til-
collaboration with the National Center for Disaster
ling 2009; Fearnley et al. 2017). Some failures
Prevention (Centro Nacional de Prevención de
resulted from different perception of the meaning of
Desastres, CENAPRED, Secretaría de Gober-
terms used to describe or define the level of volcanic
nación) and Colima University (Universidad de
hazard or risk. This is in fact a problem related to
Colima, UCOL). Therefore, many definitions of the
semiotics. Definitions of scientific, social and legal
relevant legal terminology are based on Mexican
concepts that are fundamental for volcanic crisis
law and may not be applicable verbatim to other
management having semantic contents that can be
jurisdictions. However, this document has been
translated into other languages are a necessary first
designed carefully to provide a common set of
step to homogenize terminologies and improve risk
terms that are fundamental for volcanic unrest crisis
communications. It is thus necessary to develop
management. This should provide a common
efficient communication protocols that univocally
platform for the interaction amongst different actors
use the same meaning of the terminology (semantic
and stakeholders involved in the management of
content), integrate that meaning into a communi-
volcanic crises. The proposed definitions inherently
cation code that can be understood by all stake-
follow on from investigations and interactions
holders (syntactic content), and translate the effects
between scientists and decision-makers within the
and consequences that volcanic activity may cause
VUELCO project. It is not our intention to impose
(pragmatic content). The following sections pro-
vide some context for a shared terminology.

S. De la Cruz-Reyna
Universidad Nacional Autónoma de México (UNAM),
Mexico City, Mexico
A. T. Mendoza Rosas
Universidad Nacional Autónoma de México (UNAM),
Mexico City, Mexico
J. Gottsmann
School of Earth Sciences, University of Bristol, Bristol, UK

Advs in Volcanology (2019) 299–313

DOI 10.1007/978-3-319-58412-6
© The Editor(s) (if applicable) and The Author(s) 2019
300 Appendix

Scientific Terminology

Term Término Meaning Significado Further reading

(English) (Español)
Unrest Activación; A deviation from a background or Incremento del nivel de actividad Phillipson et al.
(volcanic) reactivación; baseline level behavior of a volcánica respecto a un nivel de (2013)
agitación volcano which might be a prelude fondo o nivel base, el cual puede
(volcánica) to an eruption, or to another ser preludio de una erupción o de
hazardous event alguna otra manifestación
peligrosa
Crisis Crisis Situation during which a volcano Situación en la que un volcán Tilling (1989)
(volcanic) (volcánica) shows signs of unrest, that can be muestra signos de agitación que
interpreted as indication for pueden interpretarse como
potential impending eruptive precursora de actividad eruptiva
activity and associated hazard. inminente y de los peligros que
A crisis may or may not conlleva. Una crisis volcánica
culminate in a dangerous puede o no culminar en una
eruption, but it may cause anxiety erupción peligrosa, pero puede
and socioeconomic unrest among producir ansiedad e inquietud
the affected population socioeconómica en la población
afectada
Volcanic Peligro Potentially damaging Fenómenos potencialmente De la
hazard (H) volcánico (H) pre-eruptive, syn-eruptive or dañinos pre- eruptivos, co- Cruz-Reyna and
post- eruptive phenomena such as eruptivos o post-eruptivos, como Tilling (2008)
pyroclastic flows, windborne ash, los flujos piroclásticos, ceniza
lava flows, volcanic gases, lahars, transportada por el viento, flujos
ground deformation, major de lava, gases volcánicos,
volcano-tectonic earthquakes, and deformación del suelo,
landslides. In probabilistic lahares, sismos
assessments, ‘‘hazard’’ is volcano-tectónicos de magnitud
sometimes referred to the considerable y deslizamientos.
probability that a specific Para la evaluación probabilística,
volcanic manifestation or a veces ‘‘peligro’’ es la
phenomenon will occur in a given probabilidad de que una
area, within a given time frame manifestación volcánica
(i.e., a threat) específica ocurra en una zona
determinada, en un intervalo de
tiempo dado
Volcanic Riesgo Probable loss (assets such as Pérdida probable (de activos Tilling (1989)
Risk (R) volcánico (R) lives, property, productive como vidas, propiedades, De la
capacity, etc.) in the event of the capacidad productiva, Cruz-Reyna and
manifestation of a volcanic infraestructura, etc.) en un área Tilling (2008)
hazard. It is a conditional sujeta a un peligro volcánico. Es
probability resulting from the la probabilidad condicional que
product of the probability (P) that resulta de multiplicar la
a specific volcanic hazard will probabilidad (P) de que una
occur (i.e., a threat) times the manifestación o fenómeno
value (L) of the exposed assets volcánico específico ocurra en un
times their vulnerability (V): área y en un tiempo determinados
R=P*L*V (peligro), por el valor (L) de los
activos por: R=P*L*V
(continued)
Appendix 301

Term Término Meaning Significado Further reading

(English) (Español)
Acceptable Riesgo Risk that an individual or Riesgo que un individuo o Peterson (1988)
risk aceptable community is willing to accept, or comunidad están dispuestos a
that a public official is prepared to aceptar, o que un funcionario
allow persons in his or her charge público puede permitir que las
to accept. Acceptable risk is a personas bajo su responsabilidad
function of the benefits of risk acepten. La aceptación del riesgo
mitigation (safety and avoided depende de los beneficios de su
losses) and its costs (losses of mitigación (seguridad y pérdidas
jobs and business, community evitadas) y su costo (pérdida de
disruption, the costs of trabajos y negocios, disfunción de
transportation, housing, and food la comunidad, gastos de
for evacuees, and costs of any transporte, alojamiento y
structures to divert hazards). alimento para evacuados, y costos
Tolerance for risk varies de estructuras para reducir
considerably, mainly because the peligros). La tolerancia al riesgo
benefits and costs of mitigating es muy variable, ya que los
risks vary greatly from individual beneficios y costos de mitigación
to individual and community to de riesgos pueden ser muy
community diferentes entre individuos y entre
comunidades
Vulnerability Vulnerabilidad The susceptibility of physical and La susceptibilidad de sistemas Jolly and De la
human systems to be affected by a físicos y humanos que pueden ser Cruz-Reyna
natural hazard. It can be afectados por fenómenos (2015)
expressed as a probability of naturales peligrosos. Se puede De la
damage, or as the proportion of expresar como una probabilidad Cruz-Reyna and
the total exposed assets expected de daño, o como la proporción del Tilling (2008)
to be affected by a given total de activos expuestos a daño
manifestation. In probabilistic por una manifestación
assessment “vulnerability” is the determinada. Para evaluación
expected loss of the exposed probabilística, ‘‘vulnerabilidad’’
value should a hazardous es el porcentaje esperado de
manifestation occur (i.e., the perdida de los valores expuestos
probability of loss) dado que una manifestación o
evento volcánico ocurra (es decir,
probabilidad de perdida)
Preparedness Preparación A state of readiness or preparation Estado de preparación y Jolly and De la
for use or action in the case of conciencia para prevenir y actuar Cruz-Reyna
volcanic activity or any other en el caso de alguna actividad (2015)
threat. It involves a clear volcánica o cualquier otra De la
understanding by the population amenaza. Involucra una clara Cruz-Reyna and
and authorities of the natural comprensión de los fenómenos Tilling (2008)
phenomena and their destructive naturales y de sus efectos
effects. Series of measures to destructivos por parte de la
reduce vulnerability población y las autoridades. Serie
de medidas para reducir la
vulnerabilidad
Volcanic Desastre An event or a series of events Evento o serie de eventos Jolly and De la
disaster volcánico marked by a significant loss of caracterizados por una pérdida Cruz-Reyna
life and/or other assets that significativa de activos y del (2015)
exceed(s) the local response valor, que excede la capacidad de
capacity and results from one or respuesta local y que resultan de
more volcanic hazards uno o más peligros volcánicos
Alert Alerta A condition of heightened Condición de atención, vigilancia
watchfulness or preparation for y preparación para la acción. Un
(continued)
302 Appendix

Term Término Meaning Significado Further reading

(English) (Español)
action. A warning method or método o sistema de aviso para Jolly and De la
system to make people aware of que la gente perciba un peligro Cruz-Reyna
impending danger inminente (2015)
Alert code Código de A set of rules or a system used for Un conjunto de reglas o un Jolly and De la
Alerta transmitting alert messages sistema utilizado para la Cruz-Reyna
requiring brevity and clarity. The transmisión de mensajes de alerta (2015)
symbolic arrangement of breves y claros. Estructura
instructions in a warning system simbólica de instrucciones en un
sistema de alerta.
Consensus Consenso General agreement in the Acuerdo general en el diagnóstico Jolly and De la
diagnostics and the prognosis of y el pronóstico de la mayoría de Cruz-Reyna
the majority of scientists involved los científicos involucrados en el (2015)
in the study of an episode of estudio de un episodio de
volcanic activity. Consensus of actividad volcánica. El consenso
the involved scientists is de los científicos involucrados es
important for the decision making muy importante para la toma de
of authorities decisiones de las autoridades
Crisis Crisis An unstable situation of increased Una situación inestable de Jolly and De la
danger. A crucial stage in the amenaza en aumento. Etapa Cruz-Reyna
course of volcanic unrest, when crucial en el curso de una (2015)
the threat of an eruption, or the reactiviación/agitación volcánica
possibility of an eruption, exceeds en que la amenaza de una
some threshold or reference level erupción, o la posibilidad de una
erupción exceden un nivel de
referencia
Pre-event Pre-evento Period of time when there is Período de tiempo en que se tiene Jolly and De la
significant evidence that a hazard evidencia significativa de que un Cruz-Reyna
could occur fenómeno potencialmente (2015)
causante de daños puede ocurrir
Volcanic Semáforo de A basic communications protocol Protocolo de comunicación Fearnley (2013)
Traffic Light Alerta that translates volcanic threat into básica que traduce las amenzas
Alert System Volcánica seven levels of preparedness for volcanicas en siete niveles de
(VTLAS) the emergency-management preparación para las autoridades
authorities, but only three de gestión de emergencias, pero
unambiguous levels of alert for en solo tres niveles de alerta sin
the public (color coded green– ambigüedades para el público
yellow–red). The changing status (código de colores
of the volcano threat is verde-amarillo-rojo). Los
represented as the most likely cambios en el estado del volcán
scenarios according to the se representan como los
opinions of an official scientific escenarios de riesgo más
committee that analyzes all probables de acuerdo al consenso
available data de un Comité Científico oficial
que analiza los datos disponibles
Seismic Enjambre A group of many earthquakes of Grupo de múltiples sismos de McNutt and
swarm sísmico similar size occurring closely tamaño similar estrechamente Roman (2015)
clustered in space and time with agrupados en el espacio y en el
no dominant main shock tiempo sin evento principal
dominante
Forecast Pronóstico A general description of future Una descripción general de Connor et al.
events, including rough estimates eventos futuros incluyendo (2015)
of time, location, and likely estimaciones de tiempo,
activity localización y actividad probable
(continued)
Appendix 303

Term Término Meaning Significado Further reading

(English) (Español)
Precursor Precursor A geological event that occurs Un evento geológico que se National
prior to an eruption and is related produce antes de una erupción y Academies of
to the preparation processes of the se relaciona con los procesos de Sciences,
forthcoming eruption preparación de la próxima Engineering
erupción Medicine (2017)
Prediction Predicción A specific description of future Una descripción específica de los National
events, including time, size, type, acontecimientos futuros, Academies of
location, and formal errors for incluyendo el tiempo, el tamaño, Sciences,
each el tipo, la ubicación y los errores Engineering
formales para cada uno Medicine (2017)
Probabilistic Probabilístico A statement of the relative Una declaración de la Connor et al.
likelihood of an event based on probabilidad relativa de un evento (2015)
study of a population of similar basado en el estudio de una
events that have occurred in the población de eventos similares
past que han ocurrido en el pasado

Legal Terminology

Term (English) Término Meaning Significado Further reading

(Español)
Accident (legal Siniestro Critical and harmful Situación crítica y dañina Ley General Protección
definition, (definición situation caused by the generada por la incidencia Civil, 6 junio 2012,
Mexico) legal, México) incidence of one or more de uno o más fenómenos Capitulo l, Artículo 2,
hazards in a building or perturbadores en un Fracción LV (Mexican
facility affecting the inmueble o instalación Civil Protection General
occupants and afectando a su población y Law, 6 June 2012, Chapter
installations, with possible equipo, con posible 1, 2nd Article,
effects on surrounding afectación a instalaciones Section LV)
facilities circundantes
Alert (legal Alerta Warning of the proximity El aviso de la proximidad Reglamento de la Ley
definition, (definición of an anthropogenic or de un Fenómeno General Protección Civil,
Mexico) legal, México) natural hazard, or of the Antropogénico o Natural 6 junio 2012, Artículo 2,
increase of risk associated Perturbador, o el Fracción I (Bylaws of the
to it incremento del Riesgo General Civil Protection
asociado al mismo Law, June 6, 2012, 2nd
Article, Fraction 1)
Brigade (legal Brigada Basic Civil Protection Grupo de personas que se Ley General Protección
definition, (definición Unit. A group of people organizan dentro de un Civil, 6 junio 2012,
Mexico) legal, México) organized within a facility inmueble, capacitadas y Capitulo l, Artículo 2,
trained in basic emergency adiestradas en funciones Fracción VI (Mexican
response functions such básicas de respuesta a Civil Protection General
as: first aid, fighting fires, emergencias tales como: Law, 6 June 2012, Chapter
evacuations, search and primeros auxilios, 1, 2nd Article, Section VI)
rescue; designated in the combate a conatos de
Internal Civil Protection incendio, evacuación,
Unit as responsible for the búsqueda y rescate;
development and designados en la Unidad
implementation of Interna de Protección Civil
prevention, rescue and como encargados del
(continued)
304 Appendix

Term (English) Término Meaning Significado Further reading

(Español)
recovery based on the desarrollo y ejecución de
provisions of the internal acciones de prevención,
civil protection program of auxilio y recuperación,
the facility con base en lo estipulado
en el Programa Interno de
Protección Civil del
inmueble
Disaster (legal Desastre A consequence of the Resultado de la ocurrencia Ley General Protección
definition, (definición occurrence of one or more de uno o más agentes Civil, 6 junio 2012,
Mexico) legal, México) severe (and/or extreme) perturbadores severos y/o Capitulo l, Artículo 2,
hazards, associated or not, extremos, concatenados o Fracción XVI. (Mexican
with a natural origin, or no, de origen natural, de la Civil Protection General
caused by human activity, actividad humana o Law, 6 June 2012, Chapter
or from outer space that aquellos provenientes del 1, 2nd Article,
cause an amount of espacio exterior, que Section XVI)
damage in a given time cuando acontecen en un
and in a determined area tiempo y en una zona
that exceeds the response determinada, causan daños
capacity of the affected que por su magnitud
community exceden la capacidad de
respuesta de la comunidad
afectada
Disaster area Zona de Specified area upon which Espacio territorial Ley General Protección
(legal desastre a competent authority determinado en el tiempo Civil, 6 junio 2012,
definition, (definición formally declares a por la declaración formal Capitulo l, Artículo 2,
Mexico) legal, México) temporal condition of de la autoridad Fracción LIX (Mexican
disrupted social structure, competente, en virtud del Civil Protection General
which precludes the desajuste que sufre en su Law, 6 June 2012, Chapter
normal community estructura social, 1, 2nd Article,
activities. Such declaration impidiéndose el Section LIX)
may involve funding from cumplimiento normal de
FONDEN (A special las actividades de la
federal fund that can be comunidad. Puede
only used to finance the involucrar el ejercicio de
recovery of disaster areas.) recursos públicos a través
del Fondo de Desastres
Disruptive Fenómeno Hazard of natural origin Agente perturbador Ley General Protección
natural natural producido por la Civil, 6 junio 2012,
phenomenon perturbador naturaleza Capitulo l, Artículo 2,
(legal (definición Fracción XXII (Mexican
definition, legal, México) Civil Protection General
Mexico) Law, 6 June 2012, Chapter
1, 2nd Article,
Section XXII)
Early warning Sistemas de A set of elements El conjunto de elementos Reglamento de la Ley
systems (legal alerta temprana providing timely and para la provisión de General Protección Civil,
definition, (definición efficient information, información oportuna y 6 junio 2012, Artículo 2,
Mexico) legal, México) allowing individuals eficaz, que permiten a Fracción XIII (Bylaws of
exposed to a threat to take individuos expuestos a una the General Civil
avoidance actions aimed to amenaza tomar acciones Protection Law, June 6,
prevent or reduce risk, and para evitar o reducir su 2012, 2nd Article,
to be prepared for an riesgo, así como Fraction XIII)
effective response. Early prepararse para una
Warning Systems include respuesta efectiva. Los
(continued)
Appendix 305

Term (English) Término Meaning Significado Further reading

(Español)
hazard recognition and Sistemas de Alerta
mapping; monitoring and Temprana incluyen
forecasting of impending conocimiento y mapeo de
events; processing and amenazas; monitoreo y
dissemination of warnings pronóstico de eventos
or alerts understandable to inminentes; proceso y
authorities and population difusión de Alertas
and adopting appropriate comprensibles a las
and timely response autoridades y población;
actions to such alerts así como adopción de
medidas apropiadas y
oportunas en respuesta a
tales Alertas
Emergency Emergencia Abnormal situation in Situación anormal que Ley General Protección
(legal which damage to society is puede causar un daño a la Civil, 6 junio 2012,
definition, likely, posing an excessive sociedad y propiciar un Capitulo l, Artículo 2,
Mexico) risk to the safety and riesgo excesivo para la Fracción XVII (Mexican
integrity of the general seguridad e integridad de Civil Protection General
population, generated or la población en general, Law, 6 June 2012, Chapter
associated with the generada o asociada con la 1, 2nd Article,
presence, imminence, or inminencia, alta Section XVII)
high probability of probabilidad o presencia
occurrence of a hazard de un agente perturbador
Evacuee (legal Evacuado Person who withdraws or Persona que, con carácter Ley General Protección
definition, (definición is removed from a place of preventivo y provisional Civil, 6 junio 2012,
Mexico) legal, México) residence in a situation of ante la posibilidad o Capitulo l, Artículo 2,
emergency or likelihood of certeza de una emergencia Fracción XIX (Mexican
disaster, as a preventive o desastre, se retira o es Civil Protection General
and temporal measure to retirado de su lugar de Law, 6 June 2012, Chapter
assure his or her safety and alojamiento usual, para 1, 2nd Article,
survival garantizar su seguridad y Section XIX)
supervivencia
Geological Fenómeno Hazard caused by actions Agente perturbador que Ley General Protección
phenomenon geológico and movements of the tiene como causa directa Civil, 6 junio 2012,
(legal (definición earth's crust. This category las acciones y Capitulo l, Artículo 2,
definition, legal, México) includes earthquakes, movimientos de la corteza Fracción XXIII (Mexican
Mexico) volcanic eruptions, terrestre. A esta categoría Civil Protection General
tsunamis, unstable slopes, pertenecen los sismos, las Law, 6 June 2012, Chapter
flows, landslides, terrain erupciones volcánicas, los 1, 2nd Article,
sinking and subsidence, tsunamis, la inestabilidad Section XXIII)
and ground cracks de laderas, los flujos, los
caídos o derrumbes, los
hundimientos, la
subsidencia y los
agrietamientos
Hazard (legal Peligro Probability of occurrence Probabilidad de ocurrencia Ley General Protección
definition, (definición of a potentially damaging de un agente perturbador Civil, 6 junio 2012,
Mexico) legal, México) and disruptive potencialmente dañino de Capitulo l, Artículo 2,
phenomenon with a cierta intensidad, durante Fracción XXXVII
certain intensity, in a un cierto periodo y en un (Mexican Civil Protection
specific region and in a sitio determinado General Law, 6 June 2012,
given time interval Chapter 1, 2nd Article,
Section XXXVII)
(continued)
306 Appendix

Term (English) Término Meaning Significado Further reading

(Español)
Help (legal Auxilio Support to people at risk Respuesta de ayuda a las Ley General Protección
definition, (definición or victims of a mishap, personas en riesgo o las Civil, 6 junio 2012,
Mexico) legal, México) emergency or disaster, as a víctimas de un siniestro, Capítulo l Artículo 2,
response by official or emergencia o desastre, por Fracción V (Mexican Civil
private specialized groups, parte de grupos Protection General Law, 6
or by internal civil especializados públicos o June 2012, Chapter 1, 2nd
protection units, as well as privados, o por las Article, Section V)
actions to protect other unidades internas de
exposed people and valued protección civil, así como
non-human assets las acciones para
salvaguardar los demás
agentes afectables
High-risk zone Zona de riesgo Human settlement located Asentamiento humano que Ley General Protección
(legal grave within an area of high risk se encuentra dentro de una Civil, 6 junio 2012,
definition, (definición caused by a hazard zona de grave riesgo, Capitulo l, Artículo 2,
Mexico) legal, México) originado por un posible Fracción LXI (Mexican
fenómeno perturbador Civil Protection General
Law, 6 June 2012, Chapter
1, 2nd Article,
Section LXI)
Imminent Riesgo A risk situation that, Aquel riesgo que según la Ley General Protección
(impending) Inminente according to a specialized opinión de una instancia Civil, 6 junio 2012,
risk (legal (definición technical body, requires técnica especializada, debe Capitulo l, Artículo 2,
definition, legal, México) immediate actions as there considerar la realización Fracción L (Mexican Civil
Mexico) are conditions or increased de acciones inmediatas en Protection General Law, 6
probabilities of adverse virtud de existir June 2012, Chapter 1, 2nd
effects on an endangered condiciones o altas Article, Section L)
person or valued probabilidades de que se
non-human asset produzcan los efectos
adversos sobre un agente
afectable
Incident (legal Incidente A minor event that may be El suceso que sin Reglamento de la Ley
definition, (definición considered normal and not constituir una situación General Protección Civil,
Mexico) legal, México) necessarily caused by anormal ni haber sido 6 junio 2012, Artículo 2,
hazards, but that may provocado por fenómenos Fracción XI (Mexican
generate conditions for the perturbadores severos, Civil Protection General
occurrence of a mishap, an puede crear condiciones Law, 6 June 2012, Chapter
emergency or a disaster precursoras de Siniestros, 1, 2nd Article,
Emergencias o Desastres Section XXVIII)
Injured party Damnificado Person affected by a Persona afectada por un Ley General Protección
(legal (definición hazard who suffered agente perturbador, ya sea Civil, 6 junio 2012,
definition, legal, México) damage to their person or que haya sufrido daños en Capitulo l, Artículo 2,
Mexico) to their property in such a su integridad física o un Fracción XIV (Mexican
degree that requires perjuicio en sus bienes de Civil Protection General
external assistance for his tal manera que requiere Law, 6 June 2012, Chapter
subsistence, remaining is asistencia externa para su 1, 2nd Article,
such condition as long as subsistencia; Section XIV)
the emergency persists, or considerándose con esa
the pre-disaster normality condición en tanto no se
is restored concluya la emergencia o
se restablezca la situación
de normalidad previa al
desastre
(continued)
Appendix 307

Term (English) Término Meaning Significado Further reading

(Español)
Integrated Gestión integral Set of actions aimed to the El conjunto de acciones Ley General Protección
management of de riesgos identification, analysis, encaminadas a la Civil, 6 junio 2012,
risks (legal (definición evaluation, control and identificación, análisis, Capitulo l, Artículo 2,
definition, legal, México) reduction of risk, evaluación, control y Fracción XXVIII
Mexico) considering that it has a reducción de los riesgos, (Mexican Civil Protection
multifactorial origin, and considerándolos por su General Law, 6 June 2012,
that it is a process in origen multifactorial y en Chapter 1, 2nd Article,
continuous construction. un proceso permanente de Section XXVIII)
Such actions involve all of construcción, que
the three levels of involucra a los tres niveles
government, and all other de gobierno, así como a
sectors of society los sectores de la sociedad,
promoting the design and lo que facilita la
implementation of public realización de acciones
policies, strategies and dirigidas a la creación e
procedures that, within a implementación de
sustained development políticas públicas,
scheme, reduce the estrategias y
disaster-prone structural procedimientos integrados
vulnerabilities, and al logro de pautas de
increase the resilience or desarrollo sostenible, que
strength of society. It combatan las causas
involves identification of estructurales de los
risks and their sources, and desastres y fortalezcan las
development of prevision, capacidades de resiliencia
prevention, mitigation, o resistencia de la
preparedness, relief, sociedad. Involucra las
recovery and etapas de: identificación de
reconstruction schemes los riesgos y/o su proceso
de formación, previsión,
prevención, mitigación,
preparación, auxilio,
recuperación y
reconstrucción
Internal Civil Programa Planning and operational Es un instrumento de Ley General Protección
Protection interno de tool circumscribed to an planeación y operación, Civil, 6 junio 2012,
program (legal Protección Civil agency, or organization circunscrito al ámbito de Capitulo l, Artículo 2,
definition, (definición public or private and una dependencia, entidad, Fracción XLI (Mexican
Mexico) legal, México) social sector, which is institución u organismo Civil Protection General
composed by the operating del sector público, privado Law, 6 June 2012, Chapter
plan for the Internal Civil o social; que se compone 1, 2nd Article,
Protection Unit, the plan por el plan operativo para Section XLI)
for continuity of la Unidad Interna de
operations and Protección Civil, el plan
contingency plan, and para la continuidad de
aims to mitigate the operaciones y el plan de
previously identified risks contingencias, y tiene
and define preventive and como propósito mitigar los
response to be able to meet riesgos previamente
the event of an emergency identificados y definir
or disaster acciones preventivas y de
respuesta para estar en
condiciones de atender la
eventualidad de alguna
emergencia o desastre
(continued)
308 Appendix

Term (English) Término Meaning Significado Further reading

(Español)
Mitigation Mitigación Any action aimed to Es toda acción orientada a Ley General Protección
(legal (definición reduce the impact or disminuir el impacto o Civil, 6 junio 2012,
definition, legal, México) damage by a hazard on a daños ante la presencia de Capitulo l, Artículo 2,
Mexico) person or valued un agente perturbador Fracción XXXVI
non-human asset sobre un agente afectable (Mexican Civil Protection
General Law, 6 June 2012,
Chapter 1, 2nd Article,
Section XXXVI)
Monitoring Sistemas de A set of elements allowing El conjunto de elementos Reglamento de la Ley
systems (legal monitoreo detection, measurement, que permiten detectar, General Protección Civil,
definition, (definición processing, prediction and medir, procesar, 6 junio 2012, Artículo 2,
Mexico) legal, México) study of the behavior of pronosticar y estudiar el Fracción XIV (Bylaws of
hazards, in order to assess comportamiento de los the General Civil
them and their risks agentes perturbadores, con Protection Law, June 6,
la finalidad de evaluar 2012, 2nd Article,
Peligros y Riesgos Fraction XIV)
National Risk Atlas Nacional Integral geographic Sistema integral de Ley General Protección
Atlas (legal de Riesgos information system of información sobre los Civil, 6 junio 2012,
definition, (definición threats and expected agentes perturbadores y Capitulo l, Artículo 2,
Mexico) legal, México) damages, resulting from a daños esperados, resultado Fracción IV (Mexican
spatial and temporal de un análisis espacial y Civil Protection General
analysis of the interaction temporal sobre la Law, 6 June 2012, Chapter
between hazards, interacción entre los 1, 2nd Article, Section IV)
vulnerability and exposure peligros, la vulnerabilidad
of affected entities y el grado de exposición
de los agentes afectables
Prevention Prevención Group of actions Conjunto de acciones y Ley General Protección
(legal (definición implemented in advance of mecanismos Civil, 6 junio 2012,
definition, legal, México) the occurrence of hazards implementados con Capitulo l, Artículo 2,
Mexico) aimed to foresee and antelación a la ocurrencia Fracción XXXIX
reduce the related risks, de los agentes (Mexican Civil Protection
identifying, reducing or perturbadores, con la General Law, 6 June 2012,
eliminating them, and to finalidad de conocer los Chapter 1, 2nd Article,
avert or reduce their peligros o los riesgos, Section XXXIX)
destructive impact on identificarlos, eliminarlos
populace, property, o reducirlos; evitar o
infrastructure, as well as mitigar su impacto
prevent the social destructivo sobre las
construction of risks personas, bienes,
infraestructura, así como
anticiparse a los procesos
sociales de construcción
de los mismos
Prevision (legal Previsión To acquire awareness of Tomar conciencia de los Ley General Protección
definition, (definición the risks, and of what is riesgos que pueden Civil, 6 junio 2012,
Mexico) legal, México) needed to face them causarse y las necesidades Capitulo l, Artículo 2,
through defined steps of para enfrentarlos a través Fracción XL (Mexican
risk identification, de las etapas de Civil Protection General
prevention, mitigation, identificación de riesgos, Law, 6 June 2012, Chapter
preparedness, emergency prevención, mitigación, 1, 2nd Article,
response, recovery and preparación, atención de Section XL)
reconstruction emergencias, recuperación
y reconstrucción
(continued)
Appendix 309

Term (English) Término Meaning Significado Further reading

(Español)
Reconstruction Reconstrucción Transient actions aimed to La acción transitoria Ley General Protección
(legal (definición recover the social and orientada a alcanzar el Civil, 6 junio 2012,
definition, legal, México) economic environment entorno de normalidad Capitulo l, Artículo 2,
Mexico) that prevailed among social y económica que Fracción XLIV (Mexican
people before suffering the prevalecía entre la Civil Protection General
effects of a hazard in a población antes de sufrir Law, 6 June 2012, Chapter
given area or jurisdiction. los efectos producidos por 1, 2nd Article,
This process should be un agente perturbador en Section XLIV)
aimed as much as possible un determinado espacio o
to reduce existing risks, jurisdicción. Este proceso
ensuring that no new risks debe buscar en la medida
would be created, so de lo posible la reducción
improving the pre-existing de los riesgos existentes,
conditions asegurando la no
generación de nuevos
riesgos y mejorando para
ello las condiciones
preexistentes
Recovery (legal Recuperación Process starting during an Proceso que inicia durante Ley General Protección
definition, (definición emergency, consisting of la emergencia, consistente Civil, 6 junio 2012,
Mexico) legal, México) actions aimed to return to en acciones encaminadas Capitulo l, Artículo 2,
the normality of the al retorno a la normalidad Fracción XLV (Mexican
affected community de la comunidad afectada Civil Protection General
Law, 6 June 2012, Chapter
1, 2nd Article,
Section XLV)
Regulation Agente Set of actions, tools, Lo constituyen las Ley General Protección
(legal regulador standards, works and more acciones, instrumentos, Civil, 6 junio 2012,
definition, (definición generally everything normas, obras y en general Capitulo l, Artículo 2,
Mexico, as legal, México) intended to protect people, todo aquello destinado a Fracción l. (Mexican Civil
opposite to property, strategic and proteger a las personas, Protection General Law, 6
disruption) productive infrastructure bienes, infraestructura June 2012, Chapter 1, 2nd
and the environment, and estratégica, planta Article, Section I)
to reduce risk and control productiva y el medio
and prevent adverse effects ambiente, a reducir los
from a hazard riesgos y a controlar y
prevenir los efectos
adversos de un agente
perturbador
Resilience Resiliencia It is the ability of a Es la capacidad de un Ley General Protección
(legal (definición hazard-exposed system, sistema, comunidad o Civil, 6 junio 2012,
definition, legal, México) community or society to sociedad potencialmente Capitulo l, Artículo 2,
Mexico) resist, assimilate, expuesta a un peligro para Fracción XLVIII
withstand and quickly and resistir, asimilar, adaptarse (Mexican Civil Protection
efficiently recover from y recuperarse de sus General Law, 6 June 2012,
damaging effects, efectos en un corto plazo y Chapter 1, 2nd Article,
preserving and recovering de manera eficiente, a Section XLVIII)
its basic and functional través de la preservación y
structures, and achieving a restauración de sus
better protection and estructuras básicas y
improved risk reduction funcionales, logrando una
for the future mejor protección futura y
mejorando las medidas de
reducción de riesgos
(continued)
310 Appendix

Term (English) Término Meaning Significado Further reading

(Español)
Risk (legal Riesgo Damage or probable loss Daños o pérdidas Ley General Protección
definition, (definición of an asset, resulting from probables sobre un agente Civil, 6 junio 2012,
Mexico) legal, México) the interaction between its afectable, resultado de la Capitulo l, Artículo 2,
vulnerability and the interacción entre su Fracción XLIV. (Mexican
presence of a hazard vulnerabilidad y la Civil Protection General
presencia de un agente Law, 6 June 2012, Chapter
perturbador 1, 2nd Article,
Section XLIV)
Risk Identificación Recognizing and assessing Reconocer y valorar las Ley General Protección
identification de riesgos potential damage or pérdidas o daños Civil, 6 junio 2012,
(legal (definición probable loss from hazards probables sobre los Capitulo l, Artículo 2,
definition, legal, México) in terms of their agentes afectables y su Fracción XXXI (Mexican
Mexico) geographical distribution distribución geográfica, a Civil Protection General
through the analysis of través del análisis de los Law, 6 June 2012, Chapter
hazard and vulnerability peligros y la 1, 2nd Article,
vulnerabilidad Section XXXI)
Risk reduction Reducción de Preventive intervention of Intervención preventiva de Ley General Protección
(legal riesgos individuals, institutions individuos, instituciones y Civil, 6 junio 2012,
definition, (definición and communities aimed to comunidades que nos Capitulo l, Artículo 2,
Mexico) legal, México) remove or reduce the permite eliminar o reducir, Fracción XLVI (Mexican
adverse impact of disasters mediante acciones de Civil Protection General
through preparedness and preparación y mitigación, Law, 6 June 2012, Chapter
mitigation actions. It el impacto adverso de los 1, 2nd Article,
involves the identification desastres. Contempla la Section XLVI)
of risks, and the analyses identificación de riesgos y
of vulnerabilities, el análisis de
resilience and response vulnerabilidades,
capabilities. Also involves resiliencia y capacidades
developing a culture of de respuesta, el desarrollo
civil protection, the public de una cultura de la
engagement and the protección civil, el
development of an compromiso público y el
institutional framework, as desarrollo de un marco
well as the implementation institucional, la
of environmental implementación de
protection measures, use medidas de protección del
of land and urban medio ambiente, uso del
planning, protection of suelo y planeación urbana,
critical infrastructure, protección de la
building partnerships for infraestructura crítica,
the implementation of generación de alianzas y
financial and risk transfer desarrollo de instrumentos
instruments, and the financieros y transferencia
development of early de riesgos, y el desarrollo
warning systems de sistemas de
alertamiento
Risk zone Zona de riesgo Defined territorial space Espacio territorial Ley General Protección
(legal (definición upon which there is a determinado en el que Civil, 6 junio 2012,
definition, legal, México) likelihood of harm caused existe la probabilidad de Capitulo l, Artículo 2,
Mexico) by a hazard que se produzca un daño, Fracción LX (Mexican
originado por un Civil Protection General
fenómeno perturbador Law, 6 June 2012, Chapter
1, 2nd Article,
Section LX)
(continued)
Appendix 311

Term (English) Término Meaning Significado Further reading

(Español)
Shelter (legal Albergue Facility settled to provide Instalación que se Ley General Protección
definition, (definición shelter to people whose establece para brindar Civil, 6 junio 2012,
Mexico) legal, México) habitations have been resguardo a las personas Capítulo l Artículo 2,
affected by a hazard. Such que se han visto afectadas Fracción III (Mexican
people may dwell in the en sus viviendas por los Civil Protection General
shelter until the recovery efectos de fenómenos Law, 6 June 2012, Chapter
or reconstruction of their perturbadores y en donde 1, 2nd Article, Section III)
habitations permanecen hasta que se
da la recuperación o
reconstrucción de sus
viviendas
Sheltered (legal Albergado Person who is temporarily Persona que en forma Ley General Protección
definition, (definición granted help, temporal recibe asilo, Civil, 6 junio 2012,
Mexico) legal, México) accommodation and amparo, alojamiento y Capítulo l Artículo 2,
protection against a threat, resguardo ante la amenaza, Fracción II (Mexican Civil
imminence or occurrence inminencia u ocurrencia de Protection General Law, 6
of a hazard un agente perturbador June 2012, Chapter 1, 2nd
Article, Section II)
Temporary Refugio Physical installation La instalación física Ley General Protección
shelter (legal temporal enabled to temporarily habilitada para brindar Civil, 6 junio 2012,
definition, (definición provide protection and temporalmente protección Capitulo l, Artículo 2,
Mexico) legal, México) welfare to people who do y bienestar a las personas Fracción XLVII (Mexican
not have immediate que no tienen Civil Protection General
accessibility to a safe posibilidades inmediatas Law, 6 June 2012, Chapter
dwelling in case of de acceso a una habitación 1, 2nd Article,
impending hazard, an segura en caso de un Section XLVII)
emergency, a major riesgo inminente, una
accident or a disaster emergencia, siniestro o
desastre
Vulnerability Vulnerabilidad Susceptibility or Susceptibilidad o Ley General Protección
(legal (definición propensity of an asset to propensión de un agente Civil, 6 junio 2012,
definition, legal, México) suffer damage or loss in afectable a sufrir daños o Capitulo l, Artículo 2,
Mexico) the presence of a pérdidas ante la presencia Fracción LVIII. (Mexican
disrupting phenomenon de un agente perturbador, Civil Protection General
determined by physical, determinado por factores Law, 6 June 2012, Chapter
social, economic and físicos, sociales, 1, 2nd Article,
environmental factor económicos y ambientales Section LVIII)
 Bibliography

Bretton RJ, Gottsmann J, Christie R (2017) The role of 1985 eruption of Nevado del Ruiz, Colombia. J Vol-
laws within the governance of volcanic risks advances canol Geoth Res 42:101–115
in volcanology. Springer, Berlin, Heidelberg Jolly G, De la Cruz-Reyna S (2015) Chapter 68—
Connor C, Bebbington M, Marzocchi W (2015) Chapter volcanic crisis management A2. In: Sigurdsson H
51—probabilistic volcanic hazard assessment A2. In: (ed) The encyclopedia of volcanoes, 2nd edn. Aca-
Sigurdsson H (ed) The encyclopedia of volcanoes, demic Press, Amsterdam, pp 1187–1202
2nd edn. Academic Press, Amsterdam, pp 897–910 McNutt SR, Roman DC (2015) Chapter 59—volcanic
De la Cruz-Reyna S, Tilling RI (2008) Scientific and seismicity A2. In: Sigurdsson H (ed) The encyclope-
public responses to the ongoing volcanic crisis at dia of volcanoes, 2nd edn. Academic Press, Amster-
Popocatépetl Volcano, Mexico: importance of an dam, pp 1011–1034
effective hazards-warning system. J Volcanol Geoth National Academies of Sciences, Engineering Medicine
Res 170:121–134 (2017) Volcanic eruptions and their repose, unrest,
Fearnley C, Winson AEG, Pallister J, Tilling R (2017) precursors, and timing. The National Academies
Volcano crisis communication: challenges and solu- Press, Washington, DC.
tions in the 21st Century: Berlin. Springer, Berlin Peterson DW (1988) Volcanic hazards and public
Heidelberg, Heidelberg, pp 1–19 response. J Geophy Res: Solid Earth 93(B5):4161–
Fearnley CJ (2013) Assigning a volcano alert level: 4170
negotiating uncertainty, risk, and complexity in Phillipson G, Sobradelo R, Gottsmann J (2013) Global
decision-making processes. Environ Plan A volcanic unrest in the 21st century: an analysis of the
45:1891–1911. first decade. J Volcanol Geoth Res 264:183–196
Fiske RS (1984) Volcanologists, journalists, and the Tilling RI (1989) Volcanic hazards and their mitigation—
concerned local public: a tale of two crises in the progress and problems. Rev Geophy 27(2):237–269
eastern Caribbean, 110–121 p Tilling RI (2009) El Chichón’s “surprise” eruption in
Hall ML (1990) Chronology of the principal scientific and 1982: lessons for reducing volcano risk. Geofísica
governmental actions leading up to the November 13, Internacional 48:3–19

Advs in Volcanology (2019) 313–313

DOI 10.1007/978-3-319-58412-6
© The Editor(s) (if applicable) and The Author(s) 2019