Deliverable D2.8 Initial and boundary conditions for quasi-steady magmatic eruption scenarios, caldera/edifice collapse scenarios and dome eruption scenarios based on numerical studies

naivenorthΤεχνίτη Νοημοσύνη και Ρομποτική

8 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

112 εμφανίσεις

EXPLORIS Deliverable D2.8


1



Project number: EVR1
-
CT
-
2002
-
40026

Project Coordinator: Dr. Augusto Neri




Deliverable D2.8


Initial and boundary conditions

for quasi
-
steady magmatic eruption
scenarios, caldera/edifice collapse scenarios and dome eruption
scenarios based on numerical

studies



Part A:

Initial and boundary conditions for quasi
-
steady
magmatic
eruptions and
caldera/edifice collapse scenarios

A.Folch, J.Martì,
A.Felpeto,
G.Ma
cedonio, A.Neri (CSIC

and INGV)



Part B:

Conduit
f
low
characteristics

and the impact they have o
n explosive
eruptions

K.Diller, A.Clarke,
B.Voig
h
t,
A.Neri,

(ASU,

PSU
, and INGV
)


Part C: Large abundance of carbon dioxide in magma below Vesuvius

Paolo Papale and Roberto Moretti

(
INGV
)




Report period 24

months

EXPLORIS Deliverable D2.8


2

Part A
.



Initial and boundary condition
s for quasi
-
steady
magmatic eruptions and
caldera/edifice
collapse scenarios


Arnau Folch
1
, Joan Martì
1
,
Alicia Felpeto
1
,
Giovanni Macedonio
2
, Augusto Neri
3


1
Institut Jaume Almeira, Consejo Superior de Investigaciones Cientificas

2
Osservatorio Vesuviano,
INGV Napoli

3
Centro per la Modellistica Fisica e Pericolosità dei Processi Vulcanici

INGV Pisa


Objective



Proportionate adequate initial and boundary conditions to the modelling of pyroclastic flows and
fallout dispersal during explosive eruptions. The
goal can be split into two contributions:

(i)
Quasi
-
steady magmatic eruptions scenario
. In this case it is essential to quantify the main
flow variables at the vent and also throughout the eruptive column in order to proportionate realistic
boundary condit
ions to the pyroclastic flows and fallout dispersal models.

(ii)
Caldera
-
forming eruptions scenario
. Collapse calderas are nearly circular or elliptical
volcanic depressions produced by a partial or total roof collapse of a shallow
-
level magma chamber
due
to its decompression during the course of an eruptive event. The occurrence of such a
catastrophic phenomena is controlled by multiple aspects related to geometry, flow regime within
the magma chamber plus conduit system, and syn
-
eruptive response of the e
mbedding country
rocks. The goal here is to assess the syn
-
eruptive stress field around the chamber plus conduit in
order to predict the likelihood of such phenomena. In this sense, one task of the project is to develop
a physical model coupling the fluid
dynamics within the chamber and the conduit with the structural
response of the volcanic edifice.


First year achievements



Definition and implementation of a model for the chamber plus conduit system coupled together
(see [Macedonio et al., 2005] for mor
e details). The purpose of the model, called CPIUC, is to have
an starting point for both quasi
-
steady eruptions (when coupled with models for the eruptive column
and dispersal of pyroclasts) and caldera forming eruptions (when coupled with models for the
host
rock behavior). The chamber model assumes an homogeneous composition of magma, a vertical
profile of volatile content and can have an arbitrary geometry. On the other hand, the conduit model
is based on the averaged mass and momentum balance equations

solved along an arbitrary
-
shaped
conduit but assuming choked
-
flow conditions at the exit. Bubble nucleation is considered when the
homogeneous flow pressure drops below the nucleation pressure given the total water content and
the solubility law. Above th
e nucleation level, bubbles and liquid magma are considered in
mechanical equilibrium. The same equilibrium assumption is assumed above the fragmentation
level between gas and pyroclasts. The integration of the density distribution in the chamber allows
to

obtain the total mass in the chamber as a function of pressure at the chamber top and, through the
conduit model, as a function of time (Figure 1). Successful simulation tests pertaining to rhyolitic
and basaltic magmas have been performed.


EXPLORIS Deliverable D2.8


3


-11
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
D
e
p
t
h

(
K
m
)
0
25
50
75
100
125
150
175
200
225
250
275
P
r
e
s
s
u
r
e

(
M
P
a
)
t

=

0

h
t

=

4

h
t

=

2
2

h
C
o
n
d
u
i
t
C
h
a
m
b
e
r
V
e
n
t


Figure 1.
-

Numerical results from the coupled magma chamber plus conduit model, indicating the
variation of pressure (above) and magma void fraction (below) along the whole system




Definition and implementation of a model for the host
rock and investigation of geometrical
constrains for caldera
-
forming eruptions (see [Folch and Martí, 2004] for details). The model,
solves via a Finite Element Method several rehologies such as elastic, thermoelastic or poroelastic
and contemplates standa
rd basic fracture criteria. Its main application to project has been to
investigate which chamber geometries may induce a syn
-
eruptive stress field suitable for the
formation of ring
-
fault calderas without needing previous fracturing or without the need of

an
extensional regional tectonic regime. It has been found that two dimensionless geometrical
parameters (chamber extension to chamber depth ratio and chamber eccentricity) determinate two
different regions of ring fault calderas with different associated

collapse regimes. Ring
-
fault region
A is related to large collapse calderas (i.e. Andean calderas or Western US calderas), for which few
depressurization is needed to set up a collapse initially governed by flexural bending of the chamber
roof. In contras
t, ring
-
fault region B is related to small to moderate size calderas (i.e. stratovolcano
calderas), for which much depressurization is needed. The collapse requires, in this case,
reactivation of pre
-
existing fractures and therefore is more complex and his
tory dependent.


EXPLORIS Deliverable D2.8


4

0
1
2
3
4
5
6
-
2
-
1
.
5
-
1
-
0
.
5
0
0
.
5
1
1
.
5


0
1
2
3
4
5
6
-
2
-
1
.
5
-
1
-
0
.
5
0
0
.
5
1
1
.
5


Figure 2.
-

Numerical model illustrating the conditions to form ring
-
fault calderas, maximum tensile
stress at surface (above) and maximum shear stress at a vertical distance on the magma chamber
border (below).


Second year achieveme
nts



The CPIUC model has been extended to other conduit geometries different from the cylindrical in
order to contemplate also multiple vents with and arbitrary inclination and ring
-
fissure conduits. It
allows to investigate the influence of the conduit g
eometry on the dynamics of caldera
-
forming
eruptions. A future publication [Folch and Martí, 2005] will be devoted to this subject.




Coupling of the CPIUC model (first year preliminary results) with a model for the volcanic plume
(see [Folch and Felpeto,

2005] for details). The outcome is a one
-
way coupled model that involves
all the spatial domains simultaneously and has a good compromise between physical accuracy and
computational efficiency. The coupled model outputs several relevant eruptive parameter
s such as
column height, mass eruption rate, or duration of the eruption. Time
-
dependency is introduced via a
quasi
-
steady approach, that is, considering a succession of steady states. The purpose of the
coupling is to proportionate time
-
dependent boundary

conditions according to the formats and
needs demanded by pyroclastic flow and fallout dispersal models developed in parallel in other
WP's.




Implementation of the coupling between the fluid (CPIUC) and the structural models developed
during the first y
ear of the project. This objective have been reach only in part since further
numerical tests are required and computational efficiency must be improved before launching a
demonstrator software.


EXPLORIS Deliverable D2.8


5

Future work (up to month 30)


Despite the general objectives

of D2.8 have been successfully covered it still remain some
work to do. A extension the month 30 would allow:



To finish the fluid
-
structure coupling properly. In addition, it would be interesting to develop a
graphical user
-
friendly interface to input d
ata and visualize the results.



To couple the CPIUC+plume with the fall3d fallout model developed also within the project. It
would proportionate, for the first time, a model covering ALL the spatial domains (from the
chamber to the deposit). The disposal

of such a tool would be an important aid not only in terms of
hazard assessment but also to reconstruct past events.


References cited

Authors

Date

Title

Journal

Reference

Folch, A. and J.
Martí

2004

Geometrical and mechanical
constraints on the formatio
n of
ring
-
fault calderas

Earth and Planetary
Science Letters

221, 215
-
225,
2004

Macedonio, G.,
A. Neri, J. Martí,
and A. Folch

2005

Temporal evolution of flow
conditions in sustained magmatic
explosive eruptions

Journal of
Volcanology and
Geothermal Resea
rch

Accepted
In press

Folch, A., and
A. Felpeto

2005

A coupled model for dispersal of
tephra during sustained explosive
eruptions

Journal of
Volcanology and
Geothermal Research

Accepted
In press

Folch, A., and J.
Martí

Manuscrip
t to be
submitted

Time
-
depe
ndent vent conditions
at the onset of explosive caldera
-
forming eruptions

Journal of
volcanology and
geothermal Research



EXPLORIS Deliverable D2.8


6

Part
B.


Conduit f
low
c
haracteristics

and the impact they have on explosive eruptions


Kristina Diller
1
, Amanda Clarke
1
,
Barry Voig
ht
2
and Augusto Neri
3


1
Arizona State University, Phoenix AZ (USA)

2
Penn State University, PN (USA)

3
Centro per la Modellistica Fisica e Pericolosità dei Processi Vulcanici

INGV Pisa


Background


This research is building on previous work published by Cla
rke et al. in 2002 that explored
unsteady vent conditions of Vulcanian explosions. Their research paired a simple, static conduit
formulation with a pyroclast dispersal model, PDAC2D (Neri, 1998), which is a numerical,
axisymmetric model with three partic
le sizes. Their conduit model assumed that magma flow had
stagnated in the conduit prior to the explosion. Their resulting simulations produced behavior that
generally mimicked many qualitative and quantitative characteristics of observed Vulcanian
explo
sions that occurred at Soufrière Hills Volcano, Montserrat, in August of 1997. Based on their
models, they concluded that conduit permeability played a significant role in controlling large
-
scale
eruption processes.


Table 1 contains input conditions used

in both Clarke et al.’s study and the research presented here.
These pre
-
eruptive subsurface initial conditions were derived from published data of the 1997
Vulcanian eruptions at Soufrière Hills Volcano.









TABLE 1

Conduit diameter

30 m

Vent plug

20 m thick

Initial conduit gas overpressure

10 MPa

Dissolved water content of the melt

4.3 wt% in chamber

Crystal volume fraction

45% in chamber

Melt composition

rhyolitic




This ongoing research has implemented a new, more rheologically complex, dyn
amic conduit
formulation. The new formulation models 1
-
D, steady
-
state, homogeneous flow through a vertical
conduit with constant cross section. And, allows for magma viscosity and density changes due to
degassing, crystallization, and the presence of bu
bbles. This formulation creates high overpressure
in the shallow conduit. This dynamic conduit formulation is based on equations from Melnik and
Sparks 1999 (shown in Figure 1). The model solves for equations of conservation of mass for both
the melt an
d gas phases and conservation of momentum. Darcy’s Law allows the melt and gas
phases to move at different veloc
ities relative to one another.

EXPLORIS Deliverable D2.8


7

Figure 1



This research focused on three main questions. First of all, how does the dynamic conduit model
compare with the previously employed static conduit model? Second, which conduit formulation
provides a reasonable fit to data derived from pumice samples? And finally, how does added
complexity in the conduit formulation affect the resulting explosion s
imulations?

Results

One of the goals of this research was to find output from the conduit model that provides a
reasonable fit to data recently derived from Soufrière Hills pumice samples (Stephens et al., 2004).
The study assumes that the fragmentation w
ave of Vulcanian explosions quenches the magma in its
pre
-
eruptive state, preserving a record within the pumice samples of pre
-
eruption conduit
conditions. We utilized a relationship derived experimentally by Couch (2002) & Couch et al.
(2003) to determin
e the pressure at which a pumice sample was quenched, based on the percent of
microlites in the sample. Figure 2 shows bulk density of pumice clasts collected from a pyroclastic
flow deposit on Montserrat in the summer of 1997 vs. the estimated quench pre
ssure, which is used
as a proxy for depth. There is a general trend of decreasing bulk density up through the conduit and
then a sharp increase in bulk density near the surface. Figure 2 shows the trend of magma porosity
with depth in the conduit.

Figure

2


Conservation
of Mass

Conservation
of Momentum

Darcy’s Law

Permeability

0
10
20
30
40
50
60
0
500
1000
1500
2000
2500
Bulk Density (kg/m
3
)
Total Pressure (MPa)
EXPLORIS Deliverable D2.8


8

The only free parameter in our conduit model that is not constrained by field data is permeability.
Therefore, we explored permeability as a variable in the dynamic conduit model. Figure 3 shows a
formulation considering only vertical permeability
, gases escaping only through the top surface of
the conduit. The graph shows the conduit model output as bulk density vs. total pressure (again as a
proxy for depth) for 4 different values of permeability coefficient. Bulk density increases as
permeabil
ity increases. However, no reasonable vertical permeability in conjunction with a high
magma flux rate of 7.5 m
3
/s (which is thought to be the approximate flux rate for the period of
Vulcanian explosions) captures the sharp increase in bulk density near t
he surface that is suggested
by the pumice samples.


Figure 3



The curves considering only vertical permeability do not capture the trend of the pumice data. We
thus realized that this increased density near the surface was an important factor to cons
ider,
primarily because it is the high density material near the surface which serves as a ‘cap,’ creating
the very conditions necessary for Vulcanian explosions. Our question then becomes: what is
causing this increased density in the pumice samples in t
he shallow subsurface? Our list of possible
culprits includes: macroscale permeability, the presence of fractures near the surface allowing for
enhanced gas loss; interconnectivity of bubbles near the surface, creating pathways that allow for
increased ga
s loss; unsteady dynamics in the conduit, in which there is rapid ascent of magma
followed by stagnation allowing time for extensive near
-
surface degassing; and lateral permeability,
where gases are escaping radially out of the conduit into the country roc
k. In terms of practical
integration of one of these concepts into our conduit formulation, macroscale permeability and
interconnectivity of bubbles and their effect on permeability would require additional data.
Accounting for unsteady dynamics in the c
ode would require an extensive reformulation of the
current code. Therefore, we decided that the most efficient option to test first was lateral
permeability. We did this by integrating the following equation into the dynamic conduit model


Q
gas loss la
teral

= A
vs
k
cr
(

p/

r)


The rate of gas lost laterally is equal to the vertical surface area (
A
vs
) of the conduit times the
permeability coefficient of the country rock (
k
cr
) times the radial pressure gradient
(

p/

r)
.




0
10
20
30
40
50
60
0
500
1000
1500
2000
2500
Bulk Density (kg/m
3
)
Total Pressure (MPa)
pumice samples
k0 = 0
k0 = 1
k0 = 10
k0 = 100
EXPLORIS Deliverable D2.8


9

Figure 4 shows the results from the conduit formul
ation that considers both vertical and lateral
permeability, using various vertical permeability coefficients, compared against the pumice data.
The axes again are bulk density vs. total pressure. These model curves do capture the trend toward
increasing

density near the surface, due to significant overpressure in the shallow sub
-
surface.
Results using vertical permeability coefficient K
0

of 0.1 and lateral permeability of 10
-
14

m
2

provided the most reasonable match to the pumice data. We therefore deci
ded to run an explosion
simulation with this information, believing that the best match to the pumice data should provide
the best reflection of the pre
-
eruptive conduit conditions prior to the Vulcanian explosions at
Soufrière Hills volcano in 1997. We t
hus used the output from that conduit formulation (black stars
Fig. 4) as initial conditions for an explosion simulation.


Figure 4












Figure 5 shows the conduit model output in terms of pressure vs. depth and gas volume fraction
with depth, tha
t were used as input for explosion simulations. Sim B is a run previously performed
by Clarke et al. (2002) using a static conduit formulation that considered lateral permeability. Sim
K is the dynamic conduit model output using both vertical and lateral

permeability which provided
the best match to the pumice data (See also Fig. 4).




















Figure 5







Sim B = red; Sim K = green.



0
1000
2000
3000
4000
5000
0
20
40
60
80
Total Pressure (MPa)
Depth (m)
0
1000
2000
3000
4000
5000
0
0.2
0.4
0.6
0.8
Gas Volume Fraction
Depth (m)
0
10
20
30
40
50
60
0
500
1000
1500
2000
2500
Bulk Density (kg/m
3
)
Total Pressure (MPa)
pumice samples
k0 = 0.1
k0 = 1
k0 = 10
EXPLORIS Deliverable D2.8


10

Results of these two explosions simulations are presented as two animations (included with this
report) o
f the first 120 seconds of the eruptions. The plots are of total particulate volumetric
fraction in the atmosphere, where colored contours indicate log to the base 10 from

8 to

1, with
red being the highest concentration and the outer yellow colors bein
g the limit of what is visible to
the naked eye. Important things to note are the increased energy and greater particulate volume
fractions of Sim K. Qualitatively the two plumes are different, but both produce features present in
the actual eruptions at

Soufrière Hills volcano


overhangs, collapsing columns, pyroclastic density
currents, and co
-
pyroclastic density current buoyant thermal plumes.


Figure 6 shows the explosion simulation results (dashed lines) on a graph of time after start of
explosion

vs. height of the plume, with the points representing data obtained from digital video
from explosions that occurred on August 6
th

and 7
th

1997 at Soufrière Hills Volcano, is it clear that
from 10 seconds on, Sim K is much more energetic than Sim B, and i
t obviously did not provide an
accurate model of the real plume ascent. This excess energy can be attributed to the fact that Sim K
had much greater gas volume fractions both at depth and in the shallow subsurface.


Figure 6




























The simulation based on the dynamic conduit formulation with both vertical and lateral
permeability is much more energetic than the actual explosions. But, why did it not match the real
eruptions, when it was the best match to the pumice data? One factor

to consider is that our
explosion model does not account for energy dissipation during fragmentation. Another possibility
is that post
-
fragmentation bubble expansion played a role in the eruptions. Stephens et al.’s
analysis of the pumice data used the
assumption that the pumice clasts were quenched immediately
at the time of fragmentation, so that the porosity of the pumice represents the gas volume fraction of
the magma at the time of fragmentation. If post
-
fragmentation bubble expansion did occur, th
e data
derived from the pumices would not accurately reflect the pre
-
eruptive gas volume fraction in the
conduit.


Recently, we attempted to quantify the amount of post
-
fragmentation bubble expansion for these
pumice samples. Using concepts concerning pos
t
-
fragmentation bubble expansion during the 1997
Soufri
è
re Hills Vulcanian explosions (Formenti and Druitt, 2003), we have adjusted the bulk
EXPLORIS Deliverable D2.8


11

density data of Stephens and Clarke to account for this expansion (Figure 7). We quickly created
new conduit resul
ts in hopes of matching the adjusted pumice data. This gold curve, which uses a
higher vertical permeability coefficient, appears to be a reasonable match. If this conduit output
was used as input for our next explosion simulation, the lower gas volume f
ractions throughout the
conduit should help to alleviate the excessive energy seen in Sim K (Figure 8). However, this was
our first attempt at quantifying the post
-
fragmentation bubble expansion and we plan to further
investigate this in the future, with
the overarching goal of understanding the relationship between
conduit dynamics and eruption plume dynamics.


Figure 7


0
10
20
30
40
50
60
500
1000
1500
2000
2500
3000
Bulk Density (kg/m
3
)
Total Pressure
(MPa)
pumice data
adjusted pumice data
k0 = 0.1
k0 = 2.5






EXPLORIS Deliverable D2.8


12

Figure 8





Recommendations and Future Work


Future work will involve the continued effort to qu
antify post
-
fragmentation bubble expansion of
the pumice samples. This will include an investigation of literature concerning post
-
fragmentation
bubble expansion and data from Montserrat pumice. Also, SEM images of our Montserrat pumice
samples will be a
nalyzed to search for evidence of post
-
fragmentation bubble expansion. In
addition, we will continue to improve the dynamic conduit code to create output results that more
accurately match the existing pumice data. We plan to investigate how the addition

of
crystallization in the conduit and different fragmentation mechanisms may affect conduit processes.
Finally, we will continue to perform explosion simulations in hopes of producing results that mimic
the actual explosions, thereby revealing informatio
n about the true connection between conduit
dynamics and eruptive plume dynamics.


0
2500
5000
0.25
0.45
0.65
Gas Volume Fraction
Depth (m)
Sim B
Sim K
New Sim?
EXPLORIS Deliverable D2.8


13

Part

C


Large abundance of carbon dioxide in magma below Vesuvius


Paolo Papale
1

and Roberto Moretti
2


Istituto Nazionale di Geofisica e Vulcanologia

1
Sez. Roma1, Via dell
a Faggiola 32, 56126 Pisa

2
Sez. Osservatorio Vesuviano, Via Diocleziano 328, 80124 Napoli


Carbon dioxide is commonly assumed to be about 10 times less abundant than water in most
natural magmas. However, direct measurements of the total H
2
O and CO
2

conten
ts in magmas do
not exist, as the available measurements refer to either the dissolved amounts in melt inclusions
(hereafter MI’s) within crystals, or the concentrations in gas plumes and fumaroles. Due to high
H
2
O solubility, MI’s can provide a reasonable

approximation of total water contents in magmas. On
the contrary, the very low CO
2

solubility in natural magmas results in its preferential partition in the
gas phase, making melt inclusions unsuitable for a direct estimate of its abundance. MI’s in olivi
ne
crystals from past eruptions of Vesuvius show maximum water contents around 5 wt%, while
dissolved carbon dioxide is at most about 0.4 wt%. Recent investigation (Papale, JGR in press)
shows that MI’s can be used in conjunction with multi
-
component gas
-
l
iquid equilibrium modelling
to constrain the total amounts of H
2
O and CO
2

consistent with the dissolved quantities. Here it is
shown that MI’s in crystals erupted during the 1944, 1906, and AD 79 eruptions of Vesuvius
indicate the existence of a large magm
a body with top at around 10 km depth and extending down
to at least 18
-
20 km. This magma body contains carbon dioxide in the order of at least some wt%,
and a gas phase made of 80
-
90 wt% CO
2
.

Any pair of H
2
O and CO
2

dissolved in a liquid at given
P
-
T
-
comp
osition conditions can
originate from an infinite combination of total H
2
O and CO
2

contents lying on a straight line called
TV (Total Volatile) line. A set of MI’s formed in closed system conditions relative to the gas phase
produces therefore a correspond
ing set of TV lines, the intersection of which corresponds to the
actual total H
2
O and CO
2

contents in the system. Crystallisation of magma corresponds to a shift of
TV lines, which can be easily corrected by including the amount of crystallisation. This q
uantity
can be determined for any pair of MI’s on the basis of the concentration of incompatible elements.
Open system conditions produce a progressive decrease of the slope of TV lines. In this case, the
envelope of all TV lines approximates the evolution

of total volatile contents in that system. These
simple concepts, which have been rigorously demonstrated and successfully applied to basaltic and
rhyolitic test cases, open new perspectives in the investigation of magmatic volatiles. In fact, the TV
line

concept provides a simple, sound, and straightforward method to constrain the amount of the
two major magmatic volatiles water and carbon dioxide during the various stages of magmatic
evolution.

The basic assumption underlying the TV method is that of loc
al equilibrium conditions, that is,
the liquid composition measured in MI’s must be representative of the local liquid
-
gas equilibrium.
The same assumption is commonly adopted in MI’s studies carried out all over the world. We make
this assumption througho
ut this study, although we stress the importance of improving much the
present
-
day knowledge of the kinetics of volatile diffusion at the boundary of crystals growing in a
magmatic melt.

Any pair of equilibrium dissolved H
2
O and CO
2

content corresponds to
a unique entrapment
pressure and composition of the coexisting gas phase. The composition of the gas phase is also
required in the definition of the TV line corresponding to that given H
2
O and CO
2

pair. In order to
be determined, pressure and gas phase com
position need an independent estimate of the entrapment
temperature, and the availability of a multi
-
component gas
-
liquid equilibrium model. Here we adopt
a fully non
-
ideal, non
-
Henrian model based on 10 major oxides and dissolved H
2
O and CO
2

volatiles (Pa
pale, AM 1999), after having re
-
calibrated the model by updating the database with the
EXPLORIS Deliverable D2.8


14

H
2
O, CO
2
, and H
2
O+CO
2

saturation data produced in the literature after its first calibration in 1997.
This means an increase of 537 data points which nearly double the o
riginal database. The linear
robust estimate technique adopted in model calibration is described in Papale (AM, 1999). Figure 1
shows the comparison between the experimental and calculated saturation conditions. A total of 856
data on H
2
O solubility, 182 d
ata on CO
2

solubility, and 85 data on H
2
O+CO
2

saturation in silicate
liquids is satisfactorily reproduced by the model. These data span pressure conditions from a few
tens to thousands MPa, temperature conditions from 800 to 2000 °C, and volatile
-
free comp
osition
conditions from synthetic two
-
component oxide systems to natural silicate melts from basaltic to
rhyolitic and peraluminous to peralkaline. The CO
2

saturation data produced before the eighties are
not included in the database, since they are affect
ed by systematic errors due to the
14
C

-
tracking
analytical technique used at that time. The dissolved amounts of H
2
O and CO
2

from MI’s in the
Vesuvius products considered here correspond to the initial portion of the 1:1 line in Fig. 1a,b,
where the corr
espondence between calculated and experimental conditions is maximum.

Figure 2 reports the MI’s data relative to the concentration of H
2
O and CO
2

measured in
samples from the AD79, 1906, and 1944 eruptions of Vesuvius. Most of the data cluster between 2
an
d 5 wt% H
2
O, and 500


4000 ppm of CO
2
. Two samples from MI’s in leucite crystals of the
1944 eruption have about 1 wt% H
2
O and 300
-
400 ppm CO
2
, and one sample from the interstitial
liquid of a cumulitic dunite of the same eruption shows about 1.5 wt% H
2
O
and 100


200 ppm
CO
2
. Figure 3 shows the application of the TV line concept to these MI’s, entrapment temperatures
having been calculated via a composition
-
based empirical function specifically calibrated on
Vesuvius MI’s. MI’s belonging to the same group

defined in Fig. 2 are processed in order to
determine the total (in the liquid and gas phases) H
2
O and CO
2

contents consistent with the amounts
dissolved in the liquid phase. Uncertainties (reported in Fig. 2) are taken into account by expanding
the analy
tical uncertainty in order to include the approximation of the model in reproducing the
experimental H
2
O and CO
2

saturation contents. The conservative total uncertainty of 15% for H
2
O
and 25% for CO
2

has been assumed, on the basis of 10% analytical uncerta
inty and experimental
-
calculated relationships in Fig. 1. The fields referring to different MI groups in Fig. 3 include
therefore all the total H
2
O and CO
2

pairs which would produce the corresponding sequence of MI’s,
by taking into account the uncertainti
es in the analytical data and in model predictions. These total
H
2
O and CO
2

contents refer to the liquid+gas phase mass only. The corresponding fields for the
1944 eruption samples, which account for different amounts of crystallisation from the 1944 liqui
d
in MI’s within dunitic cumulates, are also shown. Total H
2
O and CO
2

concentrations in these fields
refer to liquid+gas+crystal phase mass.

All of the fields in Fig. 3 except that referring to MI’s in dunitic cumulates from 1944 eruption
are open towards
the height. The reason is that in order to close the field it is required that MI data
cover conditions from high to low volatile contents. Unfortunately, this is not the case, since MI
studies are normally carried out with the intention to detect the high
est volatile contents. The single
low pressure datum in Fig. 2 referring to interstitial liquid in cumulate dunite allows closing the
corresponding field in Fig. 3. That field corresponds to possible total H
2
O and CO
2

pairs consistent
with all data from 19
44 cumulate dunite. If the low pressure volatile contents were formed under
open system degassing conditions, that field represents a minimum estimate of total H
2
O and CO
2

contents. All of the fields in the figure, except that referring to 1944 MI’s in leu
cite, are closed at
the bottom. This implies that total volatile contents lower than those corresponding to each field are
not consistent with the bulk of the corresponding MI data. Having made a conservative estimate of
the uncertainties implies that thes
e minimum estimates of total volatile contents are also
conservative.

One relevant information from Fig. 3 is that the relatively abundant 1906 and 1944 data show
minimum total CO
2

contents of 3


7 wt%, and minimum total H
2
O contents of 3


4 wt%. Larger
total H
2
O contents imply larger total CO
2
, according to the shape and slope of each of the fields in
the figure. A second relevant information derives from the distinction between the various fields
corresponding to different sets of data. This distinction

means a different total CO
2
/H
2
O mass ratio
in magma. The lowest ratio is associated with melt inclusions in dunite cumulates. This ratio
EXPLORIS Deliverable D2.8


15

increases progressively for inclusions in 1906 olivine, 1944 olivine, and 1944 leucite, for which it is
maximum. Data
from the AD 79 eruption correspond to a wide possible range of CO
2
/H
2
O mass
ratios largely overlapping with those of 1944 dunite and 1906 olivine melt inclusions.

Figure 4 shows the pressure and gas phase composition corresponding to all MI data considered

here. The corresponding depth is obtained by assuming an average rock density below Vesuvius of
2600 kg/m
3
. Most data correspond to a depth range between 10 and 20 km, with an upper limit at 8


10 km representing a sort of barrier below which the data cl
uster. This depth corresponds closely
to that of a prominent converted P
-
to
-
S phase obtained by 2D seismic tomography, and interpreted
as the possible top of an extended magma body below Vesuvius (Zollo et al., 1996, 1998; Auger et
al., 2001). CO
2

concentr
ation in the gas phase for these high pressure MI’s range 70


95 wt%, with
most data clustering above 80 wt%. The two high total CO
2
/H
2
O mass ratio inclusions in leucite
from the 1944 eruption, and one MI from the AD 79 eruption, correspond to a depth of
4
-

7 km,
just above or at the contact with the inferred bottom of the carbonatic sequence (5
-
6 km of depth),
in good agreement with estimates on fluid inclusions hosted in clinopyroxenes found in cumulates
and subeffusive nodules from 3800 y.B.P., 79 AD a
nd 472 AD Vesuvius plinian events (Belkin and
De Vivo, 1993). Inclusions in leucite correspond to the highest CO
2

concentration in the gas phase
of 95


97 wt%. Finally, the relatively low volatile concentrations in the interstitial liquid of the
dunitic c
umulate correspond to a depth of only about 1.4 km, and were probably formed during
continuous gas exsolution from the liquid upon magma rise. Accordingly, the corresponding CO
2

concentration in the gas phase results to be the lowest, in the range 58


75
wt%.

An important implication of the high to very high CO
2

concentrations in the gas phase shown in
Fig. 4 is that the stability curve of leucite for Vesuvius magmas is strongly shifted towards large
pressure/depth conditions. The curve reported in Fig. 4
refers to H
2
O fugacity of 100 MPa, above
which leucite is commonly assumed to be unstable in Vesuvius magmas at magmatic temperatures
(1050


T


1200 °C). Although this may represent an erroneous conclusion due to the shrinking of
the leucite field as H
2
O

enters the liquid in the nepheline
-
kalsilite
-
SiO
2

system
ref.some
, it is known
from experiments on 1944 volcanics that leucite can be the primary phase at liquidus under CO
2
-
rich conditions (Trigila and De Benedetti, 1993). In the range of conditions perta
ining to the
considered melt inclusions, leucite can easily start forming down to a pressure of 400 MPa, or to a
depth of 15 km. Leucite phenocrystals were indeed observed in 1944 lava fountain deposits
carrying clinopyroxenes and olivines hosting the vol
atile
-
rich MI’s (Marianelli et al., 1999)
.

Accordingly, the common and simplistic use of leucite crystals as an indicator of low pressure (i.e.
P
tot



1 kbar) crystallisation conditions is not justified.

The picture that emerges for magma below Vesuvius fr
om the major XX century eruptions, and
possibly for the AD 79 eruption, is that of a large reservoir, at least 7


8 km thick and having its
top at 8


10 km depth, hosting a large carbon dioxide content of at least some wt%, comparable to,
or even larger
than, the corresponding water content, with a gas phase occupying a large portion of
total volume and dominantly made of carbon dioxide. A second minor reservoir appears at the
shallow depth of 4


7 km. The progressive CO
2

to H
2
O enrichment from dunite to

1906 to deep
1944 to shallow 1944 magma is not simply consistent with an open system degassing path, since
loss of CO
2
-
rich gas would rather result in progressive decrease of total CO
2

relative to H
2
O. A
progressive CO
2

accumulation from a CO
2
-
rich gas ph
ase bubbling through the deep magma body
and locally re
-
equilibrating with the melt phase could instead explain such a trend. The large CO
2

enrichment in MI’s from shallow crystallising leucite is consistent with this view, although a
contribution by decar
bonation reactions at the contact with the carbonatic rocks of the Campanian
Plain can not be ruled out. It is worth noting that our barometric estimate for these inclusions is
consistent with that from a pyroxene
-
hosted MI from the 79 AD event and with th
e maximum depth
of metamorphism (5
-
6 km) inferred by carbonate ejecta from Vesuvius Plinian eruptions (Barberi
and Leoni, 1980). Nevertheless, this latter depth value must not be taken as representative of the
bottom of magma chambers feeding plinian event
s, but simply as the bottom of the carbonatic
sequence. This implies that roots of a unique magma chamber can extend downward this depth,
coherently with our barometric estimates. Moreover, the decarbonation contribution to the observed
EXPLORIS Deliverable D2.8


16

CO
2

enrichment is e
xpected to be of minor entity for the investigated products, since: i) stable
isotope behaviour shows that the process of skarn formation at the wall of the 79 AD Vesuvius
magma chamber was dominated by the infiltration of magmatic fluids, and only to a mi
nor extent
by decarbonation of the host limestones (Fulignati et al., 2005); ii) a very limited mass exchange
between magma and carbonate wall rocks was estimated for the 1944 magma chamber, generating a
solidification front and a skarn shell effectively i
solating the interior of the chamber from new
inputs of contaminants from the surroundings carbonate rocks (Del Moro et al., 2001). These
arguments confirm previous studies dismissing large
-
scale assimilation of shallower sedimentary
carbonatic formations
beneath Vesuvius (Taylor et al., 1979; Civetta et al., 1981; Cortini and
Hermes, 1981; Ayuso et al., 1998), and lead to share the conclusion, also corroborated by Helium
isotope evidences (Martelli et al., 2004), that most of CO
2

is exsolved from magma and

originated
in a contaminated mantle (Federico et al., 2001). Since Vesuvius magmatic rocks lie on a general
trend about mantle contamination following the Apennine subduction (Martelli et al., 2004), we
conjecture that the large CO
2

abundances recorded in

Vesuvius MI’s might indeed represent a
common feature with other magmatic settings of this part of Italy, likely to sustain the huge
emissions of deep CO
2

spread all over the territory (Chiodini et al., 2000).

EXPLORIS Deliverable D2.8


17













































Fig. 1a

Fig. 1b

Fig. 1c

EXPLORIS Deliverable D2.8


18

Fig. 2

EXPLORIS Deliverable D2.8


19

Fig. 3

EXPLORIS Deliverable D2.8


20


Fig. 4