Lectura Obligatoria - Bonadonna, C. 2013 - Modeling Tephra Sedimentation From Volcanic Plumes

Chapter 9
Modeling tephra sedimentation from

volcanic plumes
Costanza Bonadonna and Antonio Costa
Overview 9.1 Introduction

Tephra erupted in volcanic plumes can be Explosive volcanic eruptions have intrigued scien-
transported over distances of thousands of tists because of their dramatic display of physical
kilometers, causing respiratory problems to processes, their crucial role in the geological evo-
humans and animals, serious damage to build- lution of Earth, and their potentially catastrophic
ings and infrastructure, and affecting economic consequences for society. A key way of improving
sectors such as aviation, agriculture, and tour- our understanding of explosive volcanism is to
ism. Models with different degrees of com- study the resulting pyroclastic deposits, which
plexity have been developed over the last few often represent the only direct evidence of explo-
decades to describe tephra dispersal. Depending sive eruptions. Tephra deposits retain a consid-
on the application, different simplifications and erable amount of information about the nature
assumptions can be introduced to make the of the eruption, such as erupted mass, bulk
problem tractable. Highly sophisticated models grain-size distribution, and eruption intensity.
are not suited for the computationally expensive However, tephra falls also represent significant
probabilistic calculations required by long-term hazards for people living close to active volca-
hazard assessments. In contrast, the simplified noes. These hazards include collapse of build-
models typically used for probabilistic assess- ings, disruption to water and electricity supplies,
ments have to compromise the sophistication disruption to transportation networks, as well as
of the physical formulation for computational health hazards from respirable ash, crop pollu-
speed. A comprehensive understanding of tephra tion, and lahar generation. Developing an under-
deposits and hazards can only result from a crit- standing of tephra fall is crucial to public safety.
ical and synergistic application of models with In this chapter tephra is used in the original sense
different levels of sophistication, ranging from of Thorarinsson (1944) as a collective term for all
purely empirical to fully numerical. A review of particles ejected from volcanoes, irrespective of
the main approaches to tephra dispersal model- size, shape, and composition, whereas tephra fall
ing is presented in this chapter. indicates the process of particle fallout.
Modeling Volcanic Processes: The Physics and Mathematics of Volcanism, eds. Sarah A. Fagents, Tracy K. P. Gregg, and Rosaly M. C.
Lopes. Published by Cambridge University Press. © Cambridge University Press 2013.
Downloaded from https://www.cambridge.org/core. Durham University Library, on 03 Oct 2018 at 00:39:49, subject to the Cambridge Core terms of use, available at
https://www.cambridge.org/core/terms. https://doi.org/10.1017/CBO9781139021562.009
174 CO STAN ZA BONADONNA AND ANTONIO CO S TA
The preservation of tephra deposits is typ- extended source (i.e., fissure eruption or pyro-
ically incomplete. In most eruptions proximal clastic density current), eventually develops into
areas are buried or collapsed, and distal depos- a turbulent buoyant current whose dynamics are
ition occurs in the sea or becomes eroded. For strongly controlled by the degree of interaction
small eruptions, the whole tephra deposit may with the atmosphere. If the plume upward vel-
be eroded away within a few years of deposition. ocity is much stronger than the wind velocity, the
As a result, empirical, analytical, and numerical initial jet phase (gas thrust) evolves into a vertical
models have been developed to allow quantita- buoyant column that then eventually spreads
tive interpretation of tephra deposits and to fully laterally as a gravity current (i.e., umbrella cloud)
understand the nature of ancient eruptions that around the neutral buoyancy level Hb (i.e., strong
are incompletely preserved. Dedicated analytical plume; Fig. 9.1(a)). In contrast, if the wind vel-
and numerical models have also been produced ocity is much stronger than the plume upward
to investigate plume dynamics and particle velocity, the turbulent current will be bent over
sedimentation, and to provide long-term assess- above the basal jet before spreading laterally
ments for land-use planning and rapid response around Hb (i.e., weak plume; Fig. 9.1(c)). It is import-
during volcanic crises. Model validations have ant to distinguish between vigorous and low-energy
shown good agreement with field data, which weak plumes: both plumes are bent over by the
justifies the use of these models for hazard wind but they are characterized by different ener-
applications. getics (e.g., steepness of plume trajectory rela-
This chapter describes: (1) empirical and tive to wind speed). Typically, vigorous weak plumes
analytical models used to determine eruption characterize the beginning of low-intensity sus-
parameters, such as column height, eruption tained eruptions (e.g., Ruapehu, 17 June 1996;
duration, magnitude, and intensity, (2) analyt- Bonadonna et al., 2005a), whereas low-energy weak
ical and numerical models developed for the plumes characterize the last phase of an eruption
study of the dynamics of volcanic plumes and when wind eventually dominates and the cloud
particle sedimentation, and (3) models com- starts propagating as a lens of aerosol (e.g., Mount
monly used for hazard assessments and fore- St. Helens, 22 July 1980; Sparks et al., 1997).
cast of plume spreading. Model assumptions Volcanic clasts (juvenile and lithic frag-
and caveats are discussed, and a key case study ment) are carried up within the turbulent cur-
is presented to facilitate comprehension of the rent according to their settling velocity, which
application of the models described (the 22 July depends on both particle and atmospheric char-
1998 explosive eruption of Mt Etna, Italy; Coltelli acteristics. When particle settling velocities are
et al., 2006; Scollo et al., 2008a). larger than the upwards component of the tur-
bulent current, they fall out and are advected
by local winds. Particles that are sufficiently
small will typically aggregate into micron- to
9.2 Plume dynamics and particle millimeter-sized clusters having greater settling
sedimentation velocities (Sparks et al., 1997). In addition, the
deposition of fine particles is also enhanced by
Before describing the models used to characterize pronounced convective instabilities and mam-
tephra dispersal and deposits, it is important to matus that often form at the base of the sedi-
understand some basic concepts of plume dynam- menting turbulent current (Bonadonna et al.,
ics and particle sedimentation (see Chapter 8 for 2002b; Durant et al., 2009). As a result, the char-
a detailed review of plume dynamics). Volcanic acteristics of tephra deposits are the result of
plumes are typically associated with explosive plume dynamics (e.g., plume height, velocity
activity and consist of a mixture of lithics (wall profile, weak-plume vorticity), particle param-
rock), volcanic gas, and juvenile particles (frag- eters (e.g., size, density and shape), atmospheric
mented magma), which, whether generated characteristics (e.g., wind field, atmospheric
from a point source (i.e., single vent) or from an density and viscosity) and sedimentation
MODELING T EPH RA S EDI MENTAT IO N F RO M VOL C A NIC P LUMES 175
(a) turbulent fallout from the plume margins and

Strong plume turbulent, intermediate, and laminar fallout
Ht
from the umbrella cloud. These fallout regimes
plume corner (x0) turbulent
result in four exponential segments in semi-log
Hb
current
plots of deposit thickness vs. distance from vent
Hcb (Fig. 9.1(b)). Resulting isopach or isomass maps
convective
range from nearly concentric in the absence of
region wind (Fig. 9.2(a)) to strongly elongated in the
gas thrust
case of significant wind advection (Figs. 9.2(b,c)).
Because of the strong dependence of the set-
tling velocity and Reynolds number on particle
(b) size, the final deposit morphology is also con-
1000
Seg0
trolled by the initial grain-size distribution. In
Thickness (arbitrary units)
100 comparison to strong plumes, tephra fall from

10
weak plumes is characterized by a lack of up-
Seg1 wind sedimentation, more pronounced prox-
1
Seg2
imal thinning due to the bent-over structure
0.1 Seg3
of the turbulent current, and narrower tephra
0.01 x0 deposits due to the presence of vortex struc-
0.001 tures underneath the plume. For example,
0 20 40 60 80 100
Distance (arbitrary units)
blocks, lapilli, and a proportion of coarse ash
were deposited from the rising phase of the 17
(c) June 1996 Ruapehu weak plume (within ~30 km
Weak plume of the vent), whereas the horizontal spreading
was characterized by deposition of coarse and
Ht
fine ash (Bonadonna et al., 2005a). The resulting
nt t
rre en
WIND ADVECTION
deposit (Fig. 9.2(c)) shows maximum accumula-

cu bul
Hb
r
tu
Hcb tion along the dispersal axis, but as soon as the

plume was no longer sustained the eruption lost
its vigor and the spreading turbulent current
gas thrust
started bifurcating due to vorticity conservation
(~800 km from the vent).
x0
Figure 9.1 Diagrams showing (a) the main characteristics

of a strong volcanic plume: Hb is the neutral buoyancy
9.3 Empirical and analytical models
level of the plume, Ht and Hcb are the top and the base of used for the characterization of
the spreading turbulent, umbrella cloud (modiied from tephra deposits
Bonadonna and Phillips, 2003); (b) sedimentation from plume
margins (Seg0) and umbrella cloud (Seg1 = turbulent regime,
Field investigations of tephra deposits are cru-
Reynolds number, Re > 500; Seg2 = intermediate regime,
6 < Re < 500; Seg3 = laminar regime, Re < 6) of a strong cial for characterization of volcanic eruptions
plume (fallout regimes deined as in Bonadonna and Phillips, and their hazards. In particular, distributions of
2003; (c) the main characteristics of a weak volcanic plume tephra thickness and mass per unit area (isopach
(modiied from Bonadonna et al., 2005a). and isomass maps; Fig. 9.2) are necessary for
estimating erupted volume or mass (Pyle, 1989;
processes (e.g., particle aggregation, convective Fierstein and Nathenson, 1992; Bonadonna and
instabilities, mammatus). Houghton, 2005), whereas the distribution of lar-
Tephra fall from strong plumes is charac- gest clasts (isopleth maps) is typically used for
terized by four main sedimentation regimes: estimating column height and wind speed at the
Figure 9.2 (a) Isopach map of

the Basal Fall of Pululagua 2450 b.p.,
Ecuador (cm) (Volentik et al., 2010);
(b) isopach map of tephra deposit of
the 18 May 1980 eruption of Mount
St. Helens, USA (mm) (Sarna-Wojcicki
et al., 1981); (c) isomass map of the
tephra deposit of the 17 June 1996
eruption of Ruapehu, New Zealand
(kg m−2) (Bonadonna et al., 2005a).
time of the eruption (Carey and Sparks, 1986). a result, the logarithm of tephra thickness can
Isopach, isomass and isopleth maps can also be be described by straight lines (i.e., exponential
used to determine vent location and to classify segments) when plotted against distance from
eruptive style (Walker, 1973; Walker, 1980; Pyle, vent or square root of the area enclosed by each
1989). Mass eruption rate and the duration of the isopach ( A ):
sustained phase of the eruption can be calculated
from plume height and erupted mass, respect- T To exp (−kk A ) (9.1)
ively (Sparks, 1986; Wilson and Walker, 1987;
Carey and Sigurdsson, 1989). Inferences of frag- where To is the maximum thickness of the
mentation mechanisms can also be made from deposit and k defines the rate of thinning of the
the study of particle sizes (Kaminski and Jaupart, deposit (i.e., slope of the associated exponential
1998; Neri et al., 1998; Zimanowski et al., 2003). segment). All notation is summarized in Section
The empirical and analytical models used for 9.9. Assuming that isopachs have elliptical
these purposes, together with their assumptions shapes, the volume of tephra deposit is:
and limitations, require thorough analysis to
assess the variability of resulting eruption param- V .08 To bt2
(9.2)
eters. This is crucial not only because these erup-
tion parameters are used to characterize volcanic where bt ( ) (k ) .
eruptions, but also because they are used as input Fierstein and Nathenson (1992), Pyle (1995),
to numerical models and to construct potential and Bonadonna and Houghton (2005) developed
activity scenarios for hazard assessment. this method to account for abrupt changes in
the rate of thinning of some tephra deposits:
9.3.1 Determination of erupted volume
based on the assumption of 2T10 ⎡ k S + 1 k S + 1⎤
V = + 2T10 ⎢ 2 12 − 1 12 ⎥ exp ( − k1S1 )
exponential thinning of tephra k 1
2
⎣ k2 k1 ⎦
deposits ⎡ k S + 1 k S + 1⎤
This approach was introduced by Pyle (1989), + 2T20 ⎢ 3 22 − 2 22 ⎥ exp ( − k2S2 ) + ... + 2T(n − 1)
⎣ k3 k2 ⎦ 0
adopting the preliminary observation of

⎡ k n S(n − 1) + 1 k(n − 1)S(n 1) + 1 ⎤
Thorarinsson (1954) that both thickness and
grain-size of tephra deposits follow an expo-
×⎢
⎢⎣ k n2
−
k(2n − 1) ⎥⎦
(
⎥ exp − k(n − 1)S(n 1) )
nential decay with distance from the vent. As (9.3)
where Tn0, kn, and Sn are the intercept, slope, and field data. In particular, the power-law best fit
position of the break in slope of line segment n. can be described as:
Their approach to estimating volume by defin-
m
ing several exponential segments (i.e., different ⎛ A ⎞
( A)
−m
values of k from proximal to distal portions of T T0 ⎜ 0 ⎟ = C pl with C pl T0 A0m 2

⎝ A ⎠
the deposit) is consistent with observations of
well-preserved tephra deposits (Hildreth and (9.5)
Drake, 1992; Scasso et al., 1994) and with the
where Cpl and m are the power-law coefficient
results of some analytical models (Bursik et al.,
and exponent respectively. The associated vol-
1992a; Sparks et al., 1992; Bonadonna et al.,
ume can be calculated as:
1998).
The approach of Pyle (1989) was also modi-
( ),
2C pl 2 m 2 m
fied to estimate erupted volume for cases in V= Adist A (9.6)
2 −m
which only one proximal isopach can be defined
based on the available data (Legros, 2000). This
which is equivalent to:
technique was derived from empirical investi-
gation of 74 tephra deposits and gives estimated ⎛ A 2− m ⎞
2
minimum volumes of the same order of magni- V = T0 A0 ⎜ dist − 1⎟ , (9.7)
2−m ⎝ A0 ⎠
tude as for the case in which only the first two
segments on semi-log plots of thickness T vs.
square root of area A are available (Seg0 and
Seg1 in Fig. 9.1(b)): where A0 and Adist are two arbitrary inte-
gration limits. In particular, Adist ideally repre-
V 3 69 Tx A x (9.4) sents the area of the isoline of zero thickness,
whereas A0 is area enclosed by the isoline cor-
where Ax (m2) is the area enclosed within the responding to the maximum deposit thickness.
isopach with thickness Tx (m). Sulpizio (2005) Note that, to guarantee convergence in the limit
presented additional techniques for the deter- of large distances, m has to be > 2.
mination of distal volume when most distal
data are missing, based on extrapolation of the Caveats
distribution of proximal deposits to distal areas Sensitivity analyses of volume calculations have
(up to thickness > 1 cm). Finally, Mannen (2006) shown that integration of less than three expo-
suggested an analytical method to derive the nential segments can underestimate deposit vol-
total erupted mass of relatively small eruptions ume when distal data are missing (Bonadonna
by adopting the model of Bursik et al. (1992a) and Houghton, 2005). For example, integration
and integrating two exponential segments of only two exponential segments described by
determined from isopleth maps. data within 10 km of the 1996 Ruapehu erup-
tive vent resulted in an underestimation by half
9.3.2 Determination of erupted of the actual deposit volume. Such an underesti-
volume based on the assumption mation does not affect the classification of the
of power-law thinning of eruption in terms of the volcanic explosivity
tephra deposits index (VEI), but is significant when simulating
Based on the results of analytical investigations tephra dispersal and compiling hazard assess-
of Bonadonna et al. (1998) and on observations ments. In contrast, the power-law fit is a good
of well-preserved deposits for which thinning approximation to well-preserved deposits and is
can be described by a power-law fit on a semi- consistent with theoretical models, but it is also
log plot of T vs. A , Bonadonna and Houghton problematic because integration limits have to
(2005) suggested deriving the total erupted vol- be chosen. In particular, the volume of deposits
ume by integrating the power-law best fit to having limited dispersal (m > 2) is very sensitive
to the choice of A0 but not to the choice of Caveats

Adist . In contrast, the volume of deposits hav- The method of Carey and Sparks (1986) provides
ing wide dispersal (m < 2) is very sensitive to a maximum column height from the distribution
Adist but not to A0 . Given that Adist is dif- of largest clasts (pumice and lithic fragments).
ficult to constrain, the power-law method may However, there are some important caveats to
be used for small deposits (m > 2) but is not rec- bear in mind when applying this technique.
ommended for widely dispersed deposits (m < 2) First, the distribution of largest clasts is not
when the outer limit of the deposit cannot be unique and is very sensitive to the techniques
accurately identified. Essentially, empirical used to determine average clast size at a given
fitting of poor data sets can be problematic, outcrop and to contour maximum values (Carey
especially for large eruptions, even with the and Sparks, 1986; Barberi et al., 1995). Second,
power-law method because there is no simple the method of Carey and Sparks (1986) is based
theoretical relationship between proximal and on two main assumptions: (1) a monotonic
distal thinning. Proximal deposition is con- decrease in vertical plume velocity within a
trolled by high-Reynolds-number particles, plinian column, and (2) a mid-latitude wind pro-
whereas distal deposition is controlled by low- file such as that described by Shaw et al. (1974).
Reynolds-number particles (Fig. 9.1(b)). The first assumption requires that this method
should be applied only to deposits from plinian
9.3.3 Determination of column height and subplinian (i.e., sustained) eruptions gener-
and wind speed ating strong plumes. In addition, Woods (1988)
Even though a buoyant eruptive column is has shown that large plumes might be character-
characterized by fluctuating vertical veloci- ized by super-buoyancy (see Fig. 8.5), for which
ties, plume studies have shown that the hori- the vertical velocity profile does not decrease
zontal profile of time-averaged vertical speed monotonically. Such an effect could result in
can be represented by a Gaussian distribution a premature loss of large clasts and, therefore,
symmetrical with respect to the plume axis higher plumes than predicted by Carey and
(Turner, 1979). From comparison between this Sparks (1986). The method of Carey and Sparks
Gaussian-distributed plume velocity and the (1986) therefore gives the best results for small
settling velocities of volcanic particles, Carey clasts (< 32 mm; Papale and Rosi, 1993; Barberi
and Sparks (1986) defined a series of theoretical et al., 1995; Rosi, 1998). Finally, wind direction
“envelopes” that support the particles within is typically very variable with height, and the
plumes. When the settling velocity of the particles profiles can show large discrepancies with that
exceeds the vertical velocity of the plume (charac- of Shaw et al. (1974). For example, Carey and
teristic of a given envelope), particles will leave Sigurdsson (1986) used a modified wind profile
the plume and deposit on the ground at distances to account for the direction inversion above the
that depend on release height, and wind speed tropopause that occurred during the 1982 erup-
and direction. As a result, the column height and tion of El Chichon.
wind speed can be derived by plotting the max-
imum downwind range vs. the crosswind range of
9.3.4 Determination of mass eruption
isopleth values (i.e., the length and half-width of
rate and eruption duration
individual isopleths). Carey and Sparks (1986) pro-
The mass eruption rate Ṁ (kg s−1) can be derived
vide plots that are based on a specific wind profile
from plume height H (m) by applying the semi-
(Shaw et al., 1974) and specific particle charac-
empirical formula of Wilson and Walker (1987)
teristics (i.e., size and density). In particular, the
wind profile is assumed to be uni-directional at H CM
C $ 1/ 4
(9.8)
all heights, with a maximum velocity (5–50 m s−1)
at the tropopause level (11 km), a linear decay to where the empirical factor C = 236 m kg−1/4 s1/4.
zero at ground level, and a value above the tropo- Eruption duration can be determined by divid-
pause of 3/4 of the maximum velocity. ing the total erupted mass by Ṁ .
Mass eruption rate can also be derived from and 1000 °C, and for both tropical and temper-
the column height H (m) using the analytical ate atmospheres. The model shows good agree-
model of Sparks (1986), which was based on ment with observed data. However, to compile
buoyant plume theory (BPT) (Morton et al., 1956; diagrams from more elaborate theory, Sparks
Settle, 1978; Wilson et al., 1978) and improved (1986) made assumptions about tropopause
by accounting for a varying adiabatic lapse rate height, surface temperature, temperature gra-
and atmospheric temperature. As a result, Ṁ and dient, wind profile, and air-entrainment mod-
H show a nonlinear correlation, which strongly els that need to be carefully verified prior to
depends on eruption temperature. application.
Caveats
Equation (9.8) holds only for circular-vent
plumes < 35 km high and is supported by the- 9.4 Models based on the
oretical investigations based on BPT (Morton Advection–Diffusion–
et al., 1956; Wilson and Walker, 1987) which
show that maximum plume height is roughly Sedimentation (ADS) equation
proportional to the fourth root of the heat injec-
tion rate, and therefore to the fourth root of the Models for tephra dispersal are based on the
mass eruption rate (see also Chapter 8, Section mass conservation equation with different
8.3.1). However, there are several limitations in degrees of simplicity, following either Eulerian
extending BPT to eruption plumes in a strati- or Lagrangian formulations. The Eulerian
fied atmosphere where buoyancy flux varies approach describes changes in the fluid at
with height and crosswinds significantly affect fixed points, whereas the Lagrangian approach
plume entrainment (Bursik, 2001; Ishimine, describes changes by following a fluid par-
2006; Carazzo et al., 2008). In addition, Eq. cel along its trajectory. Each approach is use-
(9.8) is strictly valid for a plume temperature ful for different applications. For example,
of ~800 °C, appropriate for andesitic magma. weather forecasting is based on the Eulerian
Basaltic magmas are typically hotter by at least approach (fixed measurement system) because
200 °C, and therefore, to achieve the same col- it uses data from fixed stations around the
umn height the corresponding mass discharge world. The Lagrangian approach is more use-
rates are lower for basaltic magmas (Carey and ful when describing the evolution of a given
Sparks, 1986; Sparks, 1986; Woods, 1988). For material as it moves within a certain fluid (e.g.,
example, Wehrmann et al. (2006) found that chemical modeling). Tephra dispersal is often
C = 295 m kg−1/4 s1/4 in Eq. (9.8) describes the described using both approaches. In particular,
relationship between Ṁ and H for a basaltic models commonly defined as Lagrangian are
plinian eruption of Masaya volcano, Nicaragua based on an Eulerian–Lagrangian approach,
(i.e., Fontana Lapilli). Scollo et al. (2007) and which describes the dynamics of single parti-
Andronico et al. (2008) found C = 247 m kg−1/4 s1/4 cles within an Eulerian flow field. In contrast,
and C = 244 m kg−1/4 s1/4 for the 2001 and 2002 Eulerian models consider the particle phase
eruptions of Etna volcano, respectively. Finally, and the flow field as two continua. The govern-
it is important to bear in mind that Eq. (9.8) is ing equation derived from the mass conserva-
strictly valid only for values of maximum col- tion condition has the following form (Costa
umn height and therefore gives maximum et al., 2006):
values of Ṁ. As a result, the corresponding erup-
tion durations could be underestimated. In con- ∂cj ∂u x c j ∂u y c j ∂u z c j ∂v j c j
+ + + − =
trast, the model of Sparks (1986) can easily be ∂t ∂x ∂y ∂z ∂z
(9.9)
applied by using dedicated diagrams compiled ∂ u ′x c ′j ∂ u ′y c ′j ∂ uz′ c ′j
for sustained buoyant plumes with heights up − − − + S,
∂x ∂y ∂z
to 35 km, eruption temperatures between 400
where cj(x,y,z,t) = c̅j + c′j is the concentra- ∂Uc j ∂v j c j

∂v
− = 0, (9.11)
tion of particle class j (with c′j the turbulent ∂s ∂z
fluctuation and cj̅ the ensemble average),
(ux u y , uz) ( )
u x + ux u y + uy u z + uz is the ambi- where U denotes the current velocity in the
ent fluid velocity, and vj(x,y,z) is the terminal vel- s-direction (e.g., direction along the vertical
ocity of a particle class j. The first term on the plume or the horizontal umbrella cloud). Under
left-hand side of Eq. (9.9) represents the time rate these conditions, and integrating for the sedi-
of change of the average concentration cj̅ (i.e., mentation rate derived by Martin and Nokes
the transient term), whereas the second, third, (1988), the total mass of particles, M (kg), of a
and fourth terms represent advection (i.e., wind given size fraction carried by the spreading cur-
transport) and the fifth term describes sedimen- rent at a certain distance x1 is:
tation. The first three terms on the right-hand
side of Eq. (9.9) represent the diffusive transport ⎪⎧ v w ⎪⎫
x1
due to the atmospheric turbulence, whereas the M M0 exp ⎨ − ∫ Hb dx

d ⎬ (9.12)
⎩⎪ xo Q ⎭⎪
fourth term, S(x,y,z,t), denotes the source (i.e.,
the mass flux of particle class j injected per unit
volume and unit time). For most applications where M0 (kg) is the initial mass for a given
vj(x,y,z) can be considered a function of height grain size injected into the current, vHb (m s−1)
z only. Turbulent fluxes are given by the prod- is the particle terminal velocity at the neu-
uct of the fluctuation terms on the right-hand tral buoyancy level Hb, w (m) is the maximum
side of Eq. (9.9). Typically, the simplest approach crosswind width of the current at the source, Q
consists of expressing turbulent fluxes as pro- (m3 s−1) is the volumetric flow rate into the cur-
portional to the gradient of average concentra- rent at the neutral buoyancy level, and xo (m)
tion, e.g., for the x-direction: is the plume corner position (Fig. 9.1; Bursik
et al., 1992a). These models are supported by
∂c j experimental data (Sparks et al., 1991) and
u x′ c ′j ≈ − K x (9.10)
∂x have given crucial insights into tephra deposit
thinning. For example, the models of Bursik
where the coefficient Kx is the x-component of et al. (1992a) and Sparks et al. (1992) provided a
the turbulent diffusion tensor (typically in the theoretical basis for empirical methods (Pyle,
free atmosphere, KV/KH << 1, where KH = Kx = Ky 1989; Fierstein and Nathenson, 1992) of cal-
and KV = Kz). culating erupted volumes. Furthermore, these
analytical models showed that tephra fall from
9.4.1 One-dimensional analytical the plume margins and from the umbrella
sedimentation models cloud can be described by two exponential
Depending on the application, if some approxi- segments on a semi-log plot of thickness vs.
mations hold (e.g., horizontally uniform wind distance from vent, with a break in slope coin-
field, constant diffusion coefficient, negligible ciding with the plume corner. Bonadonna
vertical motion and diffusion), Eq. (9.9) can be et al. (1998) described sedimentation from the
simplified and solved analytically. For instance, umbrella cloud by also accounting for particle
one-dimensional ADS models are typically used Reynolds number and showed that the associ-
to investigate particle sedimentation along the ated tephra deposits can be better described
dispersal axis, assuming negligible diffusion, by three exponential segments (Fig. 9.1(b)). All
steady plume conditions, and a single coordin- of these results have crucial implications for
ate direction (s) along the direction of the carry- the determination of the total erupted volume
ing medium (Bursik et al., 1992a,b; Sparks et al., (Pyle, 1990; Rose, 1993). Koyaguchi and Ohno
1992; Koyaguchi, 1994; Bonadonna et al., 1998; (2001a,b) also used this approach, in combin-
Koyaguchi and Ohno, 2001a,b; Bonadonna and ation with dedicated inversion solutions of
Phillips, 2003). As a result, Eq. (9.9) becomes: observed data, to determine the expansion
rate of an umbrella cloud, the grain-size dis- ∂ cj ∂ cj ∂ cj ∂ v jc j ∂ 2c j ∂ 2c j

tribution at the top of the eruption column, + ux + uy − = KH + KH +S
∂t ∂x ∂y ∂z ∂x 2
∂ y2
and the amount of fine ash dispersed in the
atmosphere, and found good agreement with (9.13)
satellite observations. where, for simplicity, we have eliminated the
Further developments of these models have over-bar symbol to denote the average quan-
shown that the main processes affecting par- tities. A solution of Eq. (9.13) is given by a
ticle sedimentation are the effects of wind Gaussian distribution (e.g., Suzuki, 1983; Pfeiffer
advection on the spreading of the umbrella et al., 2005). The total mass on the ground is com-
cloud and on particle transport, and particle- puted as the sum of the contributions of each of
aggregation processes in the case of ash-rich the point sources distributed above the ground
plumes and concentrated flows (Bursik et al., and of each particle class.
1992b; Bonadonna and Phillips, 2003). In par- Commonly, ADS models are based on
ticular, wind advection typically shifts the pos- empirical parameters, such as the diffusion
ition of breaks in slope downwind, whereas coefficient KH and column shape parameters
aggregation processes allow fine particles to introduced for describing the term S. As a
fall in the turbulent and intermediate fall- result, they must be validated and calibrated
out regimes with the result that: (1) breaks in with field data for specific eruptions before
slope can be shifted either closer to the vent or they can be used for a reliable hazard assess-
further downwind depending on the size and ment. However, the advantage of these models
density of the aggregates, (2) the thinning rate is the simplicity of the physical parameter-
is controlled by the amount of aggregating ization and, therefore, the high computation
particles, and (3) small aggregates are likely to speed. This allows for comprehensive prob-
generate secondary maxima of mass accumu- abilistic analysis of the associated inputs and
lation when advected (Bonadonna and Phillips, outputs, and for solution of inverse problems
2003). Bonadonna et al. (2005a) further devel- to estimate eruption parameters, such as total
oped Eq. (9.12) to account for the variation of erupted mass and column height. Key find-
volumetric flux with distance from the vent to ings include the great sensitivity of ADS mod-
describe sedimentation from a vigorous weak els to both erupted mass and column height,
plume. which justifies the use of inversion solutions
for estimating these parameters (Connor and
Connor, 2006; Scollo et al., 2008a,b; Volentik
9.4.2 Two-dimensional analytical et al., 2010).
sedimentation models
Two-dimensional models are based on analyt- 9.4.3 Three-dimensional numerical
ical solution of the ADS equation under the models for particle sedimentation
assumptions of constant and isotropic atmos- When the simplifying assumptions made to
pheric diffusion and of negligible vertical wind derive Eq. (9.13) are no longer valid, or when
velocity and vertical diffusion (Suzuki, 1983; there is a need to describe three-dimensional
Armienti et al., 1988; Glaze and Self, 1991; dispersion of volcanic clouds within the
Hurst and Turner, 1999; Connor et al., 2001; atmosphere, fully numerical models must
Bonadonna et al., 2002a, 2005b; Folch and be adopted. One example requiring three-
Felpeto, 2005; Macedonio et al., 2005; Pfeiffer dimensional treatment is when most of the
et al., 2005; Connor and Connor, 2006). The erup- transport occurs within the atmospheric
tion column is typically described as a vertical boundary layer (ABL), i.e., the part of the
line source that can be characterized by vari- troposphere that is directly influenced by the
ous mass distribution functions ranging from presence of the Earth’s surface and which
uniform to exponential (e.g., Suzuki, 1983). As a responds to surface forcing with a timescale of
result, Eq. (9.9) becomes: ≤ 1 hour (Stull, 1988). Due to both topographic
effects and rapid temporal variations of wind ∂c j ∂c j ∂c j ∂c j

and temperature fields, turbulent tensor com- + ux + uy + (u z − v j ) =
∂t ∂x ∂y ∂z
ponents are significantly more complex in the
∂ ⎛ ∂ ⎞ ∂ ⎛ ∂ ⎞
ABL than in the higher (i.e., free) atmosphere. ⎜ KH cj⎟ + KH cj +
∂x ⎠ ∂y ⎜⎝ ∂y ⎟⎠
(9.14)
∂x ⎝ ∂x
Another case requiring a numerical approach
is the tracking of volcanic ash clouds for ∂ ⎛ ∂ ⎞ ∂v j →
⎜ KV cj⎟ + cj − j∇ ⋅ u + S .
diversion of aircraft flight paths routinely per- ∂z ⎝ ∂z ⎠
∂z ∂z
formed by the Volcanic Ash Advisory Centers
(VAACs) (Witham et al., 2007). The terrain-following wind speed compo-
However, depending on the application, nents u
→
( )
u x u y u z can be passed through
several simplifying assumptions can be intro- meteorological models such as CALMET (Scire
duced to make the problem tractable for prac- et al., 2000), which is used for assimilating and
tical purposes, even for this category of models. interpolating short-term forecasts (or re-analysis)
For example, the VAAC particle-tracking mod- from mesoscale meteorological prognostic mod-
els are mainly used to describe the atmos- els. The vertical component of the turbulent
pheric transport of volcanic ash for aviation diffusivity tensor KV can be described on the
safety, but, except for a few cases (Tanaka and basis of similarity theory in the ABL, and by the
Yamamoto, 2002), they are not used to calcu- Richardson number (ratio of thermally-produced
late ground deposition. Other examples of turbulence and turbulence generated by verti-
VAAC three-dimensional, time-dependent cal shear) in the free atmosphere. The horizon-
Eulerian models are VAFTAD (Volcanic Ash tal component KH can be described following a
Forecast Transport And Dispersion; Heffter large eddy simulation (LES) approach that con-
and Stunder, 1993), CANERM (CANadian sists of solving large scale motions of the flow
Emergency Response Model; D’Amours, 1998) and modeling the effect of the smaller univer-
and MEDIA (Eulerian Model for DIspersion in sal scales using a sub-grid scale (SGS) model. The
the Atmosphere; Sandu et al., 2003). The last generic particle class j is defined by three values
model, used by the Toulouse VAAC, simulates characterizing each particle (diameter, density,
the effects of advection by the average wind, and shape factor). Several semi-empirical param-
diffusion of particles by thermal and dynam- eterizations can be used for calculating particle
ical turbulence, rainout (occurring inside the terminal settling velocities. The source term S
cloud) and washout (occurring below the cloud) (i.e., the amount of mass injected per unit vol-
of particles by precipitation, and sedimentation ume and unit time) can be described as a point
due to gravitational settling. Examples of VAAC source, as a Suzuki distribution (Suzuki, 1983;
models based on Lagrangian formulations are Pfeiffer et al., 2005), or through a buoyant plume
NAME (Numerical Atmospheric-dispersion model (Bursik, 2001). The last option involves
Modelling Environment; Ryall and Maryon, the solution of the one-dimensional, radially
1998), PUFF (Searcy et al., 1998), and HYSPLIT averaged, governing equations that describe
(Hybrid Single Particle Lagrangian Integrated the convective region of an eruption column,
Trajectory Model; Draxler and Hess, 1998). Only over which the mass is released and distributed
a few models, such as the three-dimensional, in several layers according to the particle ter-
time-dependent Eulerian FALL3D (Costa et al., minal velocity. These equations are intimately
2006; Folch et al., 2009) and the Lagrangian coupled with the wind field, which, for small
VOL-CALPUFF (Barsotti and Neri, 2008; Barsotti plumes, may cause substantial bending over of
et al., 2008) were designed for both particle the plume. The numerical solution is based on a
transport and particle sedimentation. second-order finite differences scheme. Particle
In particular, FALL3D solves the ADS equa- sedimentation at the bottom of the computa-
tion for each particle class concentration in a tional domain (i.e., ground level) is calculated as
terrain-following coordinate system: the temporal integral of the outgoing mass flux
from this surface (Costa et al., 2006). In a recent the technique used, on deposit exposure, and
study (Folch et al., 2008), FALL3D was general- on data distribution and density. Most erupted
ized to the mesoscale-synoptic domain and cou- volumes derived from field data should be con-
pled with the Weather Research and Forecasting sidered minimum values unless the data sets
(WRF) meteorological model (Michalakes et al., extend hundreds of kilometers from the vent
2005; www.wrf-model.org). (the higher the plume, the larger the deposit
Finally, it is worth mentioning another cat- to be investigated). A review of several meth-
egory of models designed to describe in detail the ods can be found in Froggatt (1982), whereas
evolution of the eruption plume from its rise to the most recent techniques are summarized in
its collapse. Given the high degree of complexity Section 9.3. In addition, recent applications of
of such models, they are usually applied over rela- inversion techniques to analytical models have
tively small horizontal domains (up to few tens of shown promising results. Specifically, mass
kilometers). Some of these models are focused on per unit area and particle-size data from indi-
describing the dynamics and thermodynamics of vidual outcrops are inverted through the use
the mixture of hot gases and particles (Dobran of two-dimensional analytical models to derive
et al., 1993; Esposti Ongaro et al., 2007), whereas eruption parameters, such as erupted mass
other models such as ATHAM (Active Tracer High and column height (Connor and Connor, 2006;
Resolution Atmospheric Model; Herzog et al., Scollo et al., 2008a; Volentik et al., 2010).
1998; Oberhuber et al., 1998) are more appropri- Column height is very important for defining
ate for dealing with the chemical interactions the source term and is related to eruption inten-
and microphysical processes of volcanic and sity (i.e., mass flux). The best evaluation of col-
cloud water, cloud ice, rain, and graupel. umn height comes from well-documented and
calibrated direct observations. Column height
can also be estimated through analysis of satel-
9.5 Limitations of input lite images, based on geometry (cloud shadow
clinometry), thermal infrared (IR) data (using
parameters and a cloud-top temperature/temperature-profile
parameterizations adopted method), and correlation of cloud trajectory
by ADS models with meteorological motion (cloud stereoscopy)
(Holasek and Self, 1995; Prata and Turner, 1997;
Regardless of the complexity of different sedi- Glaze et al., 1999; Prata and Grant, 2001). Field
mentation models, the reliability and uncertain- studies have also shown that the derivation of
ties of the associated outputs strongly depend on plume height using the method of Carey and
the reliability and uncertainties of input param- Sparks (1986) gives fairly consistent results
eters (i.e., erupted mass, column height, total even for poorly exposed deposits (Wehrmann
grain-size distribution, meteorological data) and et al., 2006). Inversion of two-dimensional analyt-
of the parameterizations used to describe critical models using particle-size data also gives a
ical sedimentation processes, such as particle good constraint on plume height (Volentik et al.,
aggregation, particle terminal velocity, and col- 2010). However, it is important to bear in mind
umn dynamics. that even a small uncertainty in plume height
results in an uncertainty about four times larger
9.5.1 Input parameters: erupted mass, in mass flux Ṁ, because of the fourth power rela-
grain-size distribution, plume tionship between Ṁ and H (Eq. 9.8).
height, wind proile All tephra dispersal models are strongly
Erupted mass is the most important input dependent on the choice of initial grain-size
parameter and is one of the most difficult to distribution. Nonetheless, even though sev-
derive accurately from field data. In fact, vol- eral methods have been proposed, including
ume (and mass) estimation strongly depends on simple data averages, sectorization, Voronoi
tessellation, and analytical models (Carey and 9.5.2 Parameterizations of particle

Sigurdsson, 1982; Bonadonna and Houghton, aggregation and settling velocity
2005; Mannen, 2006; Volentik et al., 2010), none Particle aggregation is a fundamental process
of these techniques is able to reproduce missing that typically occurs in ash-rich volcanic clouds
field data. As a result, any extrapolation based for particle diameters < 100 μm, and results in
on empirical fitting of poor data sets is likely to the formation of coated crystals, dry aggregates,
be problematic. and accretionary lapilli, depending on the water
Tephra dispersal is significantly affected by content (Sparks et al., 1997). Ash-coated crystals
local meteorological conditions. The required (typically particles > 200 μm coated by < 20-μm
resolution of meteorological data depends particles) are considered responsible for scaven-
mainly on the length scale and specific problem ging only small volumes of fine particles and the
being considered. However, wind profile infor- associated ash coating is not expected to change
mation is commonly difficult to include in sim- the particle terminal velocity significantly
ulations of particle sedimentation for two main (James et al., 2002). In contrast, both accretionary
reasons: (1) wind profile data are not always lapilli and dry aggregates can significantly affect
available for a given eruption, certainly not for sedimentation of ash-rich tephra, as shown by
prehistoric events; (2) many models for tephra numerous field observations (Hobbs et al., 1981;
dispersal do not capture the complexity of the Brazier et al., 1982; Sorem, 1982; Hildreth and
local meteorological conditions (e.g., all analyt- Drake, 1992; Scasso et al., 1994) and both ana-
ical models described above). The only meteoro- lytical and numerical investigations (Carey and
logical information for prehistoric eruptions is Sigurdsson, 1982; Cornell et al., 1983; Wiesner
average wind direction, derived from the disper- et al., 1995; Veitch and Woods, 2001; Bonadonna
sal axis, and maximum wind speed at the tropo- et al., 2002a; Textor et al., 2006). In particular: (1)
pause, derived using the method of Carey and aggregation processes can make fine particles
Sparks (1986); however, wind direction typically fall closer to the vent than expected, generating
varies with height, especially in the lower part secondary maxima of accumulation (Carey and
of the atmosphere. For recent eruptions, glo- Sigurdsson, 1982) (e.g., Fig. 9.2(b)); (2) aggrega-
bal meteorological data with a grid resolution tion processes affect tephra-deposit thinning
of 2.5˚ are available from 1948 to the present and the position of thinning breaks in slope on
(Kalnay et al., 1996; www.cdc.noaa.gov/cdc/data. semi-log plots of thickness vs. distance from vent
ncep.reanalysis.html , http://data.ecmwf.int/ (Bonadonna and Phillips, 2003); (3) deposition
data/d/era40_daily/). For recent years, other use- of co-pyroclastic-density-current ash (generated
ful data come from direct soundings for many both from large ignimbrites and block-and-ash
locations worldwide (http://weather.uwyo.edu/ flows) cannot be described unless aggregation
upperair/sounding.html). Unfortunately, the is accounted for (Cornell et al., 1983; Bonadonna
grid resolution of these data sets is too coarse et al., 2002a); (4) dry aggregation processes signifi-
for many applications, so data derived from cantly affect health hazard assessments because a
mesoscale meteorological models (e.g., the considerable number of particles with diameters
Fifth-Generation NCAR/Penn State Mesoscale < 10 μm are present as small aggregates in all
Model (MM5), and the Weather Research and volcanic plumes (James et al., 2003). The complex-
Forecasting (WRF) model) are more practical. ity of aggregation processes explains why they
For example, Byrne et al. (2007) showed that are not accounted for in most dispersal models.
dispersal of tephra from the 1995 eruption of Such a simplification, combined with the empir-
Cerro Negro (Nicaragua) can be described by ical nature of analytical models, is valid when
accounting for small-scale wind variations dur- sedimentation of ash-poor tephra is considered
ing the eruption. For applications where a very (e.g., eruption of Askja 1875 D; Bonadonna and
fine scale is needed (≤ 1 km) diagnostic wind Phillips, 2003). However, even analytical models
(mass consistent) models such as CALMET can fail to accurately reproduce tephra deposits char-
be used (Scire et al., 2000). acterized by ash-rich grain-size distributions.
Several studies have shown that particle- where L is the longest particle dimension, I is
settling velocities strongly depend on particle the longest dimension perpendicular to L, and
shape (Wilson and Huang, 1979), although for S is the dimension perpendicular to both L
simplicity particles are typically assumed to be and I. In addition, Riley et al. (2003) considered
spheres, for which terminal velocities can be the determination of particle sphericity from
determined using simple expressions (Kunii two-dimensional images, which permits analysis
and Levenspiel, 1969; Arastoopour et al., 1982). of small particles. However, these methods are
The settling velocity vj of particles of size dj is all approximations that need to be tested thor-
obtained from the balance between gravity and oughly for application to calculating terminal
air drag. The drag coefficient, Cd, is a function of velocities of volcanic particles. As a result, the
the particle shape and the Reynolds number, Re = effect of particle shape on terminal velocity is
dj ρa vj /ηa, where ηa is the air dynamic viscosity a critical factor that remains to be adequately
(Pa s). The assumption of spherical particles is described. However, the drag coefficient strongly
valid as a first-order approximation only; for depends on particle shape only for relatively
non-spherical particles the determination of Cd large particles (Fig. 9.3). For example, the model
is more complicated. Walker et al. (1971) showed of Ganser (1993) was used here to investigate
that pumice clasts > 5 mm are better described the terminal velocities of irregular particles of
by cylinders than spheres, and Wilson and known shape and diameter ranging between 1.5
Huang (1979) found that, for particle diameters and 7 cm, and with a shape factor (F = (I + S) /
between 30 and 500 μm, glass and feldspar frag- 2L; Wilson and Huang, 1979) ranging between
ments have a very high proportion of flattened 0.3 and 0.9. Figure 9.3 shows the associated drag
particles, whereas pumice clasts have a greater coefficients for Reynolds numbers Re between
variety of shapes, including equant particles. 0.001 and 106. Note that the shape factor F does
Following a review of available methods for not uniquely constrain elongated and platy par-
estimating the drag coefficient of non-spheri- ticles (i.e., F < 0.7), and that rounded particles
cal particles, Chhabra et al. (1999) showed that can have similar drag coefficients to elongated
the best approach appears to be that of Ganser particles. Furthermore, Figure 9.3 shows that
(1993), which uses the equal volume sphere the drag coefficient varies significantly with
diameter and the sphericity ψ of particles, with shape only for particles falling in the intermedi-
a resulting overall error within ~16% for Re ran- ate and turbulent regimes (Re > 1). As a result,
ging from 10−4 to 5 × 105. Unfortunately, expres- future studies of the effects of particle shape on
sions for terminal velocity that account for the terminal velocity should focus on medium- and
complexity of irregular particles are commonly high-Re particles, i.e., particles with diameters >
based on particle parameters that are imprac- 63 μm and > 2 mm respectively, for heights > 10
tical to measure. For example, particle surface km above sea level.
area necessary to calculate ψ cannot be easily An accurate description of plume dynamics
determined because this would imply a com- is crucial for both analytical and numerical mod-
plicated integration over surface elements of els (Scollo et al., 2008b). Model results are very
an irregular particle. For this reason, Wadell sensitive to the choice of velocity profile within
(1933) and Aschenbrenner (1956) introduced the plume, which ultimately controls both the
the concepts of “operational” and “working mass and the grain-size distribution within the
sphericity,” based on the determination of the eruptive column (e.g., Carey and Sparks, 1986;
volume and of the three dimensions of a par- Bursik et al., 1992a). As a first approximation,
ticle respectively: models based on BPT consider a Gaussian dis-
tribution profile across the plume and a mono-
S
2 ⎡ S⎛ I⎞ S ⎛
2
I ⎞⎤
2 tonic trend with height (Carey and Sparks,
ψ work = 12.8 ⎢1 + ⎝ 1 + ⎠ + 6 1 + 2 1 + 2 ⎟ ⎥ , 1986; Bursik et al., 1992a; Sparks et al., 1992;
IL ⎢⎣ I L I ⎝ L ⎠ ⎥⎦
Bonadonna and Phillips, 2003; Bonadonna et al.,
(9.15) 2005a). However, numerical models show more
Figure 9.3 Plot of drag coeficient vs.

Reynolds number for particles with
different shapes. Drag coeficient was
calculated analytically from the model of
Ganser (1993). F denotes shape factor
and ψ is sphericity.
complicated velocity profiles, and simulations leading to tephra dispersal to the southeast. The
from three-dimensional models of weak plumes highest seismic tremor recorded for this plume-
show no Gaussian cross section at any time or forming lava fountain lasted for about 25 minutes
even as a time-averaged property. The advan- (Aloisi et al., 2002; Coltelli et al., 2006). The asso-
tage of simple column models such as those ciated tephra blanket was sampled between ~3
used in the analytical models described above and 30 km from the vent soon after deposition.
or the steady-state models based on BPT (Woods, As a result, a detailed isomass map was compiled
1988; Bursik, 2001; Ishimine, 2006) is their com- (Fig. 9.4), and a maximum column height of
putational speed and flexibility compared to 11 km a.s.l. and a maximum wind speed at the
the complex three-dimensional time-dependent tropopause of 10–30 m s−1 were determined
descriptions of plume dynamics (Dobran et al., using the method of Carey and Sparks (1986)
1993; Herzog et al., 1998; Oberhuber et al., 1998; (Andronico et al., 1999; Table 9.1). The subplinian
Esposti Ongaro et al., 2007). As a result, for some character of this eruption is suggested by com-
regimes and applications, a challenge remain- parison with other tephra deposits on a semi-
ing for the volcanology community is to develop log plot of T vs. A (Fig. 9.5). Plinian eruptions
an accurate physical model for column dynam- ranging from basaltic (e.g., Fontana Lapilli) to
ics that is also computationally fast. rhyolitic (e.g., Taupo) are characterized by larger
maximum thicknesses and more gradual thin-
ning with distance (i.e., larger bt in Eq. (9.2)) than
9.6 Case study the 1998 Etna deposit, which instead plots with
two other well-studied subplinian eruptions:
The 22 July 1998 paroxysmal event of Mt. Etna the 17 June 1996 eruption of Ruapehu volcano
provides a useful illustration of the application (New Zealand) and the 22 July 1980 eruption of
of the main empirical, analytical, and numer- Mt. St. Helens (USA). In addition, sensitivity ana-
ical models described in Sections 9.3–9.5 (Coltelli lyses based on time-series Meteosat images and
et al., 2006; Scollo et al., 2008a). This was one of the theoretical modeling using the PUFF ash tracking
strongest explosive events at Mt. Etna in the last model (Searcy et al., 1998) gave a best-fit value of
century and produced a short-lived strong plume column height of 13 km a.s.l., an eruption dur-
associated with hawaiitic magma (Corsaro and ation between 20 and 40 minutes, and horizon-
Pompilio, 2004) that rose 12 km above sea level tal and vertical diffusivity values of 5000 and 10
(a.s.l.), ~9 km above the vent in Voragine crater, m2 s−1, respectively (Aloisi et al., 2002; Table 9.1).
Table 9.1 Comparison of eruption parameters obtained using different methods. See online supplement 9A for details of the methods used to derive
parameters.
Parameter Method
Power Carey and Numerical
Observed Exponential law Sparks Wilson and Inversion: solution:
data method method (1986) Sparks (1986) Walker (1987 ) TEPHRA2 OAT: PUFF FALL3D
Erupted mass M _ 0.9(1seg) 2.0 _ _ _ 1.7 _ 1.7
(×109 kg) 1.1(2seg)
1.8(3seg)
Column height 12 _ _ 11 _ _ 13 13 12
H (km a.s.l.)
Mass eruption 0.6 (1seg)(8) _ _ 0.6 1.8 _ _ 2.4
rate Ṁ 0.7 (2seg)
(×106 kg s –1) 1.2 (3seg)
Grain-size Md φ (0.8) _ _ _ _ _ Md φ (-0.6) Md φ (0) _

distribution STDV (1.8) STDV (2.2) STDV (1.5)
Duration of 25 (total _ _ _ 24 (1seg) 8 (1seg) _ 20–40 12
sustained duration) 30 (2seg) 10 (2seg)
phase (min) 51 (3seg) 17 (3seg)
56 (PL) 19 (PL)
47 (TEPHRA2) 16 (TEPHRA2)
Maximum/ 11 / 6 _ _ 10–30 _ _ 6/6 _ _
average wind
speed (m s –1)
K H /K V (m2 s –1) _ _ _ _ _ _ 0.5 5000 / 10 5000 /
0.004–600
(mean: 50)
Downloaded from https://www.cambridge.org/core. Durham University Library, on 03 Oct 2018 at 00:39:49, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.
https://doi.org/10.1017/CBO9781139021562.009
density of 2600 kg m−3). For the power-law cal-

culation, A0 and Adist were chosen as 1.1
km (from Eq. (9.5)) and 500 km (based on the
geometry of the deposit). However, the volume
derived by integrating the power-law trend is
not very sensitive to the choice of the outermost
integration limit ( Adist ) because the associated
power-law exponent m > 2, and therefore the
deposit thins very rapidly (Table 9.1; Fig. 9.6(b)).
In contrast, the derived volume is sensitive to
the choice of A0 (Table 9.1), but this can be
fixed using Eq. (9.5). A mass eruption rate of 0.6
× 106 kg s−1 was determined from the model of
Sparks (1986) for an eruption temperature of
1000 °C, as derived by Corsaro and Pompilio
(2004). A mass eruption rate of 1.8 × 106 kg
s−1 was calculated using Eq. (9.8) for a column
height of 9 km and C = 245 m kg−1/4 s1/4 (aver-
aged from Scollo et al. (2007) and Andronico et al.
(2008)). Using the erupted volume determined
from both the exponential and power-law meth-
ods, we obtain an eruption duration between 8
and 19 minutes (method of Wilson and Walker
(1987) with C = 245 m kg−1/4 s1/4), and between 24
and 56 minutes (method of Sparks (1986) with
an eruption temperature of 1000 °C; Table 9.1).
9.6.2 Application of 1D and 2D analytical

Figure 9.4 Isomass map of the 22 July 1998 eruption of
Mt Etna (kg m−2) based on data from Andronico (1999) and
sedimentation models
Coltelli et al. (2006).
As described in Section 9.4.1, the (one-
dimensional) model of Bonadonna and Phillips
(2003) can be used to investigate the fallout
dynamics and the thinning of tephra depos-
9.6.1 Application of empirical models: its (Fig. 9.7). The application of this model to
erupted mass, mass eruption rate, the 22 July 1998 eruption of Mt. Etna explains
and eruption duration the position of the breaks in slope shown in
Because of the variation of deposit density with Figure 9.6 by accounting for the mass fractions
distance from vent, mass/area data were plotted of particles with different Reynolds numbers
against A , and were fitted by three exponential (Run 1, Table 9.2). The first field data point in
segments and by a power-law trend (Fig. 9.6 and Figures 9.6 and 9.7 corresponds to the position
Eqs. (9.3) and (9.7)). For completeness, results for of the plume corner (~2 km from the vent for
one (medial) and two (medial and distal) expo- a plume height of ~9 km). As a result, all of the
nential segments are also shown in Table 9.1. sampled deposit corresponds to fallout from
Both power-law and three-exponential-segment the umbrella cloud, and the two breaks in
fits give a total erupted mass of about 2 × 109 kg, slope correspond to the transitions between:
which corresponds to a total erupted volume of (1) turbulent and intermediate sedimentation
2 × 106 m3 (assuming an average deposit density regimes (break in slope at 95% of intermediate-
of 1050 kg m−3) and to a dense rock equivalent Re particles and 5% of high-Re particles), and
(DRE) volume of 8 × 105 m3 (assuming a magma (2) intermediate and laminar sedimentation
Figure 9.5 Semi-log plot of thickness vs. square root of isopach area for describing the thinning trend of eruptions of different
styles.Vulcanian explosion and dome collapse of Soufrière Hills volcano, Montserrat, West Indies (Bonadonna et al., 2002b).
Subplinian eruptions of Ruapehu volcano, New Zealand (17 June 1996; Bonadonna et al., 2005a); Mount St. Helens, USA (22 July
1980; Sarna-Wojcicki et al., 1981); Kilauea volcano, USA (Keanakakoi tephra, Unit 6; McPhie et al., 1990); Fuego 1974 (Rose et al.,
2008). Plinian eruptions of Askja volcano, Iceland (Unit D; Sparks et al., 1981); Hudson volcano, Chile (12–15 August 1991; Scasso
et al., 1994), Minoan eruption, Greece (Pyle, 1990), Mount St. Helens, USA (18 May 1980; Sarna-Wojcicki et al., 1981); Quizapu,
Chile (1932; Hildreth and Drake, 1992); Tarawera, New Zealand (1886; Walker et al., 1984); Taupo, New Zealand (181AD; Walker,
1980); Masaya volcano, Nicaragua (Fontana Lapilli; Costantini et al., 2008); Masaya volcano, Nicaragua (Triple Layer; Perez et al.,
2009).
regimes (break in slope at 15% of intermedi- The goodness of fit is determined as the root
ate-Re particles and 85% of low-Re particles). mean square error (RMSE):
The discrepancy between the observed and
computed tephra accumulation is probably N
( − )2
due to the under-representation of fine parti-
RMSE = ∑
a =1 Moa
(9.16)
cles in the grain-size distribution derived from
field data, which results from the dominantly where N is the number of observations and Moa
proximal exposure of the deposit. As a result, and Mca are, respectively, the observed and com-
overestimates of terminal velocities will lead puted deposit (i.e., mass per unit area) at sample
to overestimation of tephra accumulation (Eq. location a respectively. In order to illustrate the
(9.12); Fig. 9.7). distribution of minimum values of the good-
The TEPHRA2 model can be used in com- ness-of-fit measure, Figure 9.8 shows the RMSE
bination with dedicated inversion techniques corresponding to 0.2-log(mass) increments and
to determine eruption parameters. Connor and 2-km-height increments (Run 2, Table 9.2). The
Connor (2006) applied the downhill simplex wind direction was constrained based on the
method to find the optimized set of eruption dispersal axis indicated by the isomass map
parameters corresponding to a given tephra (Fig. 9.4). Results show that this technique pro-
deposit, based on comparison between observed vides a very good constraint on the erupted mass
and computed mass accumulation per unit area. but not on the column height. The optimized
1000
100
Mass/area (kg/m2)
y = 309.2e–0.71x
10
y = 27.0e–0.25x
1
y = 1.6e–0.07x
0.1
0.01
0 10 20 30 40 50
Area(1/2) (km)
Figure 9.7 Semi-log plot of mass per unit area (kg m−2) vs.
1000
distance from vent (km) showing the comparison between
Etna 1998 ield data (from Figs. 9.4 and 9.5; symbols) and the
100
Mass/area (kg/m2)
results of the strong plume model of Bonadonna and Phillips

10 (2003) modiied to account for topography (bold curve; Run
1, Table 9.2). Fractions of particles with low, intermediate, and
1 high Reynolds number are also shown (secondary axis; light
curves).Vertical dashed lines indicate the position of the two
0.1 y = 363.4x–2.27 breaks in slope shown in Fig. 9.6a (i.e., BS1 at ~6 km and BS2
at ~23 km, equivalent to BS1 at ~5 km and BS2 at ~16 km in
0.01
0 10 20 30 40 50 units of A ). Inputs are summarized in Table 9.2.
Area(1/2) (km)
Figure 9.6 Semi-log plots of mass per unit area (kg m−2) vs. most particles fell according to power-law diffu-
square root of area (km) for the Etna 1998 deposit, showing sion (Eq. (8) in Bonadonna et al., 2005b). Figure
(a) the best it of three exponential segments and (b) the 9.9 shows the forward solution of TEPHRA2
power-law itting. Best-it equations are also shown. computed using these best-fit values and the
quantitative comparison with field data.
erupted mass varies between 0.4 × 109 and 9.6.3 Application of 3D numerical
2.5 × 109 kg (corresponding to RMSE values of models for particle sedimentation
1.5 ± 0.4 kg m−2 for observed tephra accumula- In order to apply the numerical model
tions between 0.1 and 84.0 kg m−2). The erupted FALL3D (Costa et al., 2006; Folch et al., 2009),
mass is the only eruptive parameter that sig- a pseudo-sounding profile was obtained for
nificantly affects model output, but not through 37.5°N 15°E at 18:00 LT using the WRF model
interaction with other input parameters (Scollo (Skamarock et al., 2005; Table 9.2, Run 5). The
et al., 2008a). However, given that for our case profile was used to initialize the meteorological
study the column height can be well con- processor CALMET (Scire et al., 2000), which
strained from observations, the inversion was produced a finer resolution wind and tem-
run again for plume height from 10–15 km a.s.l. perature field, also incorporating the effects of
and erupted mass of 0.5 − 50 × 109 kg (a large topography.
enough interval for the algorithm to find a sig- The observed column height of 9 km above
nificant minimum). The best fit with field data the vent corresponds to a mass eruption rate Ṁ
was obtained for a plume height of 13 km a.s.l., of 2.5 × 106 kg s−1, derived using the model of
a total erupted mass of 1.7 × 109 kg, and an aver- Bursik (2001) with an exit velocity of 100 m s−1,
age wind speed of 6 m s−1 (Runs 3 and 4, Table a temperature of 1000 °C, and a volatile content
9.2). The low fall-time threshold (FTT) resulting of 0.5 wt.% H2O. In the FALL3D simulations, Ṁ
from the inversion (i.e., 278 s) indicates that was assumed to be constant for the eruption
Table 9.2 Parameters used in the simulations. See online supplement 9B for details of the methods used to
derive parameters.
Run 1 Run 2 Run 3 Run 4 Run 5

Model Strong Plume TEPHRA2 TEPHRA2 TEPHRA2 FALL3D (forward)
(inversion) (inversion) (forward)
Grain-size GS Etna 1998 Md φ (-3–3) Md φ (-3–3) Md φ (-0.6) GS Etna 1998
distribution STDV (1–3) STDV (1–3) STDV (2.2)
Plume height H 12 4–20 10–15 13 Calculated within
(km a.s.l.) model
Wind speed w 3 1–20 1–20 6 Wind Etna 1998
(m s −1) (mean below base (mean below
current) plume height)
Wind direction – 90–200 90–200 110–184 Wind Etna 1998
(˚ from N)
Erupted mass M 2 0.0001–100 0.5–50 1.7 Calculated within
(×109 kg) model (1.7)
Mass eruption rate – – – – 2.4
(×106 kg s −1)
Particle density 900–2600 900–2600 900–2600 900–2600 900–2600
(kg m−3)
K H (m2 s −1) – 0.1–8000 0.1–8000 0.5 5000

Fall-time threshold – 1–5000 1–5000 278 –
FTT (s)
duration and the source terms (i.e., injected of Etna volcano). The turbulent diffusivity ten-
mass per second, column height, mass distribu- sor was described through similarity theory
tion in the plume) were determined through the for the vertical component, and as both con-
model of Bursik (2001), based on BPT. In agree- stant and using a large eddy simulation (LES)
ment with this model, a mass eruption rate Ṁ model for the horizontal component (Folch
of 2.5 × 106 kg s−1 produces a column height of et al., 2009). Note that, when the wind field is
~9 km above the vent (~12 km a.s.l.). Hence, derived from a single sounding, LES models
taking this value for Ṁ and the duration of the can underestimate turbulent diffusion because
climactic phase of 12 minutes (from best-fit ana- there is no horizontal shear. In fact, fits to the
lysis), the total mass erupted is found to be 1.7 × data suggest a constant horizontal diffusion
109 kg. Figure 9.10 shows the deposit obtained of KH ≈ 5000 m2 s−1, compared to the average
by FALL3D using a grid of 51 × 51 × 18 km (1 km value of KH ≈ 2250 m2 s−1 predicted by the LES
horizontal spacing) and the wind field refined model. A comparison of the simulation results
by CALMET. and observed deposit is shown in Fig. 9.10(b).
Terminal velocity was calculated using the Finally, FALL3D can also be used to assess ash
model of Ganser (1993) with a sphericity of concentration in the atmosphere and identify
0.93–0.95 (consistent with the data of Coltelli hazardous zones for air traffic (Folch et al.,
et al. (2008) for the 18 December 2002 eruption 2009).
Figure 9.8 Plot of log(erupted mass)

vs. plume height (km) showing the
minimum values of the goodness-of-it
measure (RMSE; Eq. (9.16)) for the Etna
1998 tephra deposit. Erupted mass was
varied between 104 and 1011 kg with
0.2-log(mass) increments and the plume
height was varied between 4 and 20 km
with 2-km-height increments (Run 2,Table
9.2). Resulting values were interpolated
to produce 2D RMSE surfaces and the
plot was cropped to show only the
interpolated surfaces. Dark blue on the
RMSE scale indicates the minimum values
(i.e., best it). RMSE values vary between
1 and 100 kg m−2 and are contoured with
0.8 kg m−2 contours (RMSE values > 20
kg m−2 were observed only for erupted
masses > 3 × 1010 kg and were set equal
to 20 kg m−2 for a better visualization of
the minimum). See color plates section.
9.7 Discussion coefficients (Table 9.1). In particular, the stand-

ard deviation resulting from plume height cal-
Integrated application of empirical, analytical, culations using four different models is 0.8 km,
and numerical models to tephra deposits pro- with discrepancies from the observed values
vides insights into the dynamics of the associated between 0 and 8%. The standard deviation
volcanic eruptions and allows for compilation for erupted mass calculated using six differ-
of comprehensive hazard assessments. ent methods is 4 × 108 kg. Discrepancies with
respect to the mass obtained with the inver-
9.7.1 Case study sion of TEPHRA2, and the numerical solution
The comprehensive data set of the 22 July FALL3D (i.e., 1.7 × 109 kg) are 19% integrating
1998 Etna eruption provided the opportunity the power-law fit, 49% integrating one expo-
to investigate the uncertainties in eruption nential segment, 35% integrating two expo-
parameters derived from sedimentation mod- nential segments, and 9% integrating three
els with different levels of complexity, and exponential segments.
to apply these models to basaltic explosive Third, mass eruption rate is more difficult
eruptions. First, we stress that, even though to constrain; we found a discrepancy of ~67%
magma composition does not seem to affect between the values obtained using the empiri-
sedimentation dynamics (e.g., Fig. 9.5), it must cal model of Wilson and Walker (1987) and the
be accounted for when deriving eruption model of Sparks (1986) for the same plume tem-
parameters. For example, we have shown that perature (1000 °C). In addition, the mass erup-
inferred mass eruption rate strongly depends tion rate derived numerically using FALL3D is
on eruption temperature (Sparks, 1986; Wilson ~300% and 33% larger than the results from
and Walker, 1987). Sparks (1986) and Wilson and Walker (1987),
Second, to assess the variation of eruption respectively (Table 9.1). Ṁ derived by dividing
parameters derived using different modeling the calculated erupted mass by the observed
approaches, we have compared results for duration (i.e., 25 minutes) is 26–67% lower than
calculations of plume height, erupted mass, the Ṁ of Wilson and Walker (1987) and up to
mass eruption rate, wind profile and diffusion 122% larger than that of Sparks (1986) (Table 9.1).
from the six different erupted mass values in

Table 9.1 and the Ṁ of Sparks (1986) is always
~200% greater than when derived from the Ṁ of
Wilson and Walker (1987). The duration derived
using the Ṁ of Wilson and Walker (1987) and
the mass inverted from TEPHRA2 is about 23%
greater and 28% less than the minimum duration
derived from the numerical solution of FALL3D
and PUFF (Table 9.1). Discrepancies resulting
from the calculation of wind velocity are within
50% of the observed value for all methods.
Finally, there is a large discrepancy observed
for the diffusion coefficient calculated using
FALL3D and PUFF, and the horizontal diffusion
coefficient obtained using TEPHRA2. The dif-
fusion coefficient used in analytical models is
an empirical parameter that depends on vari-
ous physical processes. In addition, TEPHRA2
describes particle diffusion using two diffusion
laws for coarse and fine particles; a constant
diffusion coefficient is used only for coarse
particles (Fickian diffusion). As a result, the dif-
fusion coefficient does not have the same mean-
ing as for numerical models such as FALL3D
and PUFF. However, the horizontal diffusion
coefficients used in FALL3D and PUFF give the
same result, which is about double the value
derived from the LES model. The mean vertical
diffusion coefficient obtained from similarity
Figure 9.9 Forward solution of the TEPHRA2 model (Run
theory is also on the same order of magnitude
4 in Table 9.2) run with the best-it values obtained from the as the coefficient derived from PUFF. In gen-
inversion computation (Run 3, Table 9.2): (a) isomass map eral, it is important to thoroughly understand
with contours of 0.1, 0.3, 0.4, 0.5, 1, 1.5, 3, 10, 25, and parameters from different models before com-
80 kg m−2, shown for comparison with Fig. 9.4. Locations of paring them.
the volcano and the ield data are indicated. (b) Comparison We conclude that for the 1998 Etna eruption,
between computed (solid line) and observed (symbols) mass/ column height is the parameter that can be con-
area data. Dashed lines indicate over- or under-estimations of
strained with the least uncertainty. The total
1/5 and 5 times the observed values, respectively. TEPHRA2
erupted mass can also be constrained within an
was run using a Global Land One-km Base Elevation map
(GLOBE; www.ngdc.noaa.gov/mgg/topo/globe.html).
acceptable range using empirical (exponential
method applied to three segments and power-
law fit), analytical (inversion of TEPHRA2), and
Discrepancies in both mass eruption rate and numerical (FALL3D) models. In contrast, the
erupted mass result in a discrepancy in the erup- derivations of Ṁ, duration, wind profile, and dif-
tion duration ranging between 9 and 49% with fusion coefficients are more problematic. This is
respect to the duration obtained from the total probably due to the greater complexity associ-
mass derived using the inversion of TEPHRA2 ated with these parameters, given that they are
and FALL3D (i.e., 16 and 47 minutes using the Ṁ strongly connected to atmospheric characteris-
of Wilson and Walker (1987) and Sparks (1986), tics, and are also parameterized differently in
respectively). In contrast, the duration derived different sedimentation models. In particular,
9.7.2 Determination of eruption

parameters
As described above, eruption parameters can
be inferred by applying empirical, analytical,
and numerical models, and through inversion
solutions of analytical models. Our case study
confirms that empirical extrapolation of poor
data sets can be misleading (e.g., volume cal-
culations based on the integration of only one
or two exponential segments), but that integra-
tions of at least three exponential segments and
of the power-law fit give comparable results for
moderate eruptions. However, this result is not
necessarily expected to hold for more exten-
sively dispersed deposits. In addition, sensitiv-
ity tests carried out for the integration of the
power-law fit for the Etna deposit show larger
discrepancies caused by the choice of A0 than
of Adist (Table 9.1). This is consistent with the
findings of Bonadonna and Houghton (2005)
for deposits produced by relatively small erup-
tions characterized by a power-law exponent m
> 2 (e.g., Ruapehu 1996). However, for widely
dispersed deposits (i.e, m < 2) the power-law
method is more sensitive to the choice of Adist
than A0 . As a result, poorly exposed deposits,
in particular those of large eruptions, would be
better described by applying inversion solutions
of analytical models such as TEPHRA2 (Connor
Figure 9.10 FALL3D simulation results for the Etna and Connor, 2006), ASHFALL (Hurst and Turner,
22 July 1998 deposit (Run 5, Table 9.2): (a) Isomass map 1999), and HAZMAP (Macedonio et al., 2005).
with contours of 0.1, 0.3, 0.4, 0.5, 1, 1.5, 3, 10, 25, and Scollo et al. (2008a) have shown that the model
80 kg m−2, shown for comparison with Figs. 9.4 and 9.9. TEPHRA can be used to determine erupted
Locations of the volcano and the ield data are indicated. (b) mass because it is very sensitive to this param-
Comparison between computed (solid line) and observed eter, independent of other inputs, and that the
(symbols) mass/area data. Dashed lines indicate over- or
erupted mass can be well constrained with at
under-estimations of 1/5 and 5 times the observed values
least 10 well-distributed field data points (the
respectively.
data set of our case study has 33 thickness meas-
urements). These conclusions hold for the other
discrepancies in the calculation of mass erup- analytical models. In contrast, determination of
tion rate are due to a combination of factors, column height using TEPHRA is less straight-
including the more complex plume dynamics forward because height is an input, along with
resulting from lava-fountaining fragmentation, the total grain-size distribution, that affects the
and the effects of wind shear on air entrain- model output through interaction with other
ment within the convective region described input parameters (Scollo et al., 2008a). These
in FALL3D, which, for a given plume height, observations are confirmed by application of
requires a larger Ṁ compared to standard ana- inversion techniques using TEPHRA2, which
lytical models. shows that erupted mass is easier to constrain
than column height (e.g., Fig. 9.8). However, probabilistic and/or real-time forecasting of
Volentik et al. (2010) have shown that a better tephra dispersal. Typically, analytical models
constraint on column height is obtained when are computationally fast and therefore can
tephra accumulation is inverted for individual be used to compile fully probabilistic assess-
grain-sizes. Further studies of the application ments, whereas numerical models are better
of inversion techniques are needed to assess suited to real-time forecasting to provide accur-
the minimum amount of field data required to ate estimates of ground sedimentation and of
provide a reliable characterization. In addition, the position of the volcanic cloud with time.
inversion techniques are not always straightfor- In particular, our case study confirms that the
ward to apply; the choice of parameter ranges more sophisticated numerical models do not
can significantly affect the final result and there- necessarily provide better accuracy in terms
fore requires critical analysis. We also stress that of ground sedimentation, especially given the
studies of poorly exposed deposits have shown uncertainties associated with the input param-
that the method of Carey and Sparks (1986) gives eters (e.g., Scollo et al., 2008b; Figs. 9.9 and
good results even when the position of the vent is 9.10). However, numerical models can provide
not well known (e.g., Wehrmann et al., 2006). As crucial information not possible with analytical
a result, inversion techniques should always be models (e.g., a description of cloud movement
used in combination with other models for con- with time). In general, verified and validated
straining both erupted mass and plume height. numerical models are also more appropriate
In particular, granulometry data in proximal, for real-time forecasting because they require
medial, and distal areas should always be col- fewer empirical parameters (e.g., diffusion
lected to better constrain plume height. Finally, coefficient, mass distribution within the erup-
models based on empirical observations should tive column) and, as a result, are simpler to
always be applied in their range of validation. apply. Models of all levels of sophistication
would benefit significantly from better param-
eterization of critical sedimentation processes
9.8 Summary and outlook such as particle aggregation and settling, and
from quantification of uncertainties associated
Comprehensive characterization of tephra with input parameters such as erupted mass,
deposits and reliable hazard assessment can plume height, mass discharge rate, and total
only result from critical and synergistic appli- grain-size distribution.
cation of models with levels of sophistication
ranging from purely empirical to fully numer-
ical. First, tephra deposits need to be sampled 9.9 Notation
accurately (for both mass/area and grain-size
data) over an area proportional to the associated a sample location in application of
particle dispersal. Second, dedicated empirical inversion techniques
and analytical models can be used for deter- A area enclosed within isomass/isopach
mination of plume height, erupted mass, ini- contours (m2)
tial grain-size distribution, mass eruption rate Ax area enclosed within isopach line of
and duration. Inversion solutions of analytical thickness Tx (m2)
models can also be used to obtain independ- A0 integration limit of power-law
ent results for the same eruption parameters. function, typically taken as distance
Eruption parameters can be used to classify of maximum deposit thickness (m)
volcanic eruptions and build potential eruptive Adist integration limit of power-law
scenarios. Finally, following thorough model function, typically taken as the
validation and calibration, analytical and downwind extent of deposit (m)
numerical models can be applied to compile bt thickness half distance (m)
cj(x,y,z,t) particle concentration (kg m−3) S(x,y,z,t) source term (kg m−3 s−1)
c′j turbulent fluctuation of concentration t time coordinate (s)
of particle class j (kg m−3) T thickness of tephra deposit (m)
c̅j ensemble average of concentration Tn0 intercept of exponential segment
of particle class j (kg m−3) n (m)
C empirical factor (m kg−1/4 s1/4) T0 maximum thickness of tephra
CD drag coefficient deposit (m)
Cpl coefficient of power-law best fit Tx thickness of given isopach x (m)
(m(1 + m)) ux,uy,uz components of wind velocity vector
dj diameter of particle of class j (m) (m s−1)
F particle shape factor u̅x,u̅y,u̅z components of average wind velocity
H maximum height of volcanic plume (m s−1)
(m) u′x,u′y,u′z components of turbulent fluctuation
Hb neutral buoyancy level of volcanic of wind velocity (m s−1)
plume (m) U current velocity in s-direction of
j index of particle size plume or umbrella cloud (m s−1)
k slope of deposit exponential best-fit vHb particle terminal velocity at neutral
curve (km−1) buoyancy level Hb (m s−1)
kn slope of exponential segment vj particle terminal velocity of particle
n (km−1) class j; vj = vj(x,y,z) (m s−1)
KH horizontal atmospheric diffusion V erupted volume (m3)
coefficient (KH = Kx = Ky) (m2 s−1) w maximum crosswind width at source
Kx x-component of horizontal diffusion of spreading current (m)
coefficient (m2 s−1) xo plume corner position (m)
KV vertical atmospheric diffusion x, y, z spatial coordinates (m)
coefficient (m2 s−1) ηa dynamic viscosity of air (Pa s)
Ky y-component of horizontal diffusion ρa density of air (kg m−3)
coefficient (m2 s−1) ψ sphericity
m exponent of power-law best fit curve ψwork “working” sphericity
L, I, S dimensions of three perpendicular Re Reynolds number
axes of particle (m)
M total mass of particles of given
size fraction carried by current at List of acronyms
distance x1 (kg)
Ṁ mass eruption rate (kg s−1) ABL Atmospheric Boundary Layer
M0 initial mass for given grain size ADS Advection Diffusion Sedimentation
injected into current (kg) ATHAM Active Tracer High Resolution
Moa observed mass per unit area at Atmospheric Model
sample location a (kg m−2) BPT Buoyant Plume Theory
Mca computed mass per unit area at CANERM CANadian Emergency Response Model
DRE Dense Rock Equivalent
sample location a (kg m−2)
FTT Fall Time Threshold
n number of exponential segments
GLOBE Global Land One-km Base Elevation
N number of field observations in
map
application of inversion techniques HYSPLIT Hybrid Single Particle Lagrangian
Q volumetric flow rate into current at Integrated Trajectory Model
neutral buoyancy level (m3 s−1) IR thermal infrared
s spreading direction of current (m) LES Large Eddy Simulation
Sn position of break in slope of MEDIA Eulerian Model for DIspersion in the
exponential segment n (m) Atmosphere
MM5 Fifth-Generation NCAR/Penn State Aschenbrenner, B. C. (1956). A new method of

Mesoscale Model expressing particle sphericity. Journal of Sedimentary
NAME Numerical Atmospheric-dispersion Petrology, 26, 15–31.
Modelling Environment Barberi, F., Coltelli, M., Frullani, A., Rosi, M. and
OAT One At a Time sensitivity tests Almeida, E. (1995). Chronology and dispersal
SGS Sub-Grid Scale characteristics of recently (last 5000 years) erupted
VAAC Volcanic Ash Advisory Centers tephra of Cotopaxi (Ecuador): implications for
VAFTAD Volcanic Ash Forecast Transport And long-term eruptive forecasting. Journal of Volcanology
Dispersion and Geothermal Research, 69, 217–239.
WRF Weather Research and Forecasting Barsotti, S. and Neri, A. (2008). The VOL-CALPUFF
model model for atmospheric ash dispersal: 2.
Application to the weak Mount Etna plume of July
2001. Journal of Geophysical Research, 113, B03209,
Acknowledgments doi:10.1029/2006JB004624.
Barsotti, S., Neri, A. and Scire, J. S. (2008). The
A. Costa was supported by the MIUR-FIRB VOL-CALPUFF model for atmospheric ash
Italian project “Sviluppo Nuove Tecnologie dispersal: 1. Approach and physical formulation.
per la Protezione e Difesa del Territorio dai Journal of Geophysical Research, 113, B03208,
Rischi Naturali.” The authors are grateful to doi:10.1029/2006JB004623.
D. Andronico, M. Coltelli, and P. Del Carlo for Bonadonna, C. and Phillips, J. C. (2003).
providing the field data of the 22 July 1998 erup- Sedimentation from strong volcanic plumes.
tion of Mt. Etna and to A. Folch for providing the Journal of Geophysical Research, 108(B7), 2340–2368.
pseudo-sounding used to run CALMET. S. Scollo Bonadonna, C. and Houghton, B. F. (2005). Total
and L. Connor are thanked for useful discussion. grainsize distribution and volume of tephra-fall
deposits. Bulletin of Volcanology, 67, 441–456.
Bonadonna, C., Ernst, G. G. J. and Sparks, R. S. J.
References (1998). Thickness variations and volume estimates
of tephra fall deposits: the importance of particle
Aloisi, M., D’Agostino, M., Dean, K. G., Mostaccio, Reynolds number. Journal of Volcanology and
A. and Neri, G. (2002). Satellite analysis and Geothermal Research, 81, 173–187.
PUFF simulation of the eruptive cloud generated Bonadonna, C., Macedonio, G. and Sparks, R. S. J.
by the Mount Etna paroxysm of 22 July 1998. (2002a). Numerical modelling of tephra fallout
Journal of Geophysical Research, 107(B12), 2373, associated with dome collapses and Vulcanian
doi:10.1029/2001JB000630. explosions: application to hazard assessment on
Andronico, D., Del Carlo, P. and Coltelli, M. (1999). Montserrat. In The Eruption of Soufrière Hills Volcano,
The 22 July 1998 fire fountain episode at Voragine Montserrat, from 1995 to 1999, ed. T. H. Druitt and
Crater (Mt. Etna, Italy). Volcanic and Magmatic Studies B. P. Kokelaar. Geological Society London Memoir,
Group Annual Meeting, UK, 5–6 January. 21, 517–537.
Andronico, D., Scollo, S., Caruso, S. and Cristaldi, Bonadonna, C., Mayberry, G. C., Calder, E. et al.
A. (2008). The 2002–03 Etna explosive activity: (2002b). Tephra fallout in the eruption of Soufrière
Tephra dispersal and features of the deposits. Hills Volcano, Montserrat. In The Eruption of
Journal of Geophysical Research, 113, B04209, Soufrière Hills Volcano, Montserrat, from 1995 to 1999,
doi:10.1029/2007JB005126. ed. T. H. Druitt and B. P. Kokelaar. Geological
Arastoopour, H., Wang, C. H. and Weil, S. A. (1982). Society London Memoir, 21, 483–516.
Particle-particle interaction force in a dilute Bonadonna, C., Phillips, J. C. and Houghton, B. F.
gas-solid system. Chemical Engineering Sciences, 37, (2005a). Modeling tephra sedimentation
1379–1386. from a Ruapehu weak plume eruption.
Armienti, P., Macedonio, G. and Pareschi, M. T. Journal of Geophysical Research, 110, B08209,
(1988). A numerical model for simulation of tephra doi:10.1029/2004JB003515.
transport and deposition – applications to May Bonadonna, C., Connor, C. B., Houghton, B. F.
18, 1980, Mount St. Helens eruption. Journal of et al. (2005b). Probabilistic modeling of
Geophysical Research, 93(B6), 6463–6476. tephra dispersion: hazard assessment of a
multi-phase eruption at Tarawera, New Zealand. Coltelli, M., Puglisi, G., Guglielmino, F. and
Journal of Geophysical Research, 110, B03203, Palano, M. (2006). Application of differential SAR
doi:10.1029/2003JB002896. interferometry for studying eruptive event of 22
Brazier, S., Davis, A. N., Sigurdsson, H. and Sparks, July 1998 at Mt. Etna. Quaderni di Geofisica, 43,
R. S. J. (1982). Fallout and deposition of volcanic 15–20.
ash during the 1979 explosive eruption of the Coltelli, M., Miraglia, L. and Scollo, S. (2008).
Soufrière of St. Vincent. Journal of Volcanology and Characterization of shape and terminal velocity of
Geothermal Research, 14, 335–359. tephra particles erupted during the 2002 eruption
Bursik, M. (2001). Effect of wind on the rise height of of Etna volcano, Italy. Bulletin of Volcanology, 70,
volcanic plumes. Geophysical Research Letters, 28(18), 1103–1112.
3621–3624. Connor, L. G. and Connor, C. B. (2006). Inversion is
Bursik, M. I., Sparks, R. S. J., Gilbert, J. S. and the key to dispersion: understanding eruption
Carey, S. N. (1992a). Sedimentation of tephra by dynamics by inverting tephra fallout. In Statistics in
volcanic plumes: I. Theory and its comparison Volcanology, ed. H. Mader, S. Cole, C. B. Connor and
with a study of the Fogo A plinian deposit, L. G. Connor. Special Publications of IAVCEI, 1, pp.
Sao Miguel (Azores). Bulletin of Volcanology, 54, 231–242. London: Geological Society.
329–344. Connor, C. B., Hill, B. E., Winfrey, B., Franklin, N. M.
Bursik, M. I., Carey, S. N. and Sparks, R. S. J. (1992b). and La Femina, P. C. (2001). Estimation of volcanic
A gravity current model for the May 18, 1980 hazards from tephra fallout. Natural Hazards Review,
Mount St. Helens plume. Geophysical Research Letters, 2, 33–42.
19, 1663–1666. Cornell, W., Carey, S. and Sigurdsson, H. (1983).
Byrne, M. A., Laing, A. G. and Connor, C. (2007). Computer simulation of transport and deposition
Predicting tephra dispersion with a mesoscale of the Campanian Y-5 Ash. Journal of Volcanology and
atmospheric model and a particle fall model: Geothermal Research, 17, 89–109.
Application to Cerro Negro volcano. Journal of Corsaro, R. A. and Pompilio, M. (2004). Magma
Applied Meteorology and Climatology, 46, dynamics in the shallow plumbing system of
121–135. Mt. Etna as recorded by compositional variations
Carazzo, G., Kaminski, E. and Tait, S. (2008). On the in volcanics of recent summit activity (1995–1999).
dynamics of volcanic columns: A comparison of Journal of Volcanology and Geothermal Research, 137,
field data with a new model of negatively buoyant 55–71.
jets. Journal of Volcanology and Geothermal Research, Costa, A., Macedonio, G. and Folch, A. (2006). A
178, 94–103. three-dimensional Eulerian model for transport
Carey, S. N. and Sigurdsson, H. (1982). Influence and deposition of volcanic ashes. Earth and
of particle aggregation on deposition of distal Planetary Science Letters, 241, 634–647.
tephra from the May 18, 1980, eruption of Mount Costantini, L., Bonadonna, C., Houghton, B. F.
St. Helens volcano. Journal of Geophysical Research, and Wehrmann, H. (2008). New physical
87(B8), 7061–7072. characterization of the Fontana Lapilli basaltic
Carey, S. N. and Sigurdsson, H. (1986). The 1982 Plinian eruption, Nicaragua. Bulletin of Volcanology,
eruptions of El Chichon volcano, Mexico (2): 71, 337–355.
observations and numerical modelling of D’amours, R. (1998). Modeling the ETEX plume
tephra-fall distribution. Bulletin of Volcanology, 48, dispersion with the Canadian emergency response
127–141. model. Atmospheric Environment, 32, 4335–4341.
Carey, S. and Sigurdsson, H. (1989). The intensity Dobran, F., Neri, A. and Macedonio, G. (1993).
of Plinian eruptions. Bulletin of Volcanology, 51, Numerical simulation of collapsing volcanic
28–40. columns. Journal of Geophysical Research, 98,
Carey, S. N. and Sparks, R. S. J. (1986). Quantitative 4231–4259.
models of the fallout and dispersal of tephra from Draxler, R. R. and Hess, G. D. (1998). An overview of
volcanic eruption columns. Bulletin of Volcanology, the HYSPLIT_4 modelling system for trajectories,
48, 109–125. dispersion and deposition. Australian Meteorological
Chhabra, R. P., Agarwal, L. and Sinha, N. K. (1999). Magazine, 47, 295–308.
Drag on non-spherical particles: an evaluation of Durant, A. J., Rose, W. I., Sarna-Wojcicki, A. M.,
available methods. Powder Technology, 101, 288–295. Carey, S. and Volentik, A. C. M. (2009).
Hydrometeor-enhanced tephra sedimentation: Holasek, R. E. and Self, S. (1995). GOES weather-

Constraints from the 18 May 1980 eruption of satellite observations and measurements of the
Mount St. Helens. Journal of Geophysical Research, May 18, 1980, Mount St. Helens eruption. Journal of
114, B03204, doi:10.1029/2008JB005756. Geophysical Research, 100(B5), 8469–8487.
Esposti Ongaro, T., Cavazzoni, C., Erbacci, G., Neri, A. Hurst, A. W. and Turner, R. (1999). Performance
and Salvetti, M. V. (2007). A parallel multiphase of the program ASHFALL for forecasting ashfall
flow code for the 3D simulation of explosive during the 1995 and 1996 eruptions of Ruapehu
volcanic eruptions. Parallel Computing, 33, 541–560. volcano. New Zealand Journal of Geology and
Fierstein, J. and Nathenson, M. (1992). Another Geophysics, 42, 615–622.
look at the calculation of fallout tephra volumes. Ishimine, Y. (2006). Sensitivity of the dynamics of
Bulletin of Volcanology, 54, 156–167. volcanic eruption columns to their shape. Bulletin
Folch, A. and Felpeto, A. (2005). A coupled model of Volcanology, 68(6), 516–537.
for dispersal of tephra during sustained explosive James, M. R., Gilbert, J. S. and Lane, S. J. (2002).
eruptions. Journal of Volcanology and Geothermal Experimental investigation of volcanic particle
Research, 145, 337–349. aggregation in the absence of a liquid phase.
Folch, A., Jorba, O. and Viramonte, J. (2008). Volcanic Journal of Geophysical Research, 107(B9), 2191,
ash forecast – application to the May 2008 Chaiten doi:10.1029/2001JB000950.
eruption. Natural Hazards and Earth System Sciences, James, M. R., Lane, S. J. and Gilbert, J. S. (2003).
8, 927–940. Density, construction, and drag coefficient
Folch, A., Costa, A. and Macedonio, G. (2009). of electrostatic volcanic ash aggregates.
FALL3D: A computational model for transport Journal of Geophysical Research, 108(B9), 2435,
and deposition of volcanic ash. Computers and doi:10.1029/2002JB002011.
Geosciences, 35, 1334–1342 Kalnay, E., Kanamitsu, M., Kistler, R. et al. (1996). The
Froggatt, P. C. (1982). Review of methods estimating NCEP/NCAR 40-year reanalysis project. Bulletin of
rhyolitic tephra volumes; applications to the Taupo the American Meteorological Society, 77, 437–471.
Volcanic Zone, New Zealand. Journal of Volcanology Kaminski, E. and Jaupart, C. (1998). The size
and Geothermal Research, 14, 1–56. distribution of pyroclasts and the fragmentation
Ganser, G. H. (1993). A rational approach to drag sequence in explosive volcanic eruptions.
prediction of spherical and nonspherical particles. Journal of Geophysical Research, 103(B12),
Powder Technology, 77, 143–152. 29 759–29 779.
Glaze, L. S. and Self, S. (1991). Ashfall dispersal Koyaguchi, T. (1994). Grain-size variation of tephra
for the 16 September 1986, eruption of Lascar, derived from volcanic umbrella clouds. Bulletin of
Chile, calculated by a turbulent-diffusion model. Volcanology, 56, 1–9.
Geophysical Research Letters, 18, 1237–1240. Koyaguchi, T. and Ohno, M. (2001a). Reconstruction
Glaze, L. S., Wilson, L. and Mouginis-Mark, P. J. of eruption column dynamics on the basis of grain
(1999). Volcanic eruption plume top topography size of tephra fall deposits. 1. Methods. Journal of
and heights as determined from photoclinometric Geophysical Research, 106(B4), 6499–6512.
analysis of satellite data. Journal of Geophysical Koyaguchi, T. and Ohno, M. (2001b). Reconstruction
Research, 104(B2), 2989–3001. of eruption column dynamics on the basis of
Heffter, J. L. and Stunder, B. J. B. (1993). Volcanic grain size of tephra fall deposits. 2. Application to
Ash Forecast Transport and Dispersion (VAFTAD) the Pinatubo 1991 euption. Journal of Geophysical
Model. Weather and Forecasting, 8, 533–541. Research, 106(B4), 6513–6533.
Herzog, M., Graf, H. F., Textor, C. and Oberhuber, Kunii, D. and Levenspiel, O. (1969). Fluidization
J. M. (1998). The effect of phase changes of water Engineering. New York: Wiley and Sons.
on the development of volcanic plumes. Journal of Legros, F. (2000). Minimum volume of a tephra
Volcanology and Geothermal Research, 87, 55–74. fallout deposit estimated from a single isopach.
Hildreth, W. and Drake, R. E. (1992). Volcano Journal of Volcanology and Geothermal Resarch, 96,
Quizapu, Chilean Andes. Bulletin of Volcanology, 54, 25–32.
93–125. Macedonio, G., Costa, A. and Longo, A. (2005). A
Hobbs, P. V., Lyons, J. H., Locatelli, J. D. et al. (1981). computer model for volcanic ash fallout and
Radar detection of cloud-seeding effects. Science, assessment of subsequent hazard. Computers and
213, 1250–1252. Geosciences, 31, 837–845.
Mannen, K. (2006). Total grain size distribution of Pyle, D. M. (1989). The thickness, volume and
a mafic subplinian tephra, TB-2, from the 1986 grainsize of tephra fall deposits. Bulletin of
Izu-Oshima eruption, Japan: An estimation based Volcanology, 51, 1–15.
on a theoretical model of tephra dispersal. Journal Pyle, D. M. (1990). New estimates for the volume of
of Volcanology and Geothermal Research, 155, 1–17. the Minoan eruption. In Thera and the Aegean World,
Martin, D. and Nokes, R. (1988). Crystal settling in ed. D. A. Hardy. London: The Thera Foundation,
a vigorously convecting magma chamber. Nature, 113–121.
332, 534–536. Pyle, D. M. (1995). Assessment of the minimum
McPhie, J., Walker, G. P. L. and Christiansen, R. L. volume of tephra fall deposits. Journal of Volcanology
(1990). Phreatomagmatic and phreatic fall and and Geothermal Research, 69, 379–382.
surge deposits from explosions at Kilauea volcano, Riley, C. M., Rose, W. I. and Bluth, G. J. S. (2003).
Hawaii, 1790 A.D.: Keanakakoi Ash Member. Quantitative shape measurements of distal
Bulletin of Volcanology, 52, 334–354. volcanic ash. Journal of Geophysical Research, 108,
Michalakes, J., Dudhia, J., Gill, D. et al. (2005). The 2504, doi:10.1029/2001JB000818.
weather research and forecast model: Software Rose, W. I. (1993). Comment on “Another look at the
architecture and performance. Use of High calculation of fallout tephra volumes”. Bulletin of
Performance Computing in Meteorology, Proceedings of Volcanology, 55, 372–374.
the Eleventh ECMWF Workshop, 156–168. Rose, W. I., Self, S., Murrow, P. J. et al. (2008). Nature
Morton, B., Taylor, G. L. and Turner, J. S. (1956). and significance of small volume fall deposits at
Turbulent gravitational convection from composite volcanoes: Insights from the October
maintained and instantaneous source. Proceedings 14, 1974 Fuego eruption, Guatemala. Bulletin of
of the Royal Society of London, A234, 1–23. Volcanology, 70, 1043–1067.
Neri, A., Papale, P. and Macedonio, G. (1998). The Rosi, M. (1998). Plinian eruption columns: particle
role of magma composition and water content transport and fallout. In From Magma to Tephra:
in explosive eruptions: 2. Pyroclastic dispersion Modelling Physical Processes of Explosive Volcanic
dynamics. Journal of Volcanology and Geothermal Eruptions, ed. A. Freundt and M. Rosi. Elsevier, pp.
Research, 87, 95–115. 139–172.
Oberhuber, J. M., Herzog, M., Graf, H. F. and Ryall, D. B. and Maryon, R. H. (1998). Validation
Schwanke, K. (1998). Volcanic plume simulation of the UK Met. Office’s NAME model against
on large scales. Journal of Volcanology and Geothermal the ETEX dataset. Atmospheric Environment, 32,
Research, 87, 29–53. 4265–4276.
Papale, P. and Rosi, M. (1993). A case of no-wind Sandu, I., Bompay, F. and Stefan, S. (2003). Validation
plinian fallout at Pululagua caldera (Ecuador): of atmospheric dispersion models using ETEX data.
implications for model of clast dispersal. Bulletin of International Journal of Environment and Pollution, 19,
Volcanology, 55, 523–535. 367–389.
Perez, W., Freundt, A., Kutterolf, S. and Schmincke, Sarna-Wojcicki, A. M., Shipley, S., Waitt, J. R.,
H.-U. (2009). The Masaya Triple Layer: A 2100 year Dzurisin, D. and Wood, S. H. (1981). Areal
old basaltic multi-episodic Plinian eruption from distribution thickness, mass, volume, and
the Masaya Caldera Complex (Nicaragua). Journal of grain-size of airfall ash from the six major
Volcanology and Geothermal Research, 179, 191–205. eruptions of 1980. In The 1980 Eruption of Mount
Pfeiffer, T., Costa, A. and Macedonio, G. (2005). A St. Helens, ed. W. P. Lipman and D. R. Mullineaux.
model for the numerical simulation of tephra Washington, D.C.: U.S. Geological Survey
fall deposits. Journal of Volcanology and Geothermal Professional Paper, 1250, 577–600.
Resarch, 140, 273–294. Scasso, R., Corbella, H. and Tiberi, P. (1994).
Prata, A. J. and Grant, I. F. (2001). Retrieval of Sedimentological analysis of the tephra from
microphysical and morphological properties 12–15 August 1991 eruption of Hudson Volcano.
of volcanic ash plumes from satellite data: Bulletin of Volcanology, 56, 121–132.
Application to Mt Ruapehu, New Zealand. Quarterly Scire, J. S., Robe, F. and Yamartino, R. (2000). A User’s
Journal of the Royal Meteorological Society, 127, Guide for the CALMET Meteorological Model. Concord,
2153–2179. MA: Earth Tech, Inc.
Prata, A. J. and Turner, P. J. (1997). Cloud-top height Scollo, S., Del Carlo, P. and Coltelli, M. (2007). Tephra
determination using ATSR data. Remote Sensing of fallout of 2001 Etna flank eruption: Analysis of the
Environment, 59, 1–13.
deposit and plume dispersion. Journal of Volcanology ed. D. Shimozuru and I. Yokoyama. Tokyo: Terra
and Geothermal Research, 160, 147–164. Scientific, pp. 95–113.
Scollo, S., Folch, A. and Costa, A. (2008b). A Tanaka, H. L. and Yamamoto, K. (2002). Numerical
parametric and comparative study of different simulation of volcanic plume dispersal from Usu
tephra fallout models. Journal of Volcanology and volcano in Japan on 31 March 2000 using PUFF
Geothermal Research, 176, 199–211. model. Earth Planets Space, 54, 743–752.
Scollo, S., Tarantola, S., Bonadonna, C., Coltelli, M. Textor, C., Graf, H. F., Herzog, M. et al. (2006).
and Saltelli, A. (2008a). Sensitivity analysis and Volcanic particle aggregation in explosive eruption
uncertanity estimation for tephra dispersal columns. Part II: Numerical experiments. Journal of
models. Journal of Geophysical Research, 113, B06202, Volcanology and Geothermal Research, 150, 378–394.
doi:10.1029/2006JB004864. Thorarinsson, S. (1944). Petrokronologista Studier pa
Searcy, C., Dean, K. and Stringer, W. (1998). PUFF: A Island. Geographes Annuales Stockholm, 26, 1–217.
high-resolution volcanic ash tracking model. Journal Thorarinsson, S. (1954). The eruption of Hekla
of Volcanology and Geothermal Research, 80, 1–16. 1947–1948. In The Tephra Fall from Hekla. Reykjavik:
Settle, M. (1978). Volcanic eruption clouds and Vis Islendinga.
thermal power output of explosive eruptions. Turner, J. S. (1979). Buoyancy Effects in Fluids.
Journal of Volcanological and Geothermal Research, 3, Cambridge: Cambridge University Press.
309–324. Veitch, G. and Woods, A. W. (2001). Particle
Shaw, D. M., Watkins, N. D. and Huang, T. C. (1974). aggregation in volcanic eruption columns. Journal
Atmospherically transported volcanic glass in of Geophysical Research, 106(B11), 26 425–26 441.
deep-sea sediments: Theoretical considerations. Volentik, A., Bonadonna, C., Connor, C. B., Connor,
Journal of Geophysical Research, 79, 3087–3094. L. J. and Rosi, M. (2010). Modeling tephra dispersal
Skamarock, W., Klemp, J., Dudhia, J. et al. (2005). in absence of wind: insights from the climactic
A Description of the Advanced Research WRF Version 2. phase of the 2450BP Plinian eruption of Pululagua
Available online at: www.wrf- model.org. volcano (Ecuador). Journal of Volcanology and
Sorem, R. K. (1982). Volcanic ash clusters: tephra Geothermal Research, 193, 117–136.
rafts and scavengers. Journal of Volcanology and Wadell, H. (1933). Sphericity and roundness of rock
Geothermal Research, 13, 63–71. particles. Journal of Geology, 41, 310–331.
Sparks, R. S. J. (1986). The dimensions and dynamics Walker, G. P. L. (1973). Explosive volcanic eruptions –
of volcanic eruption columns. Bulletin of Volcanology, a new classification scheme. Geologische Rundschau,
48, 3–15. 62, 431–446.
Sparks, R. S. J., Wilson, L. and Sigurdsson, H. (1981). Walker, G. P. L. (1980). The Taupo Pumice: product of
The pyroclastic deposits of the 1875 eruption of the most powerful known (Ultraplinian) eruption?
Askja, Iceland. Philosophical Transactions of the Royal Journal of Volcanology and Geothermal Research, 8,
Society of London, 229, 241–273. 69–94.
Sparks, R. S. J., Carey, S. N. and Sigurdsson, H. (1991). Walker, G. P. L., Wilson, L. and Bowell, E. L. G.
Sedimentation from gravity currents generated by (1971). Explosive volcanic eruptions – I. The rate
turbulent plumes. Sedimentology, 38, 839–856. of fall of pyroclasts. Geophysical Journal of the Royal
Sparks, R. S. J., Bursik, M. I., Ablay, G. J., Thomas, Astronomical Society, 22, 377–383.
R. M. E. and Carey, S. N. (1992). Sedimentation Walker, G. P. L., Self, S. and Wilson, L. (1984).
of tephra by volcanic plumes. 2. Controls on Tarawera, 1886, New Zealand – A basaltic
thickness and grain-size variations of tephra fall Plinian fissure eruption. Journal of Volcanology and
deposits. Bulletin of Volcanology, 54, 685–695. Geothermal Research, 21, 61–78.
Sparks, R. S. J., Bursik, M. I., Carey, S. N. et al. (1997). Wehrmann, H., Bonadonna, C., Freundt, A.,
Volcanic Plumes. Chichester, UK: Wiley. Houghton, B. F. and Kutterolf, S. (2006).
Stull, R. (1988). An Introduction to Boundary Layer Fontana Tephra: A basaltic Plinian eruption in
Meteorology. Dordrecht: Kluwer Academic. Nicaragua. In Volcanic Hazards in Central America,
Sulpizio, R. (2005). Three empirical methods for the Geological Society of America Special Paper, 412,
calculation of distal volume of tephra-fall deposits. pp. 209–223.
Journal of Volcanology and Geothermal Research, 145, Wiesner, M. G., Wang, Y. W. and Zheng, L. (1995).
315–336. Fallout of volcanic ash to the deep South China Sea
Suzuki, T. (1983). A theoretical model for dispersion induced by the 1991 eruption of Mount Pinatubo
of tephra. In Arc Volcanism, Physics and Tectonics, (Philippines). Geology, 23, 885–888.
Wilson, L. and Huang, T. C. (1979). The influence mass/area versus square root of area and fitting
of shape on the atmospheric settling velocity of and integrating both
volcanic ash particles. Earth and Planetary Sciences (i) three exponential segments, and
Letters, 44, 311–324. (ii) a power-law curve.
Wilson, L. and Walker, G. P. L. (1987). Explosive
volcanic eruptions – VI. Ejecta dispersal in Use a deposit density of 1050 kg m–3
plinian eruptions – the control of eruption
conditions and atmospheric properties. Values of square root of area and mass/area
Geophysical Journal of the Royal Astronomical Society, of isomass lines for the Etna 1998 tephra
89, 657–679. deposit.
Wilson, L., Sparks, R. S. J., Huang, T. C. and Watkins,
N. D. (1978). The control of volcanic column height Isomass-line Isomass-line mass/
by eruption energetics and dynamics. Journal of A (km) area (kg m −2)
Geophysical Research, 83, 1829–1836.
1.9 80.0
Witham, C. S., Hort, M. C., Potts, R. et al. (2007).
3.5 25.0
Comparison of VAAC atmospheric dispersion
models using the 1 November 2004 Grimsvotn
4.9 10.0
eruption. Meteorological Applications, 14, 27–38.
7.7 3.0
Woods, A. W. (1988). The fluid dynamics and 11.7 1.5
thermodynamics of eruption columns. Bulletin of 13.1 1.0
Volcanology, 50, 169–193. 16.4 0.5
Zimanowski, B., Wohletz, K., Dellino, P. and 19.2 0.4
Buttner, R. (2003). The volcanic ash problem. 23.8 0.3
Journal of Volcanology and Geothermal Research, 38.8 0.1
122, 1–5.
Online resources available at

www.cambridge.org/fagents
Exercises • Supplement 9A: Details for Table 9.1
• Supplement 9B: Details for Table 9.2
9.1 Calculate the erupted mass of the 22 July 1998 • Supplement 9C: Additional figures
eruption of Etna volcano, Italy, by plotting the • Additional exercises
data in the table below on a semilog plot of • Answers to exercises

Lectura Obligatoria - Bonadonna, C. 2013 - Modeling Tephra Sedimentation From Volcanic Plumes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lectura Obligatoria - Bonadonna, C. 2013 - Modeling Tephra Sedimentation From Volcanic Plumes

Uploaded by

Copyright:

Available Formats

Chapter 9

Modeling tephra sedimentation from

Overview 9.1 Introduction

(a) turbulent fallout from the plume margins and

100 comparison to strong plumes, tephra fall from

deposit (Fig. 9.2(c)) shows maximum accumula-

Hcb tion along the dispersal axis, but as soon as the

Figure 9.1 Diagrams showing (a) the main characteristics

Figure 9.2 (a) Isopach map of

adopting the preliminary observation of

values of k from proximal to distal portions of T T0 ⎜ 0 ⎟ = C pl with C pl T0 A0m 2

to the choice of A0 but not to the choice of Caveats

where cj(x,y,z,t) = c̅j + c′j is the concentra- ∂Uc j ∂v j c j

due to the atmospheric turbulence, whereas the M M0 exp ⎨ − ∫ Hb dx

rate of an umbrella cloud, the grain-size dis- ∂ cj ∂ cj ∂ cj ∂ v jc j ∂ 2c j ∂ 2c j

effects and rapid temporal variations of wind ∂c j ∂c j ∂c j ∂c j

tessellation, and analytical models (Carey and 9.5.2 Parameterizations of particle

Figure 9.3 Plot of drag coeficient vs.

Grain-size Md φ (0.8) _ _ _ _ _ Md φ (-0.6) Md φ (0) _

density of 2600 kg m−3). For the power-law cal-

9.6.2 Application of 1D and 2D analytical

results of the strong plume model of Bonadonna and Phillips

Run 1 Run 2 Run 3 Run 4 Run 5

K H (m2 s −1) – 0.1–8000 0.1–8000 0.5 5000

Figure 9.8 Plot of log(erupted mass)

9.7 Discussion coefficients (Table 9.1). In particular, the stand-

from the six different erupted mass values in

9.7.2 Determination of eruption

MM5 Fifth-Generation NCAR/Penn State Aschenbrenner, B. C. (1956). A new method of

Hydrometeor-enhanced tephra sedimentation: Holasek, R. E. and Self, S. (1995). GOES weather-

Online resources available at

You might also like