Sunteți pe pagina 1din 14

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, B09201, doi:10.

1029/2009JB007175, 2010

A model for wet aggregation of ash particles in volcanic plumes


and clouds:
1. Theoretical formulation
Antonio Costa,1,2 Arnau Folch,3 and Giovanni Macedonio1
Received 30 November 2009; revised 18 March 2010; accepted 14 April 2010; published 1 September 2010.

[1] We develop a model to describe ash aggregates in a volcanic plume. The model is
based on a solution of the classical Smoluchowski equation, obtained by introducing a
similarity variable and a fractal relationship for the number of primary particles in an
aggregate. The considered collision frequency function accounts for different mechanisms
of aggregation, such as Brownian motion, ambient fluid shear, and differential
sedimentation. Although model formulation is general, here only sticking efficiency related
to the presence of water is considered. However, the different binding effect of liquid
water and ice is discerned. The proposed approach represents a first compromise between
the full description of the aggregation process and the need to decrease the computational
time necessary for solving the full Smoluchowski equation. We also perform a
parametric study on the main model parameters and estimate coagulation kernels and
timescales of the aggregation process under simplified conditions of interest in
volcanology. Further analyses and applications to real eruptions are presented in the
companion paper by Folch et al.
Citation: Costa, A., A. Folch, and G. Macedonio (2010), A model for wet aggregation of ash particles in volcanic plumes and
clouds: 1. Theoretical formulation, J. Geophys. Res., 115, B09201, doi:10.1029/2009JB007175.

1. Introduction remote sensing studies [e.g., Rose et al., 2001] and can be
transported up to global distance prior to deposition [e.g.,
[2] During large explosive volcanic eruptions enormous
Rose et al., 2003]. On the other hand, most of fine ash injected
amounts of hot gas and ash rise into the atmosphere forming
during the eruption typically aggregate because of complex
columns up to a few tens of kilometers. An eruption column
interactions of surface liquid layers, electrostatic forces, or
comprises three regions characterized by different flow differences in settling velocities [Sparks et al., 1997]. The
regimes: a lower basal gas thrust or jet region, an inter-
proportion of fine ash in a volcanic eruption varies from
mediate convective region, and an upper umbrella region
1% to 5% for explosive basaltic eruptions to 30% to 50% for
that forms around the neutral buoyancy level (NBL). In the
silicic eruptions [Rose and Durant, 2009]. A major conse-
latter region the erupted material spreads under the effect of
quence of aggregation is the change of the aerodynamic
winds and atmospheric turbulence. Sizes of erupted particles
behavior and the residence time of the finer particles. In fact
vary by several orders of magnitude, ranging from very fine
aggregates settle at different terminal velocities than single
submicron ash to clasts larger than one meter in diameter
particles, and as a consequence can form an anomalous peak in
[Sparks et al., 1997]. The largest and heaviest particles leave
the settling velocity distribution which can produce secondary
the column at lower levels. Most of the particles within the
thickness maxima, such as in the case of 18 May 1980 Mount
intermediate size range are advected by wind, diffused by St. Helens (MSH) eruption [Carey and Sigurdsson, 1982;
turbulence, settled under gravity as single particles and are
Armienti et al., 1988; Durant et al., 2009] or 18 August 1992
finally deposited on the ground at medium to distal distances
Crater Peak eruption [Durant and Rose, 2009]. Durant et al.
[Suzuki, 1983; Armienti et al., 1988; Glaze and Self, 1991; [2009] also proposed a conceptual model for explain the role
Pfeiffer et al., 2005; Costa et al., 2006]. A small fraction of
of mammatus clouds to transport very fine ash from the
very fine ash can persist in the atmosphere in detectable
18 May 1980 MSH eruption. Following their model, ash
concentration for few days to weeks as indicated by satellite
particles initiate ice hydrometeor formation high in the tro-
posphere. As mammatus develop from increased particle
1
loading, volcanic cloud rapidly subsides. At this stage rapid
Istituto Nazionale di Geofisica e Vulcanologia, Naples, Italy. fallout occurs as the cloud passes through the melting level in
2
Department of Earth Sciences, University of Bristol, Bristol, UK.
3
Barcelona Supercomputing Center, Barcelona, Spain. a process analogous to snowflake aggregation, and form the
distal mass deposition maxima observed in many recent
Copyright 2010 by the American Geophysical Union. volcanic ashfall deposits.
0148‐0227/10/2009JB007175

B09201 1 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

[3] Since volcanic ash can produce several harmful effects using a settling velocity of 0.55 m s−1. In order to explain the
such as collapse of roofs, respiratory sickness induced by ash presence of fine ash from the Campanian Ignimbrite (ash
inhalation, crop and livestock losses and risk to air traffic layer Y‐5) within a distance of 1600 km, Cornell et al.
safety [Blong, 1984; Casadevall, 1994; Casadevall et al., [1983] assumed that 50% of the 63–44 mm ash, 75% of
1996; Baxter, 1999; Pomonis et al., 1999; Spence et al., the 44–31 mm ash and 100% of the less than 31 mm ash can
2005; Costa et al., 2006; Macedonio et al., 2008; Folch be treated as aggregate particles with a diameter of 200 mm
et al., 2008; Costa et al., 2009; Folch et al., 2009], under- and a density of 200 kg m−3. Veitch and Woods [2001] first
standing aggregation processes in volcanic plume can con- developed a one‐dimensional model to describe the forma-
tribute to mitigation of the risks related to volcanic ash tion of aggregates in a volcanic eruption column based on
transport and sedimentation. the Smoluchowski [1917] equation. Introducing several sim-
[4] Concerning experimental investigations, Gilbert and plifying assumptions Veitch and Woods [2001] combined a
Lane [1994] showed that growth of lapilli in a recirculating description of the rate of collision and sticking of particles
wind tunnel is controlled by collision and aggregation of with a model of the vertical transport in the eruption column.
liquid‐coated particles. Moreover, from the analysis of their Again, even here an attempt to validate their model was done
experiments they gave some quantitative estimation of the using the deposit of 18 May 1980 MSH eruption. Other
aggregation coefficient, showing its dependence on the grain parametrization for describing ash aggregation in explosive
size: ranging from about 1 for particles of the order of 1 mm eruption columns were suggested by Textor et al. [2006a,
to 10−2 for particles with few hundreds mm diameter. 2006b] who performed some numerical experiments but did
Schumacher and Schmincke [1995] performed few experi- not compare the model with field data.
ments to show how binding among the cohesionless ash [7] In this paper, we (1) introduce the general physical‐
particles is mainly due to the capillary forces of liquid mathematical theory describing the aggregation processes,
bridges. Electrostatic attraction between charged particles (2) formulate a simplified model for volcanic ash aggrega-
is much weaker but effective over larger distances. Their tion which is computationally cheaper than the full solution
experiments also showed that the minimum amount of water of the Smoluchowski [1917] equation, and (3) perform a
necessary for particle aggregation depends on the maximum parametric study for investigating the role of the main model
grain size of ash, from about 5% for ash composed of parameters. In the companion paper, Folch et al. [2010]
particles with diameter <350mm up to 25%–30% for parti- apply the model in the framework of the ash transport
cles with diameter <2000 mm. Moreover, Schumacher and model FALL3D [Costa et al., 2006; Folch et al., 2009] for
Schmincke [1995] analyzed the characteristics of accre- reconstructing the deposits of two well known eruptions, i.e.,
tionary lapilli from the Laacher See Tephra showing how 18 May 1980 Mount St. Helens eruption and 17 September
accretionary lapilli formed from elutriation ashfall have a 1992 Mount Spurr eruption, where ash aggregation played
unimodal distribution with maximum grain size of 250– an important role during the sedimentation.
350 mm whereas lapilli from surge deposits can have a weakly
bimodal distribution with a maximum grain size of about 2. Quantification of the Aggregation Processes
2 mm. However, very recent studies by Brown et al. [2010],
based on the interpretation of field deposit at Tenerife, indi- [8] The rate of change of number density of particles can
cate that the assumptions that most occurrences of accre- be described as [Smoluchowski, 1917; Jacobson, 1999]
tionary lapilli occur in fallout deposits need to be reappraised Z v
dnv 1
because there is evidences that some types form in density ¼ ðs; v  sÞ ðs; v  sÞ nv ðsÞ nv ðv  sÞds
dt 2
currents. Z0 1
[5] More recently, James et al. [2002, 2003] experimen-  ðs; vÞ ðs; vÞ nv ðsÞ nv ðvÞds ð1Þ
tally investigated volcanic particle aggregation in the absence 0

of a liquid phase. They showed that when aggregation pro-


where s denotes the variable of integration, nv is the
cesses are driven by electrostatic charges, aggregates of
number of particles having a volume between v and v + dv
irregular shape up to 800 mm in diameter and density of
per unit of volume of the cloud (m−6), a is called aggre-
∼100–200 kg m−3 are formed. The component particle size
gation efficiency or sticking efficiency function (hereafter
analysis of the aggregates demonstrated that particle dis-
we will use the term sticking efficiency function), and b is
tributions are dominated by particles ∼10–40 mm.
called aggregation kernel or collision frequency function
[6] From a modeling perspective, there have been only
(hereafter we will use collision frequency function). The
few studies attempting to describe in a quantitative way the
sticking efficiency a and collision frequency b appear always
effect of aggregation processes on models describing trans-
as a product that is named coagulation kernel K = ab (the
port of volcanic ash in both eruption column [e.g., Veitch and
coagulation kernel is also called collection kernel or coales-
Woods, 2001; Textor et al., 2006a, 2006b] and in volcanic
cence kernel, hereafter we will use only coagulation kernel
clouds [e.g., Carey and Sigurdsson, 1982; Cornell et al.,
or simply kernel). The first term on the right‐hand side of
1983; Armienti et al., 1988; Bonadonna et al., 2002]. For
equation (1) represents the rate of formation of aggregates
instance, in order to reproduce the secondary thickness
with volume between v and v + dv, whereas the second term
maximum of the 18 May 1980 MSH eruption, known as
denotes the rate of loss of aggregates of volume between v
Ritzville bulge, Carey and Sigurdsson [1982] imposed that
and v + dv to form larger aggregates. In the case of aggre-
all particles with diameter <63mm fell with a settling velocity
gation of solid particles such as volcanic ash, we have to
of 0.35 m s−1. Armienti et al. [1988] obtained similar results

2 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

consider the conservation of particle mass instead of particle vidual contributions [e.g., Jacobson, 1999]. Considering two
volume. Therefore, equation (1) should to be written con- classes of particles i and j with associated masses of mi =
sidering the rate of change of number density of particles with ripdi3/6 and mj = rjpdj3/6, we have
mass between m and m + dm instead of particles with volume  2
between v and v + dv: 2 kb T di þ dj
 ¼ B þ S þ DS ¼
3  di dj
Z m 1  3   2  
d
nðmÞ ¼
1
K ðs; m  sÞ n ðsÞ n ðm  sÞds þ GS di þ dj þ di þ dj Vsj  Vsi  ð3Þ
dt 2 6 4
Z0 1
 K ðm; sÞ n ðsÞ n ðmÞds ð2Þ where kb is the Boltzmann constant, T is the absolute tem-
0 perature, d is the particle diameter, m is the dynamic viscosity
of air, GS is the fluid shear that is GS = dU/dz in case of a
where s denotes the variable of integration and n is the laminar flow, and GS = (1.3/n)1/2 in case of a turbulent flow
number of particles having a mass between m and m + dm (U denotes the air velocity, n the air kinematic viscosity
per unit of volume (m−3 kg−1). If for sake of simplicity we and  is the rate of dissipation of turbulent kinetic energy),
assume that all the primary particles have the same density and Vsi and Vsj are the settling velocities of the particle
rp then we have n = nv /rp. Information on mass size classes of class i and j, respectively.
has to be transferred to the particle classes considered in the 2.2. Sticking Efficiency Function
volcanic ash transport model. For example, in the case of
the FALL3D model [Costa et al., 2006; Folch et al., 2009] [11] Concerning the sticking efficiency, there is no general
a particle class is defined by a triplet of values that are model for describing a(mi, mj) in the case of volcanic ash. If,
equivalent diameter, density, and a shape factor, respectively. as a first‐order approach, we neglect electrostatic aggregation
Equation (2) should be integrated for all classes between a [Schumacher and Schmincke, 1995], two different approx-
minimum aggregating class kmin and a maximum aggregating imations can be used depending upon the presence of liquid
class kmax. In the kmin class (typically of the order of few water or ice.
microns) there is no production terms, while in kmax class [12] 1. Below the freezing temperature of water (Tf) we
(typically of the order of a few hundreds microns) the con- treat the volcanic particles as ice crystals. In the case of ice
tribution term to form larger aggregate is negligible. [Field et al., 2006] found the following empirical parame-
[9] Concerning the limitations of the proposed model, we terization for the sticking efficiency:
need to stress the fact that there are only few experimental  
studies such as those by Gilbert and Lane [1994], for the case mbi þ mbj
ij ¼ B  2   ð4Þ
of wet aggregation, and James et al. [2002], for dry aggre-  Vsj  Vsi 
4 di þ dj
gation driven by electrostatic forces. However, it is difficult
to extract all the information necessary for describing sys- where b = 0.47, B = 6.1 × 10−4 kg−b m3 s−1. However they
tematically the sticking efficiency, due to, e.g., capillary showed that even a constant efficiency of 0.09 produced
forces and electrostatic attraction, as function of pivotal very good agreement with the experimental data. Here, for
parameters such as water content or particle diameter. An sake of simplicity, in presence of ice we use
analysis of the main mechanisms that can originate volcanic
ash aggregate is reported by Schumacher and Schmincke ij ¼ 0:09 ð5Þ
[1995]. Although electrostatic forces can be an important
mechanism for aggregating fine particles [James et al., 2002, Obviously, since ash particles may not be completely cov-
2003], they are much weaker than capillary forces by several ered of ice, this assumption may overestimate the real effi-
orders of magnitude [Schumacher and Schmincke, 1995]. ciency. This is a drastic simplification to be used as starting
There are also disaggregation processes (e.g., collisions that point and need to be removed in a further development of
destroy or split aggregates) but without any research results the model when more experimental data will be available.
in this area, for simplicity here we neglect them. Here, even Freezing of water in presence of volcanic ash is a stochastic
if we formulate the model in a general way, as a first process that occurs over a temperature range of Tf < 250–
approximation, we restrict our analysis to aggregation pro- 260 K through heterogeneous nucleation initiated by vol-
cesses in presence of water. Dry aggregation processes will canic ash [Durant et al., 2008], whereas melting occurs
be considered in a further study as collision frequency always around Tm = 273 K.
function and sticking efficiency due to electrostatic forces [13] 2. Above the melting temperature of ice (Tm), for par-
will be better constrained. ticles with a roughened surface covered by liquid [Barnocky
and Davis, 1988], Liu and Litster [2002] showed that stick-
2.1. Collision Frequency Function ing is a function of the viscous Stokes number Stij based on the
[10] The collision frequency function b contains informa- binder liquid viscosity ml:
tion on the mechanisms that bring particles together. The    
main categories of mechanism are due to Brownian motion 8p Vj  Vi  di dj 8p Vsj  Vsi  di dj
Stij ¼  ð6Þ
(b B), both laminar and turbulent fluid shear (b S) and differ- 9l di þ dj 9l di þ dj
ential sedimentation (b DS). A collision frequency function
can be associated to each of these mechanisms, and the total where rp denotes the average density of the primary parti-
collision frequency function is given by the sum of the indi- cles. The ratio de = (didj)/(di + dj) is commonly denoted as

3 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 1. Sticking efficiency as function of an effective Stokes number obtained from the analysis of
Gilbert and Lane [1994] (crosses). The solid line represents equation (8) with Stcr = 1.3 and q = 0.8.

reduced diameter of two particles [Liu and Litster, 2002]. ciency function needs to be quantified also for volcanic ash
Liu and Litster [2002] presented a parametrization for the aggregation without the presence of water.
coagulation kernel as function of the particle Stokes number.
Following Liu and Litster [2002], we assume that coagulation 3. A Simplified Aggregation Model
occurs when the Stokes number is lower than a critical value:
[15] The integration of equation (2) for all the particle
 
hl classes involved in a volcanic plume and over a large com-
Stij < Stcr ¼ ln ð7Þ putational domain is computationally extremely expensive or
ha
even impossible in most cases. For this reason we need to
where hl is the liquid binder layer thickness and ha is the adopt a simplified strategy that contains the basic physics
granule surface asperity. Liu and Litster [2002] suggest that behind equation (2) but, at the same time, is computationally
the coagulation kernel is null for Stij > Stcr and assume a fast and efficient.
constant value as Stij < Stcr (with Stcr of the order of unity). [16] Several experimental results have confirmed that
However, following a Stokes number analysis, most of the aggregates have a fractal geometry [e.g., Vicsek, 1992; Li
experimental results generally show that (1) for Stij  Stcr et al., 1998; Lee and Kramer, 2004; Burd and Jackson,
all collisions are successful, (2) for Stij ≈ Stcr some collisions 2009]. For an aggregate formed from initially identical
are successful, and (3) Stij  Stcr no collision is successful primary particles having an effective diameter dj, the number
[e.g., Walker et al., 2006]. Following the results above, and of primary particles (Nj) in an aggregate can be estimated
combining them with experimental data for volcanic ash by using a well‐known fractal relationship [Jullien and Botet,
Gilbert and Lane [1994] and considerations made by Veitch 1987; Frenklach, 2002; Lee and Kramer, 2004]:
and Woods [2001], for T ≥ Tf we propose the following  Df
parameterization for the sticking efficiency: dA
Nj ¼ kf ð9Þ
dj
1
ij ¼  q ð8Þ
1 þ Stij =Stcr where dA = 2Rg with Rg the radius of gyration of an
aggregate (i.e., the root mean square distance of the
where, using the data of Gilbert and Lane [1994], we esti- aggregate parts from its center of gravity), Df the fractal
mated Stcr = 1.3 and q = 0.8 (see Figure 1). Parametrization dimension, and kf the fractal prefactor of the order of the
for aij (equation (8)) will be improved and better calibrated unity (∼0.6–1). The fractal dimension Df furnishes a mea-
as soon as more general experimental data will be available. sure of how much the aggregate fills the available space. A
[14] The model we formulate below can be extended to three‐dimensional solid object would have Df = 3. Values of
account for dry aggregation, but in this case a sticking effi- Df less than three characterize objects that do not completely
fill the three‐dimensional space available to them. Fractal

4 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

dimensions greater than three are physically impossible [21] 4. In order to simplify the algebra for calculating the
[Burd and Jackson, 2009]. The fractal dimension depends different contributions to the coagulation kernel we consider
on both the collision and the stickiness mechanisms that an averaged value of the sticking efficiency. In particular we
formed the aggregates [Jiang and Logan, 1991]. use the class‐averaged sticking efficiency am is computed as
[17] We started from relationship (9) to formulate a sim- P P
plified model for describing the aggregation processes in i j fi fj ij
m ¼ P P ð13Þ
volcanic ash. The model we suggest consists of the following i j fi fj
steps:
[18] 1. On the basis of experimental and field observations where fk is the class mass fraction, aij is the sticking effi-
[Carey and Sigurdsson, 1982; Cornell et al., 1983; Gilbert ciency between the classes i and j, and the sums on i and j
and Lane, 1994; James et al., 2002, 2003; Bonadonna et al., are performed over the aggregating classes between kmin and
2002; Durant et al., 2009] we assume that the majority of kmax. A better more rigorous quantification of the kernel,
aggregating particles form only one aggregation class of which account for the dependence of the sticking efficiency
particles with diameter dA and density rA which depend on variation with particle diameters, is the object of ongoing
the aggregation process, i.e., on the values Df that can researches.
change depending on the classes and processes. [22] 5. Considering the coagulation kernel contribution
[19] 2. On each computational element (subvolume) of due to the Brownian motion we can write [Dekkers and
dimension dx × dy × dz we can calculate the number of Friedlander, 2002]
primary particles of class j available. In fact the particle  2
number of class j per unit of volume is   2kT di þ dj
KB vi ; vj ¼ m B;ij ¼ m
3 dj dj
6Cj  
nj ¼ ð10Þ 1=Df 1=Df 2
þ vj
p dj3 2kT vi
¼ m 1=D 1=D
ð14Þ
3 v fv f
where Cj is the mass concentration of particles j and rp is i j

the typical density of primary particles. Classes of aggre- where we used the fractal size‐volume relationship di = xv1/D
i
f

gating particles will range from the class kmin to the class and the average sticking efficiency am (x is a factor which
of kmax which is the class immediately smaller than the class depends on both fractal dimension and object shape). For
of aggregates. In order to calculate the fraction of primary example in the Euclidean space (Df = 3) in the case of a
particles that potentially can participate to form aggregate we sphere we have x = (6/p)1/3 ’ 1.24, whereas for a cube x = 1.
will proceed as described in the points below. However, here we assume that, for the typical values of
[20] 3. For many aggregating systems it was observed that Df considered here, x will remain of the order of the unity.
after sufficiently long times [e.g., Tirado‐Miranda et al., Inserting relationship (14) in equation (11) and introducing
2000; Gmachowski, 2001; Dekkers and Friedlander, 2002; the similarity variable (12), we obtain
Dekkers et al., 2002; Di Stasio et al., 2002] the particle size
distribution leads to a universal time independent cluster dntot kT
¼ m n2tot
size distribution and, under some assumptions, to a self‐ dt 3
preserving form. This implies that the relative shape of the Z 1Z 1  1=Df 2
x þ y1=Df
cluster size distribution remains constant during the whole  ð xÞ ð yÞdxdy
0 0 x1=Df y1=Df
aggregation process and it is also not sensitive to the par-  
ticular initial conditions [Tirado‐Miranda et al., 2000; 2kT
¼ m 1 þ ð1=Df Þ ð1=Df Þ n2tot
Westbrook et al., 2004]. If we integrate equation (1) over 3
all particle sizes we can obtain the rate of the total par- 4kT 2
 m n  m AB n2tot ð15Þ
R 1 dntot/dt due to the aggregation process where
ticle decay 3 tot
ntot = 0 nv dv (m−3):
where x and y denote the dimensionless
R1 variables vi /vm and
Z 1Z vj /vm, respectively, and mi = 0 xiy(x)dx represents the ith
dntot 1 1    
¼ K vi ; vj n ðvi ; tÞ n vj ; t dvi dvj ð11Þ momentum of the distribution [Dekkers and Friedlander,
dt 2 0 0
2002], am is an average sticking efficiency and AB denotes
the time evolution operator in the case of agglomerates
where vi denotes the volume of the particle i. If we want to formed by Brownian aggregation. We also used the approx-
express equation (11) in terms particle mass mi we need to imation m(1/Df) ≈ 1 and m(−1/Df) ≈ 1 [e.g., Gmachowski, 1995;
replace mi = rivi. Assuming that the coagulation kernel is a Jun et al., 2004].
homogeneous function of the particle volumes, it was shown [23] 6. For calculating dntot /dt in the case of shear
[Dekkers and Friedlander, 2002] that dntot/dt can be calcu- aggregation we can write the coagulation kernel as
lated introducing a distribution y function of the similarity
variable z in the form   GS  3
KS vi ; vj ¼ m S;ij ¼ m di þ dj
ðzÞn2tot 6
n¼ ð12Þ

GS 3  1=Df 
1=D 3
¼ m vi þ vj f ð16Þ
where  is the solid volume fraction in the system and 6
z = vntot /.

5 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Inserting these results in equation (11), we can approximate and considering m(4/Df ), m(3/Df ) and m(1/Df ) about one, we have
equation for ntot as finally
 
dntot GS dntot  p   g 4 4=Df 24=Df
23=D
¼ m 3 3=Df ntot f  m  ntot
dt 6 dt 48
Z Z
1 1 1  1=Df 3 24=Df
 m ADS 4=Df ntot ð22Þ
 x þ y1=Df ð xÞ ð yÞdxdy
2 0 0
GS  
23=D where ADS denotes the time evolution operator in the case of
¼ m 3 3=Df ntot f ð3=Df Þ þ 3ð2=Df Þ ð1=Df Þ
6 agglomerates formed by differential sedimentation.
2 3 23=D 23=D [25] 8. In practical cases, the integrals above have to be
  m GS 3=Df ntot f  m AS 3=Df ntot f ð17Þ
3 calculated between the volumes corresponding to the dia-
meters dmin and dmax instead of 0 and 1. The last assumption
where we assumed that momenta of order 1/Df, 2/Df and is a reasonable approximation since we have dmin  dmax.
3/Df are close to one [Gmachowski, 1995]. The goodness However, in order to formally extend the integration limits we
of the last assumption was confirmed also by the full may define the kernels equal to zero outside of the range from
numerical solution of equation (1). AS denotes the time evo- dmin to dmax.
lution operator in the case of agglomerates formed by shear [26] 9. Finally, assuming that the collision frequency
aggregation. functions are additive, we obtain the following heuristic
[24] 7. For estimating the kernel contribution due to the equation for Dntot :
differential settling velocity we proceed in a similar way. In
 
order to simplify the algebra we assume that drag of primary 23=D 24=D
Dntot ¼ m AB n2tot þ AS 3=Df ntot f þ ADS 4=Df ntot f Dt
particles participating to the aggregation can be described by
the Stokes law, i.e., their settling velocity scale with the ð23Þ
square of the particle diameter that is typically valid for
diameter up to ∼100 mm [e.g., Pfeiffer et al., 2005]. This where the computational time step for the transport model
assumption is legitimate as most of the aggregating particles Dt should be smaller than the characteristic aggregation
have a diameter smaller than 100 mm. Under these assump- timescale [Tirado‐Miranda et al., 2000]:
tions we obtain
  2
KDS vi ; vj ¼ m DS;ij tagg ¼ ð24Þ
  ntot Kmax
 p   g  2 
 2
¼ m di  dj2  di þ dj ð18Þ [27] 10. Let be na the total number of aggregates per unit
4 18
of volume, ni the number of primary particles of class i, Dna
where r denotes the air density. Moreover, since the differ- and Dni the relative increments in a time Dt. As we men-
ential settling velocity kernel is large only between particles i tioned above, for simplicity we assume that all particles
and j having di  dj, we can use the approximation aggregate in P
one aggregation
PNp class, therefore, we will have
      (1) Dntot = 1 Dn ≈ Dni, and (2) NiDnA = Dni,
  i i
Vsj  Vsi  ¼ p   g d 2 1  d 2 =d 2   p   g d 2
i
i j i i ð19Þ where Np are the classes of primary particles of diameter di,
18 18 i = 1,…, Np. The number of primary particle in an aggre-
Therefore, we can approximate the kernel as gate Ni can be calculated using the relationship PN (9).
Combining steps 1 and 2 we have DnA ≈ Dntot / i p Ni and
     hence, on each cell and at each time step, we can calculate
   p   g 4  dj2  dj 2 the number of particle of the class i going to form an
K vi ; vj ¼ m di 1  2  1 þ
4 18  di  di aggregate as
   
 p   g 4 dj
 m di 1 þ 2 Dntot Ni
4 18 di Dni  P ð25Þ
    i Ni
 p   g 4 4=Df 3=D 1=D
¼ m vi þ 2vi f vj f ð20Þ
72  [28] 11. The class j on a cell will participate to the
aggregation process depending on the local Stokes number
Inserting the last relationship in equation (11), we have based on the binder liquid viscosity (6) and/or the presence
of ice. The parameters dA and Df are derived from ex-
  periments and field observations depending on the aggre-
dntot  p   g 4 4=Df 24=Df
 m  ntot gation process (for example the amount of water).
dt 72 [29] 12. In the FALL3D model [Costa et al., 2006; Folch
Z 1Z 1 
1 et al., 2010] each class is represented by a triplet charac-
 x4=Df þ2x3=Df y1=Df ð xÞ ð yÞdxdy
2 0 0 terizing the particles, such as equivalent diameter, particle
 
 p   g 4 4=Df 24=Df  density, and particle shape. Particle classes are commonly
¼ m  ntot ð4=Df Þ considered to be noninteracting. When we consider aggre-
144
 gation processes an interaction term (representing a sink/
þ 2ð3=Df Þ ð1=Df Þ ð21Þ source term) has to be introduced in each class and infor-
mation among classes updated at each time step.

6 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Table 1. Values Used in the Calculations for the Parametric Study From relationship (28) we can see that for Df = 3, (1 − kf)
Parameter Water Ice Unit
represents the initial porosity. If we fix Df = 3, for aggregates
of spherical particles, kf typically ranges from 0.64 for a
Ta 50 −50 °C random loose monodisperse packing to 0.91 for very efficient
ra (at 9 km) 0.45 0.45 kg m−3 polydisperse packing [Voivret et al., 2007]. In general (other
rf,air 0.45 0.45 kg m−3
rf,water 1000 1000 kg m−3 particle shapes or fractal dimensions, droplet coagulation,
rp 2500 2500 kg m−3 etc.), kf can be up to 1 or even larger. Here, for simplicity, we
GS 0.0045 0.0045 s−1 fixed kf = 1. When the interstitial spaces in aggregates are
ma 1.98 × 10−5 1.47 × 10−5 Pa s filled of fluid (typically air and vapor mixture or liquid
ml 5.43 × 10−4 – Pa s
water) the density is
A ¼ ð1  ’Þp þ ’ f ð29Þ
[30] 13. In the case of fractal aggregates the settling
where rf is the density of the interstitial fluid and ’ is
velocity may have a different scaling than that predicted by
given by equation (28).
Stoke’s law and it has to be calculated as function of Df
[33] Plots of the density of aggregates for different values
as well [Johnson et al., 1996; Li and Logan, 2001; Tian
of the fractal exponent Df and aggregates diameter dA, made
et al., 2007]. Velocities calculated not considering the fractal
of primary particles of 1 mm and 10 mm, respectively, with
properties of the aggregates can be up to an order of magni-
air (actually a mixture of air and vapor) and liquid water in
tude higher than measured velocities [Johnson et al., 1996]. If
the interstitial space, are reported in Figure 2.
all the primary particles in the aggregates have the same
density rp and the same diameter dp, then, it is possible to 4.2. Settling Velocity of Ash Aggregates
calculate the settling velocity of an aggregate falling in the [34] Settling velocity of aggregates differs from the settling
Stokes regime as [Tian et al., 2007] velocity of compact particles for two reasons [e.g., Johnson
 
p   g 2   et al., 1996; Tian et al., 2007]. First, their density can be
VsA ¼ dA ð1  ’Þ f k; dA ; dp ð26Þ notably lower than their constituents as they incorporate a
18
lighter fluid in the interstices (air and vapor or liquid water).
where k is the hydraulic permeability of the aggregate and Secondly, because their porosity, aggregates do not fall as
f is a correction function derived from the Brinkman equa- impermeable particles being affected by their intrinsic per-
tions [Tian et al., 2007] that contains few adjustable para- meability. These effects act in different ways, since a
meters. Under these simplification we can write also 1 − ’ = decrease in density decreases the settling velocity, whereas
(1 − ’0)(dA/dp)Df −3 with ’0 denoting the porosity for Df = 3. an increase in permeability increases the settling velocity
For an impermeable sphere f is equal to one. For the case [e.g., Johnson et al., 1996]. As we mentioned above, to
of polidisperse aggregates generally, and for volcanic ash account for these effects, we adopt the relationship (27) that
aggregates in particular, there is no general experimental can be expressed as
data or theory to use, therefore for sake of simplicity, we
use the following equation: VsA ¼ VsC e
bsA f
¼V

VsA ¼ VsC e ð27Þ where VsA is the settling velocity of the aggregate, VsC is the
settling velocity of a compact particle with the same diameter
where ye is an empirical correction that depends on the of the aggregate and the same density of the primary particles,
aggregate porosity and permeability, and VsC is the settling bsA is the settling velocity of a compact particle with the same
V
velocity of a compact particle with the same diameter of the diameter and density of the aggregate, and f is an empirical
aggregate and the same density of the primary particles. correction factor that accounts for the particle permeability
and shape. Considering aggregates falling in the Stokes
4. Parametric Study regime (Re < 1), equation (27) becomes
[31] In this section we perform a parametric study on the  
p   g 2
aggregation model described above. For sake of computa- VsA ¼ dA e
18
tional needs we restrict our analysis to some simplified
conditions only. In particular, we consider primary particles that implies the relationship
having the same density and two different kind of initial
distributions (i.e., monodispersed and Gaussian). We ana- e ¼ ð1  ’Þf ð30Þ
lyze aggregation in presence of either water or ice. The fixed
properties used in the calculations are reported in Table 1. For aggregate with low permeability, f ≈ 1, being equal to
one for impermeable spherical particles. In the case of an
4.1. Density of Ash Aggregates aggregate composed only of a class of primary particles, ye
[32] Let us consider aggregates composed of one class of can be estimated using some theoretical or semiempirical
primary particles only (monodispersed). Using relationship (9) corrections [e.g., Tian et al., 2007; Li and Logan, 2001;
the porosity of the aggregates can be written as Johnson et al., 1996]. Figure 3 shows f as function of the
 Df 3 aggregate porosity calculated using the model described by
dA Tian et al. [2007]. For fractal aggregates with a large
’ ¼ 1  kf ð28Þ
dp porosity, the correction to the settling velocity can be quite

7 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 2. Density of the aggregate constituted of monodispersed primary particles of (a and c) 1 mm and
(b and d) 10 mm, with air and water vapor (Figures 2a and 2b) and liquid water (Figures 2c and 2d) in the
interstitial space, as a function of diameter, for different values of the fractal exponent Df .

relevant (up to an order of magnitude), whereas for porosity the fractal exponent and for monodispersed primary particles
lower than about 0.8, f ≈ 1. of 1 mm and 10 mm with air (actually a mixture of air and
[35] Figure 4 reports settling velocities of the aggregates vapor) and liquid water in the interstices. As shown in
as a function of aggregate diameter for different values of Figure 4, when the interstices are filled with water or ice

Figure 3. Correction to the settling velocity of a fractal aggregate, f, as function of its mean porosity
calculated in accord to Tian et al. [2007] for different dA/dp ratios.

8 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 4. Comparison of the settling velocity of aggregates constituted of monodispersed primary par-
ticles of 1 mm (solid lines) and 10 mm (dotted lines) having (a) air and (b) liquid water, respectively, in the
interstitial spaces for different values of the fractal exponent.

aggregate fractal exponent and porosity do not affect sig- efficiency am (equation (13)), through equation (22) we can
nificantly the terminal velocity of aggregates, except for a calculate the total decay rate of particles 1/ntot dntot/dt (1/s)
relatively low decrease in density. due to aggregation as a function of the fractal exponent Df
for both initial distributions. Results are shown in Figure 6,
4.3. Coagulation Kernels and Timescales which also reports the single contributes of the Brownian,
[36] Assuming the physical properties as in Table 1, fluid shear and differential sedimentation processes. For Df
through relationships (3) and (8), we can estimate the three larger than 2.8 typical aggregation timescales are of the
main components of the coagulation kernel, i.e., those due order of 100 s. As it can be seen from the comparison of the
to Brownian motion (KB), to fluid shear (KS) and to differ- two cases reported in Figure 6, the initial grain size distri-
ential sedimentation (KDS). Figure 5 plots the values of the bution of primary particles has a pivotal role in determining
kernels for the binary interaction between particles of 1 mm, aggregation regimes.
10 mm and 100 mm in diameter with all particles of diameter [38] Finally, in order to compare, and independently esti-
from 1 mm to 1 mm for Df = 2 and Df = 3. mate, the typical timescales for ash aggregation we solved
[37] Considering that the initial particle distribution is a equation (1) under the simplified conditions described below.
Gaussian distribution with sF = 1F centered at mF = 6 (i.e., In particular we assumed that primary particles have initially
16 mm) and mF = 4 (i.e., 63 mm) and the average sticking a Gaussian distribution in F centered at m = 5 and variance

9 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 5. Plot of the contributes of the Brownian, fluid shear, and differential velocity kernels for col-
lision between particles of 100 mm and particles in the range between 1 mm and 1 mm. Aggregate den-
sities are evaluated according to equation (29) with (a, c, and e) Df = 2 and (b, d, and f) Df = 3.

sF = 1. These conditions were chosen as a compromise in volume of each class with respect to the following one.
between computational costs and observations. In fact, typi- We assume that the density of the primary particles is
cal values for fine volcanic ash are m ≈ 4 − 6 sF = 1–3 [Rose 2500 kg m−3, and the density of the aggregates scales
and Durant, 2009; Durant et al., 2009]. For computational according to equation (28) and equation (29) with dp = 10mm
reasons, the only considered particles range in diameter and air (actually a mixture of air and vapor) as interstitial
between 10 and 250 mm, organized in 15000 classes uni- fluid. For what concerns the sticking efficiency, it is assumed
formly spaced in the particle volume. The smallest volume of that the particles collide in presence of liquid water and ice.
the considered particles is the volume of the particles of The initial concentration of particles in the cloud is set to a
31 mm and this volume also corresponds to the increment typical value of 0.1 kg m−3. For the other parameters, see

10 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 6. Plot of the total contributes to the aggregation rate, 1/ntot dntot/dt, (s−1) as a function of the
fractal exponent (solid line), and single contributes due to Brownian processes (dash‐dotted line), fluid
shear (dotted line), and differential sedimentation (dashed line) for an initial Gaussian distribution with
variance s = 1F and centered at (a) mF = 6F, i.e., 16 mm, and (b) at mF = 4F, i.e., 63 mm. Settling velocity
model of Arastoopour et al. [1982].

Table 1. Figure 7 shows the time evolution of the grain size 107 kg s−1 at a height of about 8 km [Veitch and Woods,
spectrum obtained by solving the Smoluchowski equation 2001]) we will have aggregation timescales that are 1/4 of
with Df = 3 and kernels calculated for both conditions, i.e., those reported in Figure 7.
with liquid water and ice. For the case reported in Figure 7 [39] The estimation of aggregation timescales has impor-
we found that time needed for forming a distribution peaked tant implications for the aggregation mechanisms themselves.
around 200 mm is 60 s, which is about 5 times the char- In fact, if aggregation timescales are much shorter than the
acteristic time 1/n dn/dt = 12.5 s (using the approximate residence time of particles in the “liquid” water window of
model we propose we have 1/n dn/dt = 20 s). Further com- the column, the aggregation will be more effective and
putational results (not reported here) confirmed that aggre- most particles will aggregate in that window. On the other
gation timescales are inversely proportional to the initial hand, when aggregation timescales are larger than the time
concentration as predicted by equation (24). For instance, for crossing the “liquid” window, the aggregation will occur
for an initial concentration of particles of 0.4 kg m−3 outside of this region and particles will aggregate mostly in
(corresponding to the density of a plume with mass flux of the ice region (or during the ice remelting). For example,

11 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Figure 7. Time evolution of the grain size spectrum due to particle aggregation considering an initial
particle concentration 0.1 kg m−3 distributed following a Gaussian distribution centered at m = 5 and
variance sF = 1, for a kernel calculated in presence of liquid water (solid line) and of ice (dashed line).
Curves represent different time steps between t = 0 and t = 60 s, spaced by 10 s. Fractal exponent Df was
assumed 3.0.

for a mass flux of 107 kg s−1 a fluid element will take particularly suitable for being incorporated into volcanic ash
around 40 s to pass through the 4 km height window with transport models. We also presented a parametric study on
liquid water in the column (from 6 to 10 km [Veitch and the effects of the main model parameters in ranges and
Woods, 2001]). This timescale is comparable with aggrega- conditions of interest for volcanology. We hope our pre-
tion timescale estimated by the solution of the Smoluchowski liminary results will stimulate further theoretical and
equation. This implies that for mass fluxes much lower than experimental studies that help to better constrain the gov-
107 kg s−1 (which corresponds to a column of ⪅15 km) erning parameters and contribute to a further simplification
aggregation can be very effective and occur inside the column and to a larger applicability of the model.
forming wet aggregates. For larger mass fluxes the time
needed by a fluid element to cross the “liquid” region is much Notation
shorter and aggregation is less effective and mainly due to
the interaction of the ice crystal or to the ice remelting. t time.
nv number of particles having a volume between v and
v + dv per unit of volume.
5. Conclusion n number of particles having a mass between m and
[40] A new model for describing ash aggregation in vol- m + dm per unit of volume.
canic plume was formulated. For the first time we explored a sticking efficiency function.
the possibility to use a simplified approach that conjugate a am class‐averaged sticking efficiency.
solid theoretical background with the need to decrease the b collision frequency function.
computational costs. More realistic coagulation kernels were b B collision frequency function due to Brownian
estimated using a fractal size‐volume relationship. The motion.
model considers three different mechanisms of particle bS collision frequency function due to fluid shear.
collision, such as Brownian motion, ambient fluid shear and b DS collision frequency function due to differential
differential sedimentation. Although the model formulation sedimentation.
is general, as a first approximation, here we consider only K coagulation kernel (K = ab).
particle sticking in presence of liquid water or ice, ne- di diameter of the i particle class.
glecting for instance electrostatic aggregation (considered vi volume of the i particle class.
orders of magnitude smaller). The proposed model re- mi mass of the i particle class.
presents a compromise between previous simple models and fi mass fraction of the i particle class.
the full integration of the Smoluchowski equation and is Vsi settling velocity of the i particle class.
kmin minimum aggregating class.

12 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

kmax particle diameter corresponding to the minimum two anonymous reviewers for their useful comments that improved the
clarity of the paper.
aggregating class.
dmin particle diameter corresponding to the maximum
aggregating class.
dmax maximum aggregating class. References
kb Boltzmann constant. Arastoopour, H., S. Lin, and S. Weil (1982), Analysis of vertical pneumatic
T absolute temperature. conveying of solids using multiphase flow models, AIChE J., 28, 467–473,
doi:10.1002/aic.690280315.
Tf freezing temperature of water. Armienti, P., G. Macedonio, and M. Pareschi (1988), A numerical model
Tm melting temperature of ice. for simulation of tephra transport and deposition: Applications to
m dynamic viscosity of air. May 18, 1980, Mount St. Helens eruption, J. Geophys. Res., 93,
6463–6476, doi:10.1029/JB093iB06p06463.
ml dynamic viscosity of water. Barnocky, G., and R. Davis (1988), Elastohydrodynamic collision and
GS fluid shear. rebound of spheres: Experimental verification, Phys. Fluids, 31,
b, B empirical constants in equation (4). 1324–1329, doi:10.1063/1.866725.
Sti,j Stokes number relative to a couple of ith and Baxter, P. (1999), Impacts of eruptions on human health, in Encyclopedia
of Volcanoes, edited by H. Sigurdsson et al., pp. 1035–1043, Academic,
jth particles. San Diego, Calif.
Stcr critical Stokes number. Blong, R. (1984), Volcanic Hazards: A Sourcebook on the Effects of
hl liquid binder layer thickness. Eruptions, Academic, North Ryde, N. S. W., Australia.
Bonadonna, C., G. Macedonio, and R. Sparks (2002), Numerical modelling
ha granule surface asperity. of tephra fallout associated with dome collapses and Vulcanian explo-
q empirical constant in equation (8). sions: application to hazard assessment on Montserrat, in The Eruption
Nj number of primary particles with diameter dj of Soufriére Hills Volcano, Montserrat, From 1995 to 1999, edited by
T. Druitt and B. Kokelaar, pp. 517–537, Geol. Soc., London.
in an aggregate. Brown, R., M. Branney, C. Maher, and P. Davila‐Harris (2010), Origin of
dA diameter of gyration of an aggregate. accretionary lapilli within ground‐hugging density currents: Evidence
rA density of aggregate. from pyroclastic couplets on Tenerife, Geol. Soc. Am. Bull., 122,
rp density of primary particle. 305–320, doi:10.1130/B26449.1.
Burd, A., and G. Jackson (2009), Particle aggregation, Annu. Rev. Mater.
rf density of interstitial fluid. Sci., 1, 65–90, doi:10.1146/annurev.marine.010908.163904.
r density of air. Carey, S., and H. Sigurdsson (1982), Influence of particle aggregation on
Df aggregate fractal dimension. deposition of distal tephra from the May 18, 1980, eruption of Mount
St. Helens volcano, J. Geophys. Res., 87(B8), 7061–7072, doi:10.1029/
kf fractal prefactor. JB087iB08p07061.
ntot total number of primary particles per unit of volume. Casadevall, T. (1994), The 1989–1990 eruption of Redoubt Volcano,
nj number of primary particles with diameter dj per unit Alaska: Impacts on aircraft operations, J. Volcanol. Geotherm. Res., 62,
of volume. 301–316, doi:10.1016/0377-0273(94)90038-8.
Casadevall, T., P. Delos Reyes, and D. Schneider (1996), The 1991 Pinatubo
z similarity variable. eruptions and their effects on aircraft operations, in Fire and Mud: Erup-
y similarity distribution. tions and Lahars of Mount Pinatubo, Philippines, edited by C. Newhall
 solid volume fraction in the system. and R. Punongbayan, pp. 625–636, Univ. of Wash. Press, Seattle.
Cornell, W., S. Carey, and H. Sigurdsson (1983), Computer simulation and
x size‐volume fractal prefactor. transport of the Campanian Y‐5 ash, J. Volcanol. Geotherm. Res., 17,
mJ Jth momentum of the distribution. 89–109, doi:10.1016/0377-0273(83)90063-X.
AB time evolution operator for aggregation due to Brow- Costa, A., G. Macedonio, and A. Folch (2006), A three‐dimensional Eulerian
model for transport and deposition of volcanic ashes, Earth Planet. Sci.
nian motion. Lett., 241, 634–647, doi:10.1016/j.epsl.2005.11.019.
AS time evolution operator for aggregation due to fluid Costa, A., F. Dell’Erba, M. Di Vito, R. Isaia, G. Macedonio, G. Orsi, and
shear. T. Pfeiffer (2009), Tephra fallout hazard assessment at the Campi Flegrei
ADS time evolution operator for aggregation due to differ- caldera (Italy), Bull. Volcanol., 71, 259–273, doi:10.1007/s00445-008-
0220-3.
ential sedimentation. Dekkers, P., and S. Friedlander (2002), The self‐preserving size distribu-
Dt computational time step. tion theory: I. Effects of the Knudsen number on aerosol agglomerate
tagg aggregation timescale. growth, J. Colloid Interface Sci., 248, 295–305, doi:10.1006/jcis.
2002.8212.
na total number of aggregates per unit of volume. Dekkers, P., I. Tuinman, J. Marijnissen, S. Friedlander, and B. Scarlett
Dna relative increment of na. (2002), The self‐preserving size distribution theory: II. Comparison with
Dni relative increment of ni. experimental results for Si and Si3N4 aerosols, J. Colloid Interface Sci.,
’ aggregate porosity. 248, 306–314, doi:10.1006/jcis.2002.8213.
Di Stasio, S., A. Konstandopoulos, and M. Kostoglou (2002), Cluster‐
f correction function for aggregate settling velocity cluster aggregation kinetics and primary particle growth of soot nano-
due to permeability. particles in flame by light scattering and numerical simulations, J. Colloid
ye correction function for aggregate settling velocity. Interface Sci., 247, 33–46, doi:10.1006/jcis.2001.8095.
Durant, A., and W. Rose (2009), Sedimentological constraints on
VsA settling velocity of aggregate. hydrometeor‐enhanced particle deposition: 1992 eruptions of Crater Peak,
Vsc settling velocity of a compact particle with the same Alaska, J. Volcanol. Geotherm. Res., 186, 40–59, doi:10.1016/j.jvolgeores.
diameter of the aggregate and the same density of the 2009.02.004.
Durant, A., R. Shaw, W. Rose, Y. Mi, and G. Ernst (2008), Ice nucleation
primary particle. and overseeding of ice in volcanic clouds, J. Geophys. Res., 113,
b sA settling velocity of a compact particle with the same
V D09206, doi:10.1029/2007JD009064.
diameter and density of the aggregate. Durant, A., W. Rose, A. Sarna‐Wojcicki, S. Carey, and A. Volentik (2009),
Hydrometeor‐enhanced tephra sedimentation: Constraints from the
18 May 1980 eruption of Mount St. Helens, J. Geophys. Res., 114,
B03204, doi:10.1029/2008JB005756.
[41] Acknowledgments. This work was supported by the Italian Field, P., A. Heymsfield, and A. Bansemer (2006), A test of ice self‐collection
Department of Civil Protection, Research Project INGV‐SPeeD. A.F. is kernel using aircraft data, J. Atmos. Sci., 63, 651–666, doi:10.1175/
grateful to the Ramón y Cajal scientific program. We thank D. Morgavi JAS3653.1. (Corrigendum, J. Atmos. Sci., 63, 1674, 2006.)
for the useful discussions. We are very grateful to C. Bonadonna and

13 of 14
B09201 COSTA ET AL.: ASH AGGREGATION MODEL IN VOLCANIC PLUMES B09201

Folch, A., C. Cavazzoni, A. Costa, and G. Macedonio (2008), An auto- Rose, W., and A. Durant (2009), Fine ash content of explosive eruptions,
matic procedure to forecast tephra fallout, J. Volcanol. Geotherm. Res., J. Volcanol. Geotherm. Res., 186, 32–39, doi:10.1016/j.jvolgeores.
177, 767–777, doi:10.1016/j.jvolgeores.2008.01.046. 2009.01.010.
Folch, A., A. Costa, and G. Macedonio (2009), FALL3D: A computational Rose, W., C. Riley, L. Henderson, R. McGimsey, G. Bluth, D. Schneider,
model for transport and deposition of volcanic ash, Comput. Geosci., 35, and G. Ernst (2001), Observations of volcanic clouds in their first few
1334–1342, doi:10.1016/j.cageo.2008.08.008. days of atmospheric residence: The 1992 eruptions of Crater Peak,
Folch, A., A. Costa, A. Durant, and G. Macedonio (2010), A model for wet Mount Spurr Volcano, Alaska, J. Geol., 109, 677–694, doi:10.1086/
aggregation of ash particles in volcanic plumes and clouds: 2. Model 323189.
application, J. Geophys. Res., 115, B09202, doi:10.1029/2009JB007176. Rose, W., C. Riley, and S. Dartvelle (2003), Sizes and shapes of 10‐Ma
Frenklach, M. (2002), Method of moments with interpolative closure, distal fall pyroclasts in the Ogallala Group, Nebraska, J. Geol., 111,
Chem. Eng. Sci., 57, 2229–2239, doi:10.1016/S0009-2509(02)00113-6. 115–124, doi:10.1086/344668.
Gilbert, J., and S. Lane (1994), The origin of accretionary lapilli, Bull. Schumacher, R., and H. Schmincke (1995), Models for the origin of accre-
Volcanol., 56, 398–411. tionary lapilli, Bull. Volcanol., 56, 626–639.
Glaze, L., and S. Self (1991), Ashfall dispersal for the 16 September 1986 Smoluchowski, M. (1917), Veruch einer mathematischen Theorie der
eruption of Lascar, Chile, calculated by a turbulent diffusion model, Geo- Koagulationkinetic kolloider Lösungen, Z. Phys. Chem., 92, 128–168.
phys. Res. Lett., 18, 1237–1240, doi:10.1029/91GL01501. Sparks, R., M. Bursik, S. Carey, J. Gilbert, L. Glaze, H. Sigurdsson, and
Gmachowski, L. (1995), Mechanism of shear aggregation, Water Res., 29, A. Woods (1997), Volcanic Plumes, John Wiley, Chichester, U. K.
1815–1820, doi:10.1016/0043-1354(95)00006-7. Spence, R., I. Kelman, P. Baxter, G. Zuccaro, and S. Petrazzuoli (2005),
Gmachowski, L. (2001), Self‐preserving aggregate size distribution, Chem. Residential building and occupant vulnerability to tephra fall, Nat.
Ing. Tech., 73, 87–89, doi:10.1002/1522-2640(200101)73:1/2<87::AID- Hazards Earth Syst. Sci., 5, 477–494, doi:10.5194/nhess-5-477-2005.
CITE87>3.0.CO;2-B. Suzuki, T. (1983), A theoretical model for dispersion of tephra, in
Jacobson, M. (1999), Fundamentals of Atmospheric Modelling, 1st ed., Arc Volcanism: Physics and Tectonics, edited by D. Shimozuru and
Cambridge Univ. Press, New York. I. Yokoyama, pp. 93–113, Terra Sci., Tokyo.
James, M., J. S. Gilbert, and S. J. Lane (2002), Experimental investigation Textor, C., H. Graf, M. Herzog, J. Oberhuber, W. Rose, and G. Ernst
of volcanic particle aggregation in the absence of a liquid phase, J. Geo- (2006a), Volcanic particle aggregation in explosive eruption columns.
phys. Res., 107(B9), 2191, doi:10.1029/2001JB000950. Part I: Parameterization of the microphysics of hydrometeors and ash,
James, M. R., S. J. Lane, and J. S. Gilbert (2003), Density, construction, J. Volcanol. Geotherm. Res., 150, 359–377, doi:10.1016/j.jvolgeores.
and drag coefficient of electrostatic volcanic ash aggregates, J. Geophys. 2005.09.007.
Res., 108(B9), 2435, doi:10.1029/2002JB002011. Textor, C., H. Graf, M. Herzog, J. Oberhuber, W. Rose, and G. Ernst
Jiang, Q., and B. Logan (1991), Fractal dimensions of aggregates deter- (2006b), Volcanic particle aggregation in explosive eruption columns.
mined from steady state size distributions, Environ. Sci. Technol., 25, Part II: Numerical experiments, J. Volcanol. Geotherm. Res., 150,
2031–2038, doi:10.1021/es00024a007. 378–394, doi:10.1016/j.jvolgeores.2005.09.008.
Johnson, C., X. Li, and B. Logan (1996), Settling velocities of fractal aggre- Tian, W., T. Nakayama, J. Huang, and K. Yu (2007), Scaling behaviours in
gates, Environ. Sci. Technol., 30, 1911–1918, doi:10.1021/es950604g. settling process of fractal aggregates in water, Europhys. Lett., 78, 46001,
Jullien, R., and R. Botet (1987), Aggregation and Fractal Aggregates, doi:10.1209/0295-5075/78/46001.
World Sci., Singapore. Tirado‐Miranda, M., A. Schmitt, J. Callejas‐Fernandez, and A. Fernandez‐
Jun, F., S. Xueming, Y. Hongyong, and Z. Xin (2004), Self‐preserving size Barbero (2000), Dynamic scaling and fractal structure of small colloidal
distribution of fire soot fractal coagulation in flaming combustion, J. Fire clusters, Colloids Surfaces A, 162, 67–73, doi:10.1016/S0927-7757(99)
Sci., 22, 53–68, doi:10.1177/0734904104039250. 00216-2.
Lee, C., and T. Kramer (2004), Prediction of three‐dimensional fractal Veitch, G., and A. Woods (2001), Particle aggregation in volcanic eruption
dimensions using the two‐dimensional properties of fractal aggregates, columns, J. Geophys. Res., 106, 26,425–26,441, doi:10.1029/
Adv. Colloid Interface Sci., 112, 49–57, doi:10.1016/j.cis.2004.07.001. 2000JB900343.
Li, X., and B. Logan (2001), Settling velocities of fractal aggregates, Water Vicsek, T. (1992), Fractal Growth Phenomena, 2nd ed., World Sci,
Res., 35, 3373–3380, doi:10.1016/S0043-1354(01)00061-6. Singapore.
Li, X., U. Passow, and B. Logan (1998), Fractal dimensions of small (15– Voivret, C., F. Radjaï, J. Delenne, and M. El Youssoufi (2007), Space‐
200 mm) particles in eastern Pacific coastal waters, Deep Sea Res., Part I, filling properties of polydisperse granular media, Phys. Rev. E, 76,
45, 115–131, doi:10.1016/S0967-0637(97)00058-7. 021301, doi:10.1103/PhysRevE.76.021301.
Liu, L. X., and J. D. Litster (2002), Population balance modelling of gran- Walker, G., G. Andrew, and D. Jones (2006), Effect of process parameters
ulation with a physically based coalescence kernel, Chem. Eng. Sci., 57, on the melt granulation of pharmaceutical powders, Powder Technol.,
2183–2191, doi:10.1016/S0009-2509(02)00110-0. 165, 161–166, doi:10.1016/j.powtec.2006.03.024.
Macedonio, G., A. Costa, and A. Folch (2008), Ash fallout scenarios at Westbrook, C., R. Ball, P. Field, and A. Heymsfield (2004), Theory of
Vesuvius: Numerical simulations and implications for hazard assessment, growth by differential sedimentation, with application to snowflake for-
J. Volcanol. Geotherm. Res., 178, 366–377, doi:10.1016/j.jvolgeores. mation, Phys. Rev. E, 70, 021403, doi:10.1103/PhysRevE.70.021403.
2008.08.014.
Pfeiffer, T., A. Costa, and G. Macedonio (2005), A model for the numerical A. Costa and G. Macedonio, Istituto Nazionale di Geofisica e
simulation of tephra fall deposits, J. Volcanol. Geotherm. Res., 140,
Vulcanologia, Sezione di Napoli, Via Diocleziano 328, I‐80124 Naples,
273–294, doi:10.1016/j.jvolgeores.2004.09.001. Italy. (costa@ov.ingv.it; macedon@ov.ingv.it)
Pomonis, A., R. Spence, and P. Baxter (1999), Risk assessment of resi- A. Folch, Earth Sciences Division Barcelona Supercomputing Center,
dential buildings for an eruption of Furnas Volcano, Sao Miguel, the
Edifici Nexus II, c/ Jordi Girona 29, 08034 Barcelona, Spain. (arnau.
Azores, J. Volcanol. Geotherm. Res., 92, 107–131, doi:10.1016/ folch@bsc.es)
S0377-0273(99)00071-2.

14 of 14

S-ar putea să vă placă și