Probabilistic modeling of tephra dispersion: hazard assessment of a multi-phase eruption at Tarawera Volcano, New Zealand C. Bonadonna 1 , C.B. Connor 2 , B.F. Houghton 1 , L. Connor 2 , M. Byrne 3 , A. Laing 3 , T.K. Hincks 4 1 Department of Geology and Geophysics, University of Hawaii, Honolulu, HI 96822, USA 2 Department of Geology, University of South Florida, Tampa, FL 33620, USA 3 Department of Geography, University of South Florida, Tampa, FL 33620, USA 4 Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, UK
65
Embed
Probabilistic modeling of tephra dispersion: hazard ...pages.mtu.edu/~raman/papers2/BonadonnaJGR.pdfProbabilistic modeling of tephra dispersion: hazard assessment of a ... (Fig. 2).
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Probabilistic modeling of tephra dispersion: hazard assessment of a multi-phase eruption at
Tarawera Volcano, New Zealand
C. Bonadonna1, C.B. Connor2, B.F. Houghton1,L. Connor2, M. Byrne3, A. Laing3, T.K. Hincks4
1Department of Geology and Geophysics, University of Hawaii, Honolulu, HI 96822, USA 2Department of Geology, University of South Florida, Tampa, FL 33620, USA
3Department of Geography, University of South Florida, Tampa, FL 33620, USA 4Department of Earth Sciences, University of Bristol, Bristol, BS8 1RJ, UK
2
Abstract
The Tarawera Volcanic Complex (New Zealand) comprises 11 rhyolite domes formed
during five major eruptions between 17000 BC and AD 1886, the first four of which were
predominantly rhyolitic. The only historical event (AD 1886) erupted about 2 km3 of
purely basaltic tephra fall killing about 150 people. The AD 1305 Kaharoa eruption is the
youngest rhyolitic event and erupted a total tephra-fall volume more than two times
larger than the AD 1886 eruption. We used data from the AD 1305 Kaharoa eruption to
assess the tephra-fall hazard from a future episode at Tarawera. This eruption consisted of
a complex sequence of eruptive events with eleven discrete Plinian episodes,
characterized by VEI 4. We developed an advection-diffusion model (TEPHRA) that
includes grainsize-dependent diffusion law and particle density, a stratified atmosphere,
the horizontal diffusion time of particles within the eruptive plume and settling velocities
that account for Reynold’s Number variations along the particle trajectory. Selected
eruption parameters are sampled stochastically, possible outcomes are analyzed
probabilistically and simulations are run in parallel on multiple processors. Given the fast
computational times, TEPHRA is also characterized by a robust algorithm and high
resolution and reliability of resulting outputs (i.e. hazard maps). Therefore, TEPHRA is
an example of a class of numerical models that take advantage of new computational
tools to forecast hazards as conditional probabilities far in advance of future eruptions.
Three different scenarios were investigated on a probabilistic basis for a comprehensive
tephra-fall hazard assessment: Upper Limit Scenario, Eruption-Range Scenario and
Multiple-Eruption Scenario.
3
1. Introduction Numerical models are increasingly important in geological hazards and risk
assessments [Barberi et al., 1990; Bonadonna et al., 2002a; Connor et al., 2001; Glaze
and Self, 1991; Hill et al., 1998; Iverson et al., 1998; Wadge et al., 1998]. These models
are used to quantify assessments that are otherwise based on qualitative, sometimes
disparate geological observations [Newhall and Hoblitt, 2002]. Numerical simulations (i)
augment direct observations; (ii) characterize better the variation and uncertainty in
geologic processes that often occur on long time scales, or infrequently, compared with
the time scales of human experience; (iii) allow volcanologists to explore a much wider
range of geological processes than is possible to observe directly. Therefore, evaluating
the range of possible outcomes of geologic processes, such as earthquakes, volcanic
eruptions, and landslides, is best achieved using probabilistic techniques that propagate
uncertainty through the analysis using stochastic simulations.
This is certainly true in volcanology, a field in which hazard assessments must
strive to bound the range of possible consequences of volcanic activity, drawing from the
geologic record, analogy, and understanding of the physics of volcanic processes.
Historically, volcanology has advanced through description of volcanic processes and
Nevado del Ruiz, [Chung, 1991]). While extremely important, this approach is no longer
sufficient for mitigation of volcanic risks and numerical simulations should be used to
complement direct observations. However, such an approach is computationally
expensive, because numerical models of geologic processes are generally complex, and
because a large number of simulations is required to accurately simulate the range of
behaviors of natural phenomena, like volcanic eruptions. Nevertheless, recent advances
in parallel computing, such as the development of the Message Passing Interface (MPI)
and the advent of inexpensive computer clusters [Sterling et al., 1999], render this
approach to geological hazard assessment tractable.
As a practical example, we describe the probabilistic assessment of hazards
related to dispersion and accumulation of volcanic tephra fall, and, in particular, we
4
present the tephra-fall hazard assessment of Tarawera Volcano (New Zealand) based on
its most recent rhyolitic Plinian eruption (AD 1305 Kaharoa eruption, [Sahetapy-Engel,
2002]). Tarawera Volcano is located in the North Island of New Zealand and has been
one of New Zealand’s most destructive volcanoes in recent times. Famous for its large
eruption in 1886, Tarawera Volcano buried seven villages, killing about 150 people
[Keam, 1988]. The AD 1305 Kaharoa eruption produced a total tephra-fall volume nearly
three times larger than the AD 1886 eruption, covering a wider area northwest and
southeast of the volcano, and therefore we consider it as the appropriate scenario for our
hazard assessment of tephra fall from Tarawera Volcano. Based on the frequency of past
eruptions from this volcano complex, the annual probability of an eruption from
Tarawera with volume exceeding 0.5 km3 is approximately 10-3/yr [Stirling and Wilson,
2002], certainly a sufficiently high probability to require assessment of eruption
consequences [Woo, 2000].
This tephra-fall hazard assessment is achieved through implementation of an
advection-diffusion model (TEPHRA) derived from the integration of several modeling
approaches and theories [Armienti et al., 1988; Bonadonna et al., 1998; Bonadonna et al.,
2002a; Bonadonna and Phillips, 2003; Bursik et al., 1992a; Connor et al., 2001;
Macedonio et al., 1988; Suzuki, 1983]. TEPHRA is written for parallel computation on a
Beowulf cluster, a networked set of personal computers. As such, TEPHRA is an
example of a class of numerical models that take advantage of new computational tools to
forecast hazards as conditional probabilities far in advance of future eruptions. That is,
given that a scenario of volcanic activity takes place, what is the expected range of
tephra-fall thicknesses over a region of interest? What drives uncertainty in hazard
assessment? What eruptive conditions result in hazardous tephra-fall accumulations? Our
goal is to illustrate how numerical models, like TEPHRA, can help resolve such
questions and provide a basis for improved hazard assessment.
2. Geological setting The Tarawera Volcanic Complex is a dome complex within the Okataina
Volcanic Centre, one of the five major calderas within the Taupo Volcanic Zone, North
5
Island, New Zealand (Fig. 1). Tarawera is made of 11 rhyolite domes and a combination
of tephra-fall and flow deposits that formed during five major eruptions [Cole, 1970]: (i)
AD 1886; (ii) Kaharoa, AD 1305; (iii) Waiohau, 11000 BP; (iv) Rerewhakaaitu, 15000
BP; (v) eruption associated with the Okareka Ash, 17000 BP. The AD 1886 subplinian
eruption was basaltic. The most recent rhyolitic eruption occurred about 700 years ago
(Kaharoa eruption), and has been intensely studied in the last 5 years [Nairn, 1989; Nairn
et al., 2001; Sahetapy-Engel, 2002].
2.1 AD 1305 Kaharoa eruption
The AD 1305 Kaharoa eruption represents the most recent rhyolitic event in the
whole Taupo Volcano Zone. It consisted of a sequence of 11 Plinian eruptive episodes
with Volcano Explosivity Index (VEI) 4, column heights between about 16 and 26 km,
and volumes between about 0.2 and 1 km3 (minimum total volume is 4.6 km3, 2.2 Dense
Rock Equivalent, DRE) [Sahetapy-Engel, 2002]. Sahetapy-Engel [2002] divides the
Kaharoa tephra-fall stratigraphy in two main deposit lobes: one to the southeast and one
to the north-northwest of the volcano (Fig. 2). Detailed study of these lobes also revealed
pronounced layering of fall deposits, which were classified as Units A to L
corresponding to the 11 Plinian eruptive episodes [Sahetapy-Engel, 2002].
The Kaharoa eruption was characterized by multiple vents, but the exact location
of eruptive vents for each Unit is still uncertain. By extrapolations of the axes of dispersal
and clast size distribution Sahetapy-Engel [2002] concludes that: Units A-G were
dispersed to the southeast and were probably erupted from Crater Dome; Unit H was
dispersed to both southeast and northwest and was probably erupted from the Ruawahia
Dome; Units I-K were dispersed to the north-northwest and were also probably erupted
from the Ruawahia Dome and/or Wahanga Dome (Fig. 1).
Unit F is the one characterized by the largest volume (minimum volume of 1 km3)
and highest plume (26.4 km), whereas Units A and G have the smallest volume
(minimum volume of 0.16 km3) and lowest plumes (16.3 km). Unit K is the most
voluminous phase of the late stage of the eruption (0.79 km3) [Sahetapy-Engel, 2002].
6
The average density of analyzed pumice and lithic fragments is 1000 and 2350 kg
m-3 respectively, whereas the average bulk deposit density is 1000 kg m-3 [Sahetapy-
Engel, 2002]. High pumice density values are due to the high crystal content and low
vesicularity. An average maximum wind speed of 20-30 m s-1 and a duration of 2-6 hours
for individual eruptive episodes (with the only exception being 19 hours for episode K)
were calculated by using the method of Carey and Sparks [1986] and Sparks [1986]
[Sahetapy-Engel, 2002].
Studying the wind-field patterns throughout a whole year, Sahetapy-Engel [2002]
shows how it is likely that the whole Kaharoa eruption had a maximum duration of the
first explosive phase (episodes A-G) of 13 days. Estimates of the duration of the total
Kaharoa eruption vary between weeks to a few months, depending on the values assigned
to breaks in the eruption and the duration of dome growth and disruption [Sahetapy-
Engel, 2002].
3. TEPHRA TEPHRA consists of three main parts: (i) a physical model that describes
diffusion, transport and sedimentation of volcanic particles [Armienti et al., 1988;
Bonadonna et al., 1998; Bonadonna et al., 2002a; Bursik et al., 1992a; Connor et al.,
2001; Suzuki, 1983]; (ii) a probabilistic approach used to identify a range of input
parameters for the physical model (i.e. column height; eruption duration; mass
distribution parameter; clast exit velocity; grainsize parameters) and to forecast a range of
possible outcomes (i.e. hazard curves and probability maps); (iii) a computational
approach that uses parallel processing methods to speed up computation and make fully
probabilistic approaches practical.
3.1 Physical model
TEPHRA is written in C with MPI commands. Particle diffusion, advection and
sedimentation are computed solving a mass-conservation equation [Armienti et al., 1988;
Suzuki, 1983]. Particles of size fraction j are released from a point source i along a
7
volcanic plume. Each particle size fraction, j, and point source i is processed
independently due to the linearity of the equation. That is, the total mass oM (kg) of the
eruption is:
max
min
,0
Ho o
i ji j
M M (1)
where ,oi jM (kg) is the total mass fraction of particles with size j that fall from a point
source i at a height zi, H is the total height of the volcanic plume and min and max
indicate the maximum and minimum particle diameter respectively (with = -log2d,
where d is the particle diameter in mm). The fraction of ,oi jM (kg) that accumulates on the
ground at a certain point with coordinates (x,y) is ),(, yxm ji (kg m-2), where:
0, , ,( , ) ( , )i j i j i jm x y M f x y (2)
where ),(, yxf ji (m-2) is a function, described in detail in the following, that uses an
advection-diffusion equation to estimate the fraction of mass of a given particle size and
release height to fall around the point with coordinates (x,y). Therefore, the total mass M
accumulated per unit area (kg m-2) at a certain point on the ground (x,y) is:
max
min
,0
( , ) ( , )H
i ji
M x y m x y (3)
which is the quantity of greatest interest in forecasting volcanic hazards related to tephra
fall. Thus, the problem reduces to understanding the function, ),(, yxf ji , which controls
the horizontal dispersion of particles, and jioM , , the source term.
8
All the particles are released instantaneously [Bonadonna et al., 2002a; Connor et
al., 2001; Suzuki, 1983] and are assumed to be spherical [Bonadonna et al., 2002a] with a
settling velocity that varies according to the particle Reynolds number [Bonadonna et al.,
1998]. The atmosphere is divided into horizontal layers characterized by a uniform
horizontal wind velocity and direction specific for each layer [Bonadonna et al., 2002a].
Each point source i is located in a horizontal layer, and particles released from that point
source are initially transported by the wind specific for that layer, until they fall into a
lower layer, where they are affected by a different wind direction and velocity. This
process continues until the particles reach the ground.
For emission from an instantaneous point source, the analytical solution of the
mass-conservation equation is a Gaussian distribution of concentration in both the x and
y directions [Bonadonna et al., 2002a; Connor et al., 2001]. Particles spread
horizontally due to the combined effects of turbulent eddy diffusion and gravity
spreading of the plume, and are transported by the wind for the time jt spent in each
layer. jt is a function of the settling velocity jv of the particles and the layer thickness
z (with j
j vzt ). After the time jt , the centre of the Gaussian distribution is shifted
in the yx plane by a distance jxj twx and jyj twy on the axes x and y
respectively, where xw and yw are the horizontal components of the wind speed in that
layer. Particles falling from a point source i located at iii zyx ,, reach the ground at the
time jit , , where:
,i
i j jlayers j
zt tv
(4)
Therefore, the analytical solution of the mass-conservation equation can be
written as:
9
2,
2
,
2,
2,
, exp1),(ji
jiji
jiji
yyxxyxf (5)
where jix , and jiy , are the coordinates of the center of the bivariate Gaussian
distribution (layers
jiji xxx , ,layers
jiji yyy , ) and 2,i j is the variance of the
Gaussian distribution, which is controlled by atmospheric diffusion and horizontal
spreading of the plume [Suzuki, 1983].
3.1.1 Atmospheric diffusion
The parameter 2,i j controls diffusion of particles in the atmosphere. Effectively,
the use of 2,i j in eq. (1) lumps complex plume and atmospheric processes into a single
parameter. This greatly simplifies the model, making it much easier to implement but
also ignores processes that can affect tephra-fall dispersion. For example, the diffusion
coefficient is likely scale dependent and varies with barometric pressure in the
atmosphere [Hanna et al., 1982]. Such factors are not considered in the model.
Atmospheric turbulence is a second order effect for coarse particles, and several
models for tephra-fall dispersal are based on the assumption that the atmospheric
turbulence is negligible [Bonadonna and Phillips, 2003; Bursik et al., 1992b; Sparks et
al., 1992]. However, if the fall time of particles is large, for example for ash-sized
particles, atmospheric turbulence may not be negligible [Bursik et al., 1992a; Suzuki,
1983]. In the case of small particle-fall time, ,i jt , the diffusion is linear (Fick's law), and
the variance 2,i j is [Suzuki, 1983]:
2 'i,j ,4 i j iK t t (6)
10
where K (m2 s-1) is a constant diffusion coefficient, ,i jt (s) is the total particle fall time
(eq. (4)), and 'it (s) is the horizontal diffusion time in the volcanic plume. The horizontal
diffusion coefficient, K , is considered isotropic (K=Kx=Ky) [Armienti et al., 1988;
Bonadonna et al., 2002a; Connor et al., 2001; Suzuki, 1983]. The vertical diffusion
coefficient is small above 500 m of altitude [Pasquill, 1974], and therefore is assumed to
be negligible. The horizontal diffusion time, 'it , accounts for the change in width of the
plume as a function of height. Change in width of the ascending turbulent plume is
complex [Ernst et al., 1996; Woods, 1995] and also includes the gravitational spreading
should the plume reach neutral buoyancy [Sparks, 1986]. Such a change in plume width
simply adds to the dispersion of tephra fall, and so can be expressed as 'it [Suzuki, 1983].
Here, we approximate the radius, ir , of the spreading plume at a given height, iz , with
the relation developed by [Bonadonna and Phillips, 2003] and based on the combination
of numerical studies [Morton et al., 1956] and observations of plume expansion [Sparks
and Wilson, 1982]: 0.34i ir z . Thus, taking 3i pr , where p is the standard deviation
of the Gaussian distribution of the mass in the ascending plume [Sparks et al., 1997;
Suzuki, 1983], we have:
2' 2 50.2i it z (7)
We also assume that 'it does not depend on grainsize. However, Fig. 3a and Fig. 3b show
how 'it significantly affects the total fall time of coarse particle, i.e. '
,( )i j it t , because for
coarse particles jii tt ,' .
When the particle fall is of a time scale of hours, the scale of turbulent structures
that carry particles (and then K ) increases with time. As an example, particles with
diameter <1 mm falling from a 30 km-high plume will have an average time of fall >1
hour (based on their particle settling velocity). In this case the variance 2,i j can be
empirically determined as [Suzuki, 1983]:
11
2.52 ', ,
85i j i j iC t t (8)
where C is a constant empirically determined ( C =0.04 m2 s-1; [Suzuki, 1983]).
Therefore, once particles leave the bottom of the turbulent current, they must
experience different types of turbulent diffusion depending on their size. The linear
diffusion described by eq. (6) strongly depends on the choice of the diffusion coefficient,
whereas in the power-law diffusion described by eq. (8) the diffusion coefficient is
calculated on the basis of the apparent eddy diffusivity and the time of particle fall
[Suzuki, 1983]. Some advection-diffusion models consider only one diffusivity law
[Bonadonna et al., 2002a; Connor et al., 2001]. TEPHRA accounts for two types of
diffusivity law according to particle size: linear diffusion for ash-sized particles (Fick's
law; eq. (6)) and power-law diffusion for coarse particles (eq. (8)). If the volcanic plume
is sufficiently high, specific particles will experience a shift in diffusion law during fall
due to the decrease in fall time (e.g. particles with diameter = 0.25 mm; Fig. 3c). The
transition from one diffusion law to another (eq. (6) to eq. (8)) is not well defined based
on theory but can be determined empirically. Fig. 3c shows the power-law dependence of 2,i j with time, which makes the total diffusion more significant for fine particles.
3.1.2 Settling velocity
The settling velocity jv of falling particles of size j is obtained by the balance
between gravity and air drag:
43
pj
D a
gdv
C (9)
where g is the gravitational acceleration (m s-2), d is the particle diameter (m), p is
the particle density (kg m-3), DC is the drag coefficient and a is the air density (kg m-3).
12
The drag coefficient is a function of the particle shape and the Reynolds number,
Re a jd v, where is the air viscosity [kg m-1 s-1]. For non-spherical particles the
determination of DC is very complicated [Kunii and Levenspiel, 1969], therefore models
for tephra-fall dispersal are typically based on the assumptions of spherical particles for
which DC can be determined using simple empirical expressions [Bonadonna et al.,
1998; Bursik et al., 1992a; Bursik et al., 1992b; Koyaguchi and Ohno, 2001a; Sparks et
al., 1992]. Studies have shown that particle-settling velocities strongly depend on particle
shape [e.g. [Armienti et al., 1988; Macedonio et al., 1988; Wilson and Huang, 1979], but
given the current uncertainties in the determination of particle shape from field data
[Chhabra et al., 1999], we have decided to adopt the assumption of spherical particles
until new techniques are made available, and therefore to use the analytic expressions for
jv from Kunii and Levenspiel [1969] and modified by Bonadonna and Phillips [2003].
3.1.3 Mass distribution
The source term, ,oi jM , represents the distribution of mass as a function of particle
size and height in the eruption column. Several methods have been used to describe
particle distribution in the ascending volcanic plume [Bonadonna et al., 2002a; Sparks
and Walker, 1977; Suzuki, 1983; Woods, 1988]. Here we modify the method proposed by
Suzuki [1983] in which mass is assumed to have an exponential distribution in the plume
as a function of particle settling velocity and initial bulk velocity at the vent. The
probability density function for particle distribution as a function of height in the plume,
for a given grainsize, is:
( )( )( )1 (1 )
i
o
Y zo i
Z i Yp o
w Y z ep zv H Y e
(10)
where is a dimensionless parameter that controls the shape of the distribution function
(larger values of place more mass, proportionally, at the top of the volcanic plume;
Fig. 4), ow (m s-1) is the initial velocity of pyroclasts at the vent, pv (m s-1) is the particle
13
settling velocity as empirically calculated by Suzuki [1983] to fit data from Wilson and
Huang [1979], and:
( )( ) ii
p
w zY zv
oo
p
wYv
(11)
( ) 1 ii o
zw z wH
Large values of , say =1.0, skews )( iZ zp such that most particles are released from
high in the volcanic plume (Fig. 4a). This corresponds to a strong plume of the type that
occurs in Plinian eruptions. Small values of (e.g. = 0.01) result in a much less
skewed distribution of )( iZ zp . For coarse particles with comparatively high settling
velocities, )( iZ zp decreases nearly linearly with height in the volcanic plume (Fig. 4b).
This behavior occurs in weak plumes [Sparks et al., 1997]. Thus, should be adjusted
based on the type of volcanic eruptions being simulated.
The source term, ,oi jM , is then calculated by assuming an eruption grainsize
distribution, )(f [Suzuki, 1983]:
, ( ) ( )o oi j z iM p z f M (12)
3.1.4 Total erupted mass
Given a plume height H (m), the total erupted mass oM (kg) is derived from an
empirical power-law equation [Carey and Sigurdsson, 1989]:
4
1670o
depHM (13)
14
where dep (kg m-3) is the density of the tephra-fall deposit and (s) is the duration of
the sustained phase of the eruption.
3.2 Probabilistic determination of inputs and outputs
3.2.1 Inputs
Plume height: either an individual plume height H or a range of plume heights can be
input in TEPHRA according to the type of eruptive scenario investigated and the type of
output result desired: (i) one input value of H , together with one wind profile, is used to
compute isomass maps; one input value of H is also used to compute hazard curves and
probability maps for the worst-case eruptive episode (i.e. typically the highest plume
observed and/or considered possible, i.e. Upper Limit Scenario) that are based on the
variability of wind profiles; (ii) a range of input values of H is randomly sampled for the
computation of hazard curves and probability maps that account for the variability of
eruptive episodes and wind profiles (i.e. Eruption Range Scenario); (iii) a whole set of
input values of H is used for the computation of cumulative probability maps (i.e.
probability maps computed for a scenario of long-lasting activity (i.e. multiple eruptions)
that account for the variability of wind profiles, i.e. Multiple Eruption Scenario). In case
(ii) any probability function of H can be sampled. We have decided to randomly sample
a uniform set of values that range between min maxLog H Log H where minH
and maxH are, respectively, the minimum and the maximum plume height observed
and/or considered possible. We have chosen a logarithmic function of H to reflect a
higher frequency of low plumes. As an example the distribution of plume height
randomly sampled for a Kaharoa-type eruption is shown (i.e. H =14-26 km; Fig. 5a). The
minimum height represents the boundary between subplinian and Plinian eruptions
[Sparks et al., 1992] in agreement with the Kaharoa-type events, whereas the maximum
height is from field data [Sahetapy-Engel, 2002].
Duration of eruptive episodes: together with the plume height H , the duration of
individual eruptive episodes is used for the determination of the total erupted mass (eq.
15
(13)). TEPHRA randomly samples the duration amongst a given range of values
observed and/or considered possible. As an example the distribution of the eruptive-
episode duration randomly sampled for a Kaharoa-type eruption is shown (i.e. 2-6 hrs;
Fig. 5b).
Total erupted mass: the distribution of plume-height values described above associated
with the randomly sampled distribution of eruptive-episode duration results in a log-
normal distribution of the total erupted mass derived using eq. (13) (Fig. 5c).
Plume mass distribution ( ): the mass distribution along the eruptive plume is controlled
by a factor [Connor et al., 2001; Suzuki, 1983]. We have decided to link to the
column height, as larger plumes will be characterized by a larger accumulation of mass at
the top. Therefore, for powerful plumes can be calculated as:
minmin (1 min)
max minH H
H H(14)
where min is the minimum value of considered (Fig. 6a). The maximum value of
possible is therefore 1 (i.e. maximum of mass accumulation at the top of the plume).
Total grainsize distribution: a grainsize distribution can be defined by expressing the
corresponding minimum and the maximum particle diameter, the Median Diameter
(Md ), the Graphic Standard Deviation ( ) and the Graphic Skewness (SkG) [Inman,
1952]. However, the total grainsize distribution of pyroclastic deposits is typically
extremely difficult to determine mainly due to the methodological problems related to the
integration of grainsize analysis of single samples and to the scarcity of data points
(because of poor outcrop exposure, deposit erosion and/or tephra-fall dispersal in the
sea). As a result, only a few total grainsize distributions are available [Carey and
Sigurdsson, 1982; Hildreth and Drake, 1992; Sparks et al., 1981; Walker, 1980; Walker,
1981]. Given these uncertainties, we have applied a probabilistic approach also for the
16
determination of the total grainsize distribution stochastically sampling Md between
values observed and/or considered possible. As an example the distribution of Md
randomly sampled for a Kaharoa-type eruption is shown (i.e. Md = -0.8 and 4 ; Fig.
6b). This is based on data from comparable Plinian eruptions: Taupo, Waimihia and
A probabilistic approach is also used to forecast a range of possible outcomes (i.e.
hazard curves and probability maps), so that the probability of exceeding certain
hazardous tephra-fall accumulations can be investigated for different eruptive scenarios
and a wide area around the eruptive vent (Fig. 13-18). The calculation of tephra-fall
accumulation on a grid is also inherently parallel because the exact same computations
are performed many times for different grid points.
TEPHRA describes the particle transport at discrete atmospheric level (e.g. 1 km)
accounting for settling-velocity variations (based on particle Reynold’s Number
variation) and wind variations (i.e. direction and velocity). The accuracy of the resulting
hazard assessment is strictly related to the accuracy and number of wind profiles used and
weather fluctuations analyzed. The detailed gridded zonal and meridional wind fields
downloaded from the NCEP Reanalysis project allowed a full hazard assessment
32
including specific seasonal assessments (Fig. 17) and assessments for particular climate
conditions (e.g. ENSO phenomenon; Fig. 18).
A good statistical study of wind profiles also help constrain the occurrence time
of a given eruption. As an example, Fig. 11 shows that winds below 25 km above sea
level around the Tarawera Volcanic Complex blow between N and S with main direction
to the E, and winds above 25 km blow between W and E with main direction to the N.
Fig. 12 also shows that winds are more likely to blow to the W and N during September
through March. Given that the 5 Kaharoa units dispersed to the N and NW were produced
consecutively and by plumes smaller than 25 km (i.e. Units H, I, J, K and L; plume height
between 16-24 km [Sahetapy-Engel, 2002]), it is likely that Units H to L were produced
during the same austral spring-summer. However, Units A to G were dispersed to the SE
and therefore it is more difficult to estimate the corresponding occurrence time, given that
winds can blow to the SE during the whole year (Fig. 12). Our probabilistic analysis for
the ENSO phenomenon also indicates that the unusual wind conditions that produced the
Kaharoa tephra-fall dispersal cannot be explained as en effect of El Niño or La Niña
climate fluctuations (Fig. 18). Therefore, based on purely tephra-fall dispersal
considerations, we can conclude the whole Kaharoa eruption could have occurred during
an austral spring-summer or at least 5 consecutive units are very likely to have occurred
during the same austral spring-summer. However, magma-chamber dynamics and
volcanic edifice geometry should also be considered in order to assess magma-chamber
recharging times and the times required to establish certain eruptive conditions [Melnik,
2000; Pinel and Jaupart, 2003].
6.1 Model caveats
TEPHRA represents a great implementation of the existing advection-diffusion
models of tephra-fall dispersal. However, there are still some parameters and processes
that need to be investigated and studied in more details. First of all, the mass distribution
within the eruptive plume is controlled by the factor and by the determination of the
particle-settling velocity within the plume and the plume vertical velocity. The factor
allows to switch between sedimentation from strong plumes and sedimentation from
33
weak plumes, but is an empirical parameter not supported by a robust theory. Moreover,
the determination of particle-settling velocity within the plume and plume vertical
velocity is very complex and TEPHRA still uses empirical parameters based on the
theory from Suzuki [1983]. A more thorough model is needed to describe the plume
dynamics.
A further implementation of TEPHRA could be describing the effects of
aggregation processes and particle shape on tephra-fall dispersal. Aggregation processes
were not accounted for in our assessment because no particle-aggregation data are
available for the Kaharoa eruption. A reliable parameterization that can describe
aggregation processes during tephra fall even when direct observations are not possible
would help describe tephra-fall features also from those volcanoes that do not erupt very
frequently and therefore do not provide detailed information of their eruptive processes.
A simple parameterization of the effect of particle shape on particle-settling velocity
would also significantly improve the description of tephra-fall dispersal. Unfortunately,
modeling settling velocities without the assumption of spherical particles is still very
complex [Chhabra et al., 1999]. Nonetheless, the stochastic sampling of grainsize
parameters used in our assessment can at least provide a complete analysis regardless the
lack of reliable observations due to the difficulties of studying an old tephra-fall deposit.
Advection-diffusion models are typically characterized by particle release at
time=0, and therefore do not account for wind-profile variations with time that can
significantly affect long-lasting eruptions. Another important implementation of
advection-diffusion models in the frame of forecasting tephra-fall dispersal for hazard
assessment would be including the time factor for the simulation of tephra fall.
Finally, reliable datasets from powerful eruptions are needed to calibrate
advection-diffusion models more accurately. Our calibration was based on one very good
dataset from a weak subplinian plume. In order to have more reliable models for tephra-
fall dispersal that can provide reliable hazard assessments, we need more datasets
complete with direct observations of tephra-fall processes and accurate data processing.
34
7. Conclusions
A new advection-diffusion model, TEPHRA, was developed from the
combination of several theories and modeling approaches to provide an efficient and
reliable tool for hazard assessment of tephra fall from weak and strong plumes. After a
careful analysis of the model, we can conclude that:
1. the main implementations of TEPHRA are represented by: (i)
parallelization of the advection-diffusion code, (ii) fully probabilistic assessment of
input and output parameters and (iii) more robust algorithm.
2. parallel modeling is the ideal computational approach for hazard
assessment as, given the short computing times, it increases the hazard map resolution
and reliability because calculations can be done on more points and because the
physical models can be based on a more robust algorithm;
3. the short computing times that characterize parallel modeling also allow a
fully probabilistic assessment based on (i) a stochastic sampling of input parameters
and (ii) a probabilistic analysis of possible outcomes;
4. the main implementations of the advection-diffusion algorithm are: (i)
grainsize dependent diffusion law, (ii) particle-settling velocities that account for
Reynold’s Number variations along the particle trajectory and (iii) use of the mass-
distribution factor to simulate sedimentation from strong as well as weak plumes
(i.e. using large and small values of respectively);
5. output results are sensitive to the choice of diffusion coefficient and
values and less sensitive to the choice of fall-time cut off.
A combination of observations from the AD 1305 Kaharoa eruption and model
simulations enabled us to evaluate probabilistically the accumulations and effects in
different areas of the North Island (NZ) of tephra fall produced by Tarawera Volcano and
based on the assumption of three different eruptive scenarios: (i) Upper Limit Scenario (1
eruptive episode happening with a 26-km-high plume), (ii) Eruption Range Scenario (1
eruptive episode happening with a plume in the range 14-26 km) and (iii) Multiple
35
Eruption Scenario (10 eruptive episodes with plumes in the range 14-26 km). On the
basis of our tephra-fall hazard assessment we can also conclude that:
6. due to the prevailing winds below 25 km above sea level blowing between
N and S with main direction to the E, the areas W and S of the Tarawera Volcanic
Complex are likely to receive little tephra fall from a Kaharoa-type eruption.
Therefore key cities such as Tauranga, Hamilton and Auckland are relatively safe
from hazardous accumulation of tephra fall;
7. the most affected localities are key towns NE of Tarawera that are likely
to experience damage to vegetation in the case of ULS, ERS and MES and partial
collapse of building in the case of ULS and MES;
8. detailed wind analysis shows that the dispersal of tephra fall from
Tarawera Volcano is not significantly affected by El Niño or La Niña fluctuations,
whereas is slightly affected by seasonal variations, with the area immediately west of
the volcano being more likely to receive tephra fall during the austral spring-summer;
9. based on purely tephra-fall dispersal considerations, the whole Kaharoa
eruption might have occurred during an austral spring-summer or at least 5
consecutive units are very likely to have occurred during the same austral spring-
summer (i.e. Units H to L).
Acknowledgments
Many thanks to Steve Sahetapy-Engel for his continuous support and sharing of his
experience and knowledge of the AD 1305 Kaharoa eruption. NCEP Reanalysis data
were provided by the NOAA-CIRES Climate Diagnostics Center, Boulder, Colorado,
USA, from their Web site at http://www.cdc.noaa.gov/. This research was funded by a
subcontract from the Foundation for Research, Science and Technology via Professor
J.W. Cole and the University of Canterbury (New Zealand).
36
References
Armienti, P., G. Macedonio, and M.T. Pareschi, A numerical-model for simulation of tephra transport and deposition - applications to May 18, 1980, Mount-St-Helens eruption, Journal of Geophysical Research, 93 (B6), 6463-6476, 1988.
Barberi, F., G. Macedonio, M.T. Pareschi, and R. Santacroce, Mapping the tephra fallout risk: an example from Vesuvius, Italy, Nature, 344, 142-144, 1990.
Blong, R., The Rabaul Eruption, 1994, Australian Geographer, 25 (2), 186-190, 1994. Blong, R.J., Volcanic Hazards. A sourcebook on the effects of eruptions, 424 pp.,
Academic Press, Sidney, 1984. Bonadonna, C., Models of tephra dispersal, Ph.D. thesis, University of Bristol, Bristol,
2001.Bonadonna, C., G.G.J. Ernst, and R.S.J. Sparks, Thickness variations and volume
estimates of tephra fall deposits: the importance of particle Reynolds number, Journal of Volcanology and Geothermal Research, 81 (3-4), 173-187, 1998.
Bonadonna, C., G. Macedonio, and R.S.J. Sparks, Numerical modelling of tephra fallout associated with dome collapses and Vulcanian explosions: application to hazard assessment on Montserrat, in The eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999, edited by T.H. Druitt, and B.P. Kokelaar, pp. 517-537, Geological Society, London, Memoir, 2002a.
Bonadonna, C., and J.C. Phillips, Sedimentation from strong volcanic plumes, Journal of Geophysical Research, 108 (B7), 2340-2368, 2003.
Bursik, M.I., S.N. Carey, and R.S.J. Sparks, A gravity current model for the May 18, 1980 Mount-St-Helens plume, Geophysical Research Letters, 19 (16), 1663-1666, 1992a.
Bursik, M.I., R.S.J. Sparks, J.S. Gilbert, and S.N. Carey, Sedimentation of tephra by volcanic plumes: I. Theory and its comparison with a study of the Fogo A plinian deposit, Sao Miguel (Azores), Bulletin of Volcanology, 54, 329-344, 1992b.
Carey, S., and H. Sigurdsson, The intensity of Plinian eruptions, Bulletin of Volcanology,51, 28-40, 1989.
Carey, S.N., and H. Sigurdsson, Influence of particle aggregation on deposition of distal tephra from the May 18, 1980, eruption of Mount St-Helens volcano, Journal of Geophysical Research, 87 (B8), 7061-7072, 1982.
Carey, S.N., and R.S.J. Sparks, Quantitative models of the fallout and dispersal of tephra from volcanic eruption columns, Bulletin of Volcanology, 48, 109-125, 1986.
Chhabra, R.P., L. Agarwal, and N.K. Sinha, Drag on non-spherical particles: an evaluation of available methods, Powder Technology, 101 (3), 288-295, 1999.
Chung, M.R., The eruption of Nevado del Ruiz Volcano, Colombia, South America (November 13,1985), in Natural Disaster Studies, edited by N.R. Council, pp. 109, National Academy Press, Washington, D.C., 1991.
Cole, J.W., Structure and eruptive history of Tarawera Volcanic Complex, New Zealand Journal of Geology and Geophysics, 13, 879-902, 1970.
Coltelli, M., M. Pompilio, P. Del Carlo, S. Calvari, S. Pannucci, and V. Scribano, Etna. Eruptive activity, in Data related to eruptive activity, unrest phenomena and other observations on the Italian active volcanoes – Geophysical monitoring of
37
the Italian active volcanoes 1993-1995, edited by P. Gasparini, pp. 141-148, Acta Vulcanologica, 1998.
Connor, C.B., B.E. Hill, B. Winfrey, N.M. Franklin, and P.C. La Femina, Estimation of volcanic hazards from tephra fallout, Natural Hazards Review, 2 (1), 33-42, 2001.
Di Vito, M.A., R. Isaia, G. Orsi, J. Southon, M. D'Antonio, S. de Vita, L. Pappalardo, and M. Piochi, Volcanism and deformation since 12,000 years at the Campi Flegrei caldera (Italy), Journal of Volcanology and Geothermal Research, 91 (2-4), 221-246, 1999.
Ernst, G.G.J., R.S.J. Sparks, S.N. Carey, and M.I. Bursik, Sedimentation from turbulent jets and plumes, Journal of Geophysical Research-Solid Earth, 101 (B3), 5575-5589, 1996.
Fedotov, S.A., S.T. Balesta, V.N. Dvigalo, A.A. Razina, G.B. Flerov, and A.M. Chirkov, New Tolbachik Volcanoes, in Active volcanoes of Kamchatka, edited by O.M. Vanyukova, pp. 302, Nauka Publisher, Moscow, 1991.
Geist, A., A. Beguelin, J. Donagarra, W. Jaing, R. Mancheck, and V. Sunderam, PVM:Parallel Virtual Machine - a User's Guide and Tutorial for Networked Parallel Computing., 279 pp., The MIT Press, Cambridge, Ma, 1997.
Glaze, L.S., and S. Self, Ashfall dispersal for the 16 September 1986, eruption of Lascar, Chile, calculated by a turbulent-diffusion model, Geophysical Research Letters,18 (7), 1237-1240, 1991.
Hanna, S.R., G.A. Briggs, and R.P. Hosker, Handbook on atmospheric diffusion, 102 pp., Technical Information Center (U.S. Department of Energy), 1982.
Hildreth, W., and R.E. Drake, Volcano Quizapu, Chilean Andes, Bulletin of Volcanology,54, 93-125, 1992.
Hill, B.E., C.B. Connor, M.S. Jarzemba, P.C. La Femina, M. Navarro, and W. Strauch, 1995 eruptions of Cerro Negro volcano, Nicaragua, and risk assessment for future eruptions, Geological Society of America Bulletin, 110 (10), 1231-1241, 1998.
Hurst, A.W., and R. Turner, Performance of the program ASHFALL for forecasting ashfall during the 1995 and 1996 eruptions of Ruapehu volcano, New Zealand Journal of Geology and Geophysics, 42 (4), 615-622, 1999.
Inman, D.L., Measures for describing the size distribution of sediments, Journal of Sedimentary Petrology, 22, 125-145, 1952.
Iverson, R.M., S.P. Schilling, and J.W. Vallance, Objective delineation of lahar-inundation hazard zones, Geological Society of America Bulletin, 110, 972-984, 1998.
Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G. White, J. Woollen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K.C. Mo, C. Ropelewski, J. Wang, A. Leetmaa, R. Reynolds, R. Jenne, and D. Joseph, The NCEP/NCAR 40-year reanalysis project, Bulletin of the American Meteorological Society, 77 (3), 437-471, 1996.
Keam, R.F., Tarawera: the volcanic eruption of 1886, 472 pp., R.F. Keam, Auckland, 1988.
Kokelaar, B.P., Setting, chronology and consequences of the eruption of Soufrière Hills Volcano, Montserrat (1995-1999), in The eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999, edited by T.H. Druitt, and B.P. Kokelaar, Geological Society, London, Memoir, 2002.
38
Koyaguchi, T., and M. Ohno, Reconstruction of eruption column dynamics on the basis of grain size of tephra fall deposits. 1. Methods, Journal of Geophysical Research,106 (B4), 6499-6512, 2001a.
Kunii, D., and O. Levenspiel, Fluidisation Engineering, Wiley and Sons, New York, 1969.
Lacroix, A., La Montagne Pelée et ses eruptions, Masson, Paris, 1904. Macedonio, G., M.T. Pareschi, and R. Santacroce, A numerical simulation of the Plinian
fall phase of 79 AD eruption of Vesuvius, Journal of Geophysical Research-Solid Earth and Planets, 93 (B12), 14817-14827, 1988.
Maulucci Vivolo, F.P., Lettere di Plinio il Giovane a Tacito sull'eruzione del Vesuvio (da C. Plini Caecili Secundi Epistularum libri decem), pp. 39, Bastogi Editrice Italiana, Foggia, 1994.
Melnik, O., Dynamics of two-phase conduit flow of high-viscosity gas-saturated magma: large variations of sustained explosive eruption intensity, Bulletin of Volcanology,62 (3), 153-170, 2000.
Morton, B., G.L. Taylor, and J.S. Turner, Turbulent gravitational convection from maintained and instantaneous source, Proceedings of the Royal Society, 234, 1-23, 1956.
Nairn, I.A., Sheet V16AC-Tarawera. Geological map of New Zealand 1:50000, Department of Scientific and Industrial Research, Wellinghton, 1989.
Nairn, I.A., S. Self, J.W. Cole, G.S. Leonard, and C. Scutter, Distribution, stratigraphy, and history of proximal deposits from the c. AD 1305 Kaharoa eruptive episode at Tarawera Volcano, New Zealand, New Zealand Journal of Geology and Geophysics, 44, 467-484, 2001.
Newhall, C.G., and R.P. Hoblitt, Constructing event trees for volcanic crises, Bulletin of Volcanology, 64, 3-20, 2002.
Pasquill, F., Atmospheric Diffusion, John Wiley & Sons, Chichester, 1974. Pinel, V., and C. Jaupart, Magma chamber behavior beneath a volcanic edifice, Journal
of Geophysical Research-Solid Earth, 108 (B2), 2003. Prata, A.J., and I.F. Grant, Retrieval of microphysical and morphological properties of
volcanic ash plumes from satellite data: Application to Mt Ruapehu, New Zealand, Quarterly Journal of the Royal Meteorological Society, 127 (576), 2153-2179, 2001.
Pyle, D.M., The thickness, volume and grainsize of tephra fall deposits, Bulletin of Volcanology, 51 (1), 1-15, 1989.
Sahetapy-Engel, S., Tephra fall deposit of the AD 1305 eruption of Tarawera, New Zealand: reconstruction of eruption dynamics, Master thesis, University of Hawaii, Honolulu, 2002.
Sparks, R.S.J., The dimensions and dynamics of volcanic eruption colums, Bulletin of Volcanology, 48, 3-15, 1986.
Sparks, R.S.J., M.I. Bursik, G.J. Ablay, R.M.E. Thomas, and S.N. Carey, Sedimentation of tephra by volcanic plumes .2. Controls on thickness and grain-size variations of tephra fall deposits, Bulletin of Volcanology, 54 (8), 685-695, 1992.
Sparks, R.S.J., M.I. Bursik, S.N. Carey, J.S. Gilbert, L.S. Glaze, H. Sigurdsson, and A.W. Woods, Volcanic Plumes, 574 pp., John Wiley & Sons, Chichester, 1997.
39
Sparks, R.S.J., and G.P.L. Walker, The significance of vitric-enriched air-fall ashes associated with crystal-enriched ignimbrites, Journal of Volcanology and Geothermal Research, 2, 329-341, 1977.
Sparks, R.S.J., and L. Wilson, Explosive volcanic eruptions: V. Observations of plume dynamics during the 1979 Soufrière eruption, St. Vincent, Geophysical Journal of the Royal Astronomical Society, 69, 551-570, 1982.
Sparks, R.S.J., L. Wilson, and H. Sigurdsson, The pyroclastic deposits of the 1875 eruption of Askja, Iceland, Philosophical Transaction of the Royal Society of London, 229, 241-273, 1981.
Sparks, R.S.J., and S.R. Young, The eruption of Soufrière Hills Volcano, Montserrat (1995-1998): overview of scientific results, in The eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999, edited by T.H. Druitt, and B.P. Kokelaar, Geological Society, London, Memoir, 2002.
Sterling, T.L., J. Salmon, D.J. Becker, and D.F. Savarese, How to build a Beowulf - a Guide to implementation and application of PC clusters, 239 pp., The MIT Press, Cambridge, Ma, 1999.
Stirling, M.W., and C.J.N. Wilson, Development of a volcanic hazard model for New Zealand: first approaches from the methods of probabilistic seismic hazard analysis, Bulletin of the New Zealand Society for Earthquake engineering, 35 (4), 266-277, 2002.
Suzuki, T., A theoretical model for dispersion of tephra, in Arc Volcanism, Physics and Tectonics, edited by D. Shimozuru, and I. Yokoyama, pp. 95-113, Terra Scientific Publishing Company (TERRAPUB), Tokyo, 1983.
Wadge, G., P. Jackson, S.M. Bower, A.W. Woods, and E. Calder, Computer simulations of pyroclastic flows from dome collapse, Geophysical Research Letters, 25 (19), 3677-3680, 1998.
Walker, G.P.L., The Taupo Pumice: product of the most powerful known (Ultraplinian) eruption?, Journal of Volcanology and Geothermal Research, 8, 69-94, 1980.
Walker, G.P.L., The Waimihia and Hatepe plinian deposits from the rhyolitic Taupo Volcanic Centre, New Zealand Journal of Geology and Geophysics, 24, 305-324, 1981.
Walker, G.P.L., S. Self, and L. Wilson, Tarawera, 1886, New Zealand - A basaltic Plinian fissure eruption, Journal of Volcanology and Geothermal Research, 21,61-78, 1984.
Williams, H., Volcanoes of the Paricutin Region, USGS Bulletin, 965 (B), 165-275, 1950. Wilson, L., and T.C. Huang, The influence of shape on the atmospheric settling velocity
of volcanic ash particles, Earth and Planetary Sciences Letters, 44, 311-324, 1979.
Woo, G., The mathematics of natural catastrophes, 292 pp., Imperial College Press, London, 2000.
Woods, A.W., The fluid-dynamics and thermodynamics of eruption columns, Bulletin of Volcanology, 50 (3), 169-193, 1988.
Woods, W., The dynamics of explosive volcanic eruptions, Review of Geophysics (AGU),33 (4), 495-530, 1995.
Te
st/S
cena
rio
Fig.
G
rain
size
Dis
tribu
tion
Md(
)m
in-m
ax
Den
sity
(k
g m
-3)
pum
ices
lith
ics
Col
umn
Hei
ght
(km
)
Ven
tH
eigh
t (m
)
Erup
ted
Mas
s(x
1010
kg)
K
(m2 s-1
) D
iff. L
aw
cut
off
(
s)
Dur
.
(hrs
)
Win
ds
[max
spee
d]
(m s-1
)
Dep
osit
Thre
shol
d (k
g m
-2)
C
ALI
BR
ATI
ON
Run
s 1
Rua
pehu
(to
t dat
a)
7,8
-0.8
* 2.
4*-8
– 1
0*
1100
26
50
8.5*
30
00
0
.5*
0.01
-1
BF: 0
.2 0
.1-3
000
BF: 2
00
4-10
800
BF: 7
27*
R
uape
hu*
[28]
* -
Run
s 2
Rua
pehu
(d
ist.
data
) 7,
8 -0
.8*
2.4*
-8 –
10*
11
00
2650
8.
5*
3000
0.5
* 0.
01-1
BF
: 0.3
0.1
-300
0 BF
: 600
4-10
800
BF: 1
080
7*R
uape
hu*
[28]
*-
Run
s 3
Kah
aroa
F
8,9
1.7
2.5
-7 –
10
1000
23
50
26.4
10
00
110*
1
600
1080
3
Kah
aroa
F
[27]
-
H
AZA
RD
CU
RV
ES
Run
s 4
ULS
13
-0
.8-4
2.
5 -7
– 1
0 10
00
2350
26
10
00
127
1 60
010
80
6 19
96-1
998
-R
uns 5
ER
S 13
-0
.8-4
2.
5 -7
– 1
0 10
00
2350
14
-26
1000
4-
116
0.5-
1 60
010
80
2-6
1996
-199
8 -
PR
OB
AB
ILIT
Y M
APS
Run
s 6
ULS
14
-0
.8-4
2.
5 -7
– 1
0 10
00
2350
26
10
00
127
1 60
010
80
6 19
96-1
998
30,3
00
Run
s 7
ERS
15
-0.8
-4
2.5
-7 –
10
1000
23
50
14-2
6 10
004-
116
0.5-
1 60
010
80
2-6
1996
-199
8 30
,300
R
uns 8
M
ES
16
-0.8
-4
2.5
-7 –
10
1000
23
50
14-2
6 10
004-
116
0.5-
1 60
010
80
2-6
1996
-199
8 30
,300
R
uns 9
U
LS
17a
-0.8
-4
2.5
-7 –
10
1000
23
50
26
1000
127
1 60
010
80
6 96
-98
- Win
ter
30R
uns 1
0 U
LS
17b
-0.8
-4
2.5
-7 –
10
1000
23
50
26
1000
127
1 60
010
80
696
-98
- Sum
mer
30
Run
s 11
ULS
18
a -0
.8-4
2.
5 -7
– 1
0 10
00
2350
26
10
0012
7 1
600
1080
6
1996
- ne
utra
l 30
Run
s 12
ULS
18
b -0
.8-4
2.
5 -7
– 1
0 10
00
2350
26
10
0012
7 1
600
1080
6
97-9
8 - E
NSO
30
Run
s 13
ULS
18
c -0
.8-4
2.
5 -7
– 1
0 10
00
2350
26
10
0012
7 1
600
1080
6
1997
- El N
iño
30R
uns 1
4 U
LS
18d
-0.8
-4
2.5
-7 –
10
1000
23
50
26
1000
127
1 60
010
80
619
98- L
a N
iña
30
Val
ues
in it
alic
bol
d ar
e th
e pa
ram
eter
s va
ried
to te
st th
e se
nsiti
vity
of t
he m
odel
. Val
ues
in re
gula
r bol
d ar
e th
e pa
ram
eter
s st
ocha
stic
ally
sam
pled
dur
ing
sim
ulat
ions
. BF
is th
e be
st-f
it va
lue.
Gra
insi
ze D
istr
ibut
ion
is th
e to
tal g
rain
size
dis
tribu
tion
of th
e te
phra
-fal
l dep
osit
expr
esse
d w
ith th
e In
man
par
amet
ers (
Md
and
[Inm
an, 1
952]
) and
the
min
imum
and
max
imum
pa
rticl
e di
amet
er (
); gr
ains
ize
para
met
ers
for R
uape
hu a
re fr
om th
e 17
Jun
e 19
96 d
epos
it (R
uns
1-2)
; gra
insi
ze p
aram
eter
s fo
r Kah
aroa
F a
re a
vera
ged
from
the
grai
nsiz
e di
strib
utio
ns in
Fi
g. 6
c (R
uns
3); M
d fo
r Run
s 4-
14 is
sto
chas
tical
ly s
ampl
ed b
etw
een
the
two
end-
mem
ber d
istri
butio
ns in
Fig
. 6c.
Den
sity
is th
e m
easu
red
dens
ity o
f the
eru
pted
cen
timet
ric c
last
s (k
g m
-3);
dens
ity o
f all
pum
ice
parti
cles
is m
ade
varie
d be
twee
n th
e de
nsity
of c
entim
etric
pum
ices
and
the
dens
ity o
f cen
timet
ric li
thic
s ac
cord
ing
to th
e de
nsity
mod
el in
Fig
. 10.
Col
umn
Hei
ght i
s th
e m
axim
um h
eigh
t of t
he e
rupt
ive
plum
e (k
m).
Vent
Hei
ght i
s th
e he
ight
of t
he e
rupt
ive
vent
(km
). Er
upte
d M
ass
is th
e to
tal t
ephr
a-fa
ll m
ass
erup
ted
(kg)
det
erm
ined
by
eq.
(13)
. is
the
mas
s-di
strib
utio
n fa
ctor
(eq.
(10)
). K
is th
e di
ffus
ion
coef
ficie
nt (m
2 s-1
) (eq
. (6)
). D
iff. L
aw c
ut o
ff is
the
fall
time
at w
hich
the
diff
usio
n la
w s
witc
hes
betw
een
linea
r and
po
wer
law
(eq.
(6) a
nd (8
)). W
inds
are
the
win
d pr
ofile
s us
ed in
the
sim
ulat
ions
with
the
max
imum
val
ue in
dica
ted
in s
quar
e br
acke
ts (R
uape
hu is
the
win
d pr
ofile
obs
erve
d du
ring
the
17
June
199
6 er
uptio
n (N
CEP
Rea
naly
sis d
ata)
; Kah
aroa
F is
the
win
d de
term
ined
with
the
met
hod
of C
arey
and
Spa
rks [
1986
] for
the
Kah
aora
F e
vent
with
the
dire
ctio
n al
ong
the
disp
ersa
l ax
is o
f the
dep
osit;
199
6-19
98 a
re p
rofil
es s
ampl
ed 4
tim
es a
day
bet
wee
n 19
96 a
nd 1
998;
96-
98W
inte
r and
Sum
mer
are
the
prof
iles
for t
he a
ustra
l win
ter a
nd su
mm
er b
etw
een
1996
and
19
98; 1
996-
neut
ral a
re th
e pr
ofile
s fo
r th
e ye
ar 1
996
that
was
cha
ract
eriz
ed b
y ne
utra
l con
ditio
ns; 9
7-98
EN
SO a
re p
rofil
es f
rom
199
7-19
98 th
at w
ere
year
s ch
arac
teriz
ed b
y El
Niñ
o So
uthe
rn O
scill
atio
n ph
enom
enon
; 199
7-El
Niñ
o an
d 19
98-L
a N
iña
are
the
prof
iles
durin
g El
Niñ
o an
d La
Niñ
a flu
ctua
tions
occ
urre
d in
199
7 an
d 19
98 re
spec
tivel
y). D
ur. i
s th
e du
ratio
n of
the
erup
tive
epis
ode
in h
ours
. Dep
osit
Thre
shol
d is
the
haza
rdou
s thr
esho
ld u
sed
to c
ompi
le h
azar
d cu
rves
and
pro
babi
lity
map
s (kg
m-2
). *,
obs
erve
d da
ta.
Appendix I
Nomenclature
Dimensions of each term are given in brackets: L = length, T = time, M = mass.
jit , fall time of a particle of size j released from a point source i along the eruptive
plume [T] 'it horizontal diffusion time in the volcanic plume at a point source i [T]
vj particle terminal velocity of a particle of size j in the atmosphere [M L-1]
pv particle settling velocity within the eruptive plume as empirically calculated by
Suzuki [1983] to fit data from [Wilson and Huang, 1979] [M L-1]
wo upward velocity of particles at vent [L T-1]
xw component of the wind speed along the x axis [L T-1]
yw component of the wind speed along the y axis [L T-1]
w(zi) upward velocity of particle within the eruptive plume [L T-1]
,x y coordinates of a point on the ground
iii zyx ,, coordinates of a point source i along the eruptive plume from where
particles are released
jix , x coordinate of the center of the Gaussian distribution of mass on ground of
particles of size j and released from a point i along the eruptive plume
(layers
jiji xxx , ) [L]
jiy , y coordinate of the center of the Gaussian distribution of mass on ground of
particles of size j and released from a point i along the eruptive plume
(layers
jiji yyy , ) [L]
Y(zi) dimensionless parameter proportional to the ratio of upward particle velocity (i.e.
plume velocity) to particle settling velocity within the eruptive plume
Yo dimensionless parameter proportional to the ratio of initial particle velocity at
vent to particle settling velocity within the eruptive plume
zi height of a point source i along the eruptive plume [L]
dimensionless parameter that controls the shape of the mass-distribution function
within the eruptive plume
min minimum value of considered for a specific set of simulations
duration of the Plinian discharge [T]
jt time spent by a particle of size j within each atmospheric layer [T]
jx wind transport of a particle of size j along the x axis within an atmospheric layer
( j x jx w t ) [L]
jy wind transport of a particle of size j along the y axis within an atmospheric layer
( j y jy w t ) [L]
z thickness of each atmospheric layer [L]
µ dynamic viscosity [M L-1 T-1]
a air density [M L-3]
p particle density [M L-3]
dep density of tephra-fall deposit [M L-3]
Graphic Standard Deviation (grainsize parameter) [Inman, 1952] [L2]2,i j variance of the Gaussian mass distribution on the ground of particles of size j
released from a point source i [L2]
p standard deviation of the Gaussian distribution of the mass in the ascending plume
[L]
granulometric unit: = -log2(103d), where d is the particle diameter in m
max maximum particle diameter
min minimum particle diameter
Figure captions
Fig. 1. Map of the North Island of New Zealand showing: (a) main cities (black
diamonds), important populated towns (grey circles: Mk, Maketu; Mt, Matata; Ec, Edgecumbe; W, Whakatane; TT, Te Teko; K, Kawerau), Tarawera Volcanic Complex (small grey triangle) and the Taupo Volcanic Zone (large shaded triangle); (b) Tarawera Volcanic Complex (adapted from Nairn [1989] and Sahetapy-Engel [2002]) with the main domes, old lava deposits and the AD 1886 eruptive fissure. The names of the 4 lava domes produced during the AD 1305 Kaharoa eruption are also shown (only Crater Dome, Ruawahia Dome and Wahanga Dome produced tephra fall; [Sahetapy-Engel, 2002]).
Fig. 2. Isopach map of total tephra-fall accumulation (cm) from the AD 1305 Kaharoa eruption [Sahetapy-Engel, 2002]. Dash line indicates 5 cm isopach. The triangle indicates the Tarawera Volcanic Complex. Note the two prominent lobes to the southwest and to the northwest.
Fig. 3. Semi-log plots showing the variation, for different particle sizes, of: (a) fall time (ti,j); (b) fall time + diffusion time in the plume (ti,j + t’i); (c) variance (σi,j
2) (eq. (6) and eq. (8)). Calculations are done between 1 and 35 km (1 km step), with K= 900 m2 s-1 and fall-time threshold=3600 s. Note how the plume diffusion time (t’i) mainly affects coarse particles ((a) and (b)). Note also the step in σi,j
2 values at 5-6 km due to the shift of diffusion law for particles with diameter of 0.25 mm (2 φ) ((c); (eq. (6) and eq. (8)).
Fig. 4. Mass fraction of particles of a specific grainsize, here specified by settling velocities of 0.25, 1, and 2.5 m s-1, released as a function of height from a volcanic plume of total height 10 km, calculated using eq. (6) and (8). Initial eruption velocity at the vent, ow , is assumed to be 100 m s-1. Substantially different curves are generated for (a) 1=β and (b) 01.0=β .
Fig. 5. Frequency plots showing output results for: (a) plume heights stochastically sampled between Log(14 km) and Log(26 km) (bin = 200 m); (b) eruptive-episode duration stochastically sampled between 2 and 6 hours (bin = 0.1 hours); (c) total erupted mass determined from the combination of plume heights and eruption duration above (eq. (13)) (bin = 2.00E+10 kg).
Fig. 6. Frequency plots showing: (a) values of β stochastically sampled between 0.5 and 1 (bin = 0.01) and (b) Mdφ stochastically sampled between -0.8 and 4 φ (bin = 0.1 φ). φ = -log2(103d), where d is the particle diameter in m. (c) Cumulative plot showing the total grainsize distribution of the Waimihia Plinian (white), Hatepe Plinian (black) and Taupo Plinian (gray) from Walker [1980] and Walker [1981]. Given that all original grainsize distributions are truncated at 3 φ, the grainsize trend beyond 3 φ is extrapolated based on a log-normal distribution. Resulting Mdφ are: -0.8, 0.6 and 4 φ respectively. Resulting σφ are: 2.3, 2.3 and 2.9 respectively.
Fig. 7. Semi-log plots showing the misfit function (mf; eq. (15)) for all points (grey diamonds; primary axis; Runs 1 in Table 1) and for distal points (white triangles; secondary axis; Runs 2 in Table 1) (17 June 1996 eruption of Ruapehu, New Zealand) calculated for: (a) diffusion coefficient, K, (b) fall-time cut off and (c) mass-
distribution parameter, β. Observed data from BF Houghton (manuscript in preparation, 2003).
Fig. 8. Log plots showing best fit for: (a) Ruapehu total data (114 data) (DC = 200 m2 s-
1; fall-time threshold = 72 s, i.e. 0.02 hours); (b) Ruapehu distal data (88 points) (DC = 600 m2 s-1; fall-time threshold = 1080 s, i.e. 0.3 hours); (c) Kaharoa F using Ruapehu distal-data best fit (DC = 600 m2 s-1; fall-time threshold=1080 s). Ruapehu proximal data (< 5 km) plotted with distal-data best fit are also shown in (b). Runs 1-3 in Table 1. The equiline is also shown for comparison. Observed data for the Ruapehu and Kaharoa eruptions are from BF Houghton (manuscript in preparation, 2003) and Sahetapy-Engel [2002] respectively.
Fig. 9. Comparison between isomass maps for the AD 1305 Kaharoa eruption (Unit F) compiled from field data (from Sahetapy-Engel [2002]) and the computed deposit. The computed deposit is run with DC = 600 m2 s-1 and fall-time cut off = 1080 s (see alsoTable 1). Tephra-fall accumulation is in kg m-2. Contours for the field-based map are spaced every 50 kg m-2; the 10 kg m-2 contour is also shown. Black dots on the field-based map indicate field data. Contours for the computed map are spaced every 10 kg m-2 up to 50 kg m-2, and every 100 kg m-2 after 100 kg m-2. Some key cities (diamonds) and Tarawera Volcano (triangle) are also shown.
Fig. 10. Density parameterization for the AD 1305 Kaharoa eruption (from Bonadonna and Phillips [2003]). Density of pumices (gray circles; 1000 kg m-3) and lithics (black triangles; 2350 kg m-3) with diameter >2 mm is directly measured [Sahetapy-Engel, 2002]. Density of lithics is assumed to be constant, whereas density of pumices with diameter <2 mm is assumed to decrease linearly and to reach the lithic density value when the fragment diameter decreases below 8 µm.
Fig. 11. Plots showing (a) mean wind direction (provenance + 180°) and (b) mean wind velocity averaged for each level (i.e. every 1 km) over 3 years of wind profiles (1996-1998) sampled 4 times a day (see text for details). The standard deviation determined for each level is also shown. NCEP Reanalysis data provided by the NOAA-CIRES Climate Diagnostics Center.
Fig. 12. Plots showing: (a) the percentage of wind profiles characterized by ≥ 15 levels with wind blowing between 0° and 90°, 90° and 180°, 180° and 270° and 270° and 360° respectively (degrees are from the North); (b) the percentage of such wind profiles distributed amongst each month. Percentages are calculated out of the total wind sample analyzed (i.e. 1996-1998).
Fig. 13. Hazard curves computed for the Upper Limit Scenario (thick line) and the Eruption Range Scenario (thin line) for key cities and towns (in Fig. 1): (a) Maketu, (b) Kawerau, (c) Tauranga, (d) Taupo, (e) Wellington and (f) Rotorua. Hazard Curves show the probability of exceeding a given accumulation of tephra fall (kg m-2). Runs 4 and 5 in Table 1.
Fig. 14. Upper Limit Scenario maps computed for a deposit threshold of: (a) 30 kg m-2 (damage to agriculture) and (b) 300 kg m-2 (minimum loading for roof collapse). The map for a 600 kg m-2 threshold is not shown as it only indicates probability <5% for the populated areas around the volcano. Contours are spaced every 10% probability of reaching a given threshold. The 5% contour is also shown (thick solid line). Key cities and towns are also shown (circles; from Fig. 1). Tarawera Volcanic Complex is indicated with a triangle. Runs 6 in Table 1.
Fig. 15. Eruption Range Scenario maps computed for a deposit threshold of 30 kg m-2 (damage to agriculture). The map for a 300 kg m-2 threshold is not shown as it only indicates probability <5% for the populated areas around the volcano. Contours are spaced every 10% probability of reaching a given threshold. The 5% contour is also shown (thick solid line). Key cities and towns are also shown (circles; from Fig. 1). Tarawera Volcanic Complex is indicated with a triangle. Runs 7 in Table 1.
Fig. 16. Multiple Eruption Scenario maps computed for a deposit threshold of: (a) 30 kg m-2 (damage to agriculture) and (b) 300 kg m-2 (minimum loading for roof collapse). The map for a 600 kg m-2 threshold is not shown as it only indicates probability <5% for most of the populated areas around the volcano (apart from Kawerau, i.e. 5%). Contours are spaced every 10% probability of reaching a given threshold. The 5% contour is also shown (thick solid line). Key cities and towns are also shown (circles; from Fig. 1). Tarawera Volcanic Complex is indicated with a triangle. Runs 8 in Table 1.
Fig. 17. Upper Limit Scenario maps computed sampling wind data (a) from June through August (austral winter; winds mainly blowing between 0° and 180°, see Fig. 7b) and (b) from September to March (austral spring-summer; when winds also blow between 180° and 360°, see Fig. 7b). Deposit threshold is 30 kg m-2 (damage to agriculture). Contours are spaced every 10% probability of reaching a given threshold. The 5% contour is also shown (thick solid line). Key cities and towns are also shown (circles; from Fig. 1). Tarawera Volcanic Complex is indicated with a triangle. Runs 9 and 10 in Table 1.
Fig. 18. Upper Limit Scenario maps computed sampling wind data (a) from 1996 (neutral conditions), (b) from 1997 through 1998 (characterized by El Niño Southern Oscillation phenomenon), (c) from April 1997 through May 1998 (characterized by a strong El Niño fluctuation) and (d) between May and December 1998 (characterized by La Niña fluctuation). Deposit threshold is 30 kg m-2 (damage to agriculture). Contours are spaced every 10% probability of reaching a given threshold. The 5% contour is also shown (thick solid line). Key cities and towns are also shown (circles; from Fig. 1). Tarawera Volcanic Complex is indicated with a triangle. Runs 11-14 in Table 1.