Introduction
Volcanic ash hazards
Primary hazards associated with explosive volcanic eruptions include
pyroclastic density currents (flows and surges), the widespread deposition of
air fall tephra and the threats to aviation posed by volcanic ash in the
atmosphere. Simulation of all possible hazards with one model is difficult
due to the fact that different length scales dominate different hazards. Our
focus here is the hazard that volcanic ash poses to aircraft.
During volcanic eruptions, volcanic ash transport and
dispersion models (VATDs) are used to forecast the location and movement of ash clouds at
timescales that range from hours to days. VATDs use eruption source
parameters, such as plume height, mass eruption rate, duration and the mass
fraction distribution of erupted particles finer than about 4Φ (or
62.5 µm), which can remain in the cloud for many hours or days.
Observational data for such parameters are usually unavailable in the first
minutes or hours after an eruption is detected. Moreover, these input
parameters are subject to change during an eruption, requiring rapid
reassignment of new parameters. Usually, plume models are used to provide
these source terms for VATDs and the forecast accuracy is critically
dependent on these models. This paper reports on a new three-dimensional (3-D)
volcanic plume model designed to exploit the advantages of
mesh-free methods for 3-D modeling of such plumes that involve multiphase
free boundary flows.
Existing plume models
Several one-dimensional (1-D) volcanic plume models have been developed in
the past few decades, ranging from the most basic 1-D model
which only accounts for mass conservation to more
recently developed 1-D models
which tend to account for more comprehensive physics effects. For example,
FPLUME-1.0 accounts for wind effect, entrainment of
moisture, water phase change, particle fallout and re-entrainment, and even
wet aggregation of ash. However, in these 1-D models, the entrainment of air
is evaluated based on two coefficients: entrainment coefficient due to
turbulence in the rising buoyant jet and the crosswind field. Different 1-D
models adopt different entrainment coefficients based on specific formulation
or calibration against well-documented case studies. The feedback from plume
to atmosphere is usually ignored in 1-D models. Even though determination of
essential parameters such as the entrainment is not based on first
principles, such simple models nevertheless allow us to investigate the
importance of physical mechanisms in a volcanic plume. In addition, these
simplified models require little computational resource and can run on
standard personal computers or on websites in a very short time. As a result,
1-D software for volcanic plume development such as, combined with VATDs such as is
widely used in research and practice. While these 1-D models can generate
well-matched results with three-dimensional (3-D) models for weak plumes,
much greater variability is observed for strong plume scenarios, especially
for local variables . In addition, there is need for
greater skill in hazard forecasts especially where the plume model is used
to generate source conditions for complex 2-D (two-dimensional) and 3-D VATD
models.
The development of 2-D and 3-D time-dependent and multiphase numerical
models for volcanic plumes has provided new explanations for many features of
explosive volcanism. For example, a recent study based on a 3-D fluid dynamical
model shows that simple extrapolations of
integral models for Plinian columns to those of super-eruption plumes are not
valid, and their dynamics diverge from current ideas of how volcanic plumes
operate. One of the earliest of these is the 3-D model pyroclastic
dispersion analysis code (PDAC) which is a
non-equilibrium, multiphase, 3-D compressible flow model. Conservation
equations for each phase are solved separately with the finite volume method.
A parallel computing version of PDAC was also developed
. Advanced numerical techniques, such as a
second-order scheme and semi-explicit time stepping, were also adopted afterwards to
improve the accuracy of PDAC .
Another 3-D model, SK-3D , is a 3-D time-dependent
fluid dynamics model that attempts to reproduce the entrainment process of
eruption clouds with relatively simple physics but with high-order numerical
accuracy and high spatial resolution. A series of simulations based on SK-3D
was reported, including establishment of the relationship between the
observable quantities of the eruption clouds and the eruption conditions at
the vent , investigation of the effect of the
intensity of turbulence in the umbrella cloud on dispersion and sedimentation
of tephra , determination of the entrainment
coefficients of eruption columns as a function of height
and investigation of the effect of wind field on
the entrainment coefficient .
While SK-3D focuses on accurately capturing the entrainment caused by
turbulent mixture with higher resolution and numerical method of higher
order, PDAC takes the disequilibrium between different phases into account
and hence is a true multiphase model. Another 3-D model, the Active Tracer
High-Resolution Atmospheric Model (ATHAM) , focuses more
on microphysics within the plume. As pyroclastic flow is not the initial
concern of ATHAM; dynamic and thermodynamical equilibrium is assumed in
ATHAM. The dynamic core of ATHAM solves the compressible Euler equations for
momentum, pressure and temperature of the gas–particle mixture
. The subgrid-scale turbulence closure scheme
that differentiates between the horizontal and vertical directions
is adopted to capture turbulent mixing. The
cloud microphysics predicts the mass of hydrometeors in liquid and ice phases
. Additional modules, including gas-phase chemistry
and gas scavenging by hydrometeors
, were added lately. A further extension was made
to include particle aggregation
. However, the resolution of
ATHAM is still coarse compared with SK-3D and PDAC.
Besides adding to their special strengths (ATHAM has been adding more and
more microphysics; PDAC was extended to consider more phases), these models
are also adding core strengths. PDAC development has begun to include the
effect of microphysics into the model, while ATHAM development has extended
its ability to modeling pyroclastic flow. Both are using finer and finer
resolution.
Recently, a first-order, non-equilibrium compressible 3-D model, ASHEE
, was introduced based on three-dimensional
N-phase Eulerian transportation equations, which are a full set of mass,
momentum and energy transport equations for a mixture of gas and dispersed
particles. ASHEE is valid for low concentration and low Stokes number regions
and much faster than the N-phase Eulerian model. The model is based on the
open-source numerical solver OpenFOAM , adapting its
unstructured finite volume solver.
To summarize, each 3-D model has its own focus based on the problem of
interest and modeling/numerical choices made. Accuracy of simulation
(depending on comprehensiveness of the model, resolution of discretization,
numerical error and order of accuracy) and the simulation time (depending on
number of governing equations, resolution, numerical methods and parallel
techniques) are always conflicting considerations in 3-D plume simulations.
Features of SPH
To the best of our knowledge, all of the existing 3-D plume models use
mesh-based Eulerian methods, and there are no 3-D plume models based on mesh-free
Lagrangian methods. Lagrangian methods have several features that are
suitable for volcanic plume simulation that we outline below. Among such
Lagrangian methods, smoothed particle-hydrodynamics-based simulations have shown
good agreement with experiments for many applications in fluid dynamics.
Currently, they are, by far, the most widely used mesh-free schemes. Our
implementations of SPH in volcanology follow earlier efforts
. Specifically,
we choose SPH as the numerical method for volcanic plume simulation to
enable
better investigation of mixing phenomena;
accurate modeling of the development of the zone of flow establishment (ZFE),
zone of established flow (ZEF) investigation and relation to column collapse and
the questions relating to the development of entrainment; and
easy inclusion of particles of different sizes (phases) and investigation
of detailed mechanics of sedimentation and drag force interaction in lower plume.
These are enabled by the following features of SPH:
The advection term in the Navier–Stokes equations does not appear explicitly in
discretized formula of SPH (as illustrated in
Eqs. –).
It is easy to include various physics effects (like self gravity, radiative
cooling and chemical reaction) in the model. It does not require a major overhaul
and re-tooling every time new physics is introduced . This
implies that accounting for more physics is easier for the SPH model.
With more than one phase, each described by its own set of particles, interface
problems between phases are often trivial for SPH but difficult for mesh-based schemes.
Thus, multiphase flow can be easily handled by SPH. Adding new phases to the model also
does not require a major overhaul and re-tooling. As will be shown in later paragraphs,
adding new phases only leads to adding several lines into the source code for new
phases and additional interaction terms between existing phases and newly added phases.
Interface tracking is explicit in SPH through capturing of the locations of the
particles. Less numerical effort is required for interface construction when we attempt
to include the effects of mixing by resolving the detailed interface structure and dynamics of turbulence.
As discussed in the previous paragraph, existing 3-D plume models focus on
one or several specific aspects of the plume and have been extended to be more
comprehensive by accounting for more physics or more phases. Easy
extensibility and capability of handling multiple phase flow with less
additional numerical effort greatly facilitate future extension of SPH
models. As volcanic plumes are in nature multiphase and without predefined
boundary in the atmosphere, SPH is a suitable numerical method for plume
modeling. The core physics, such as entrainment of air and thermal expansion,
are essential for all plume modeling, while some other physics, such as water
condensation and aggregation, are important in specific scenarios. As an
initial effort towards exploiting advantages of SPH in volcanic plume modeling, we
focus on capturing basic features in plume development using a numerically
robust and computationally efficient framework with support for scalable
parallel computing.
Open-source availability and the relatively easy extensibility of SPH will
facilitate development of a more comprehensive community-driven model.
Our contributions
Even though SPH has been known for several decades, implementations of SPH
for compressible multiphase turbulent flows are few.
proposed a multiphase SPH model for numerical
simulation of air entrapment in violent fluid-structure interactions.
and proposed weakly
compressible and incompressible multiphase SPH solvers in their papers.
also proposed new formulations for weakly
compressible fluid with the speed of sound sufficiently large to guarantee that
the relative density variations are typically 1 %.
recently presented a new SPH model for weakly compressible multiphase flows
with complex interfaces and large density differences. All of them focus on
incompressible or weakly compressible flow.
Several issues endemic to classical SPH, like tensile instabilities,
compressible turbulence modeling and turbulent heat exchange, are fixed in
our implementation. The most popular applications of SPH (and their original
motivating application) have been in the simulation of free surface flow, such
as breaking waves and floods. Less attention was paid to velocity inlet and
pressure outlet boundary conditions which are required in plume modeling.
We develop methodology to impose pressure boundary conditions by adding
extra layers of static ghost particles. Additional constraints on the time step
are used to avoid the growth of numerical fluctuations near the pressure boundary.
We impose a velocity inlet boundary condition by placing several layers of ghost
particles moving with eruption velocity.
The turbulence model is crucial for reproducing the entrainment of air. There
are several turbulence models proposed for the SPH method .
We adopt a Lagrangian-averaged Navier–Stokes (LANS) turbulence model ,
which was originally proposed for incompressible flow and is extended here for compressible
flow accounting for turbulent heat exchange.
Corrected formulation of SPH is adopted to bypass the
well-known tensile instability issues of classical SPH.
Simulation of volcanic plumes with acceptable accuracy requires fine resolution
(very high particle counts) that cannot be accomplished without parallel computing using
large process counts. The core solver of our model is parallelized by distributed memory
message passing interface (MPI) standard parallelism. In addition, a dynamic load balancing strategy is also developed.
Imposition of some types of boundary conditions (such as eruption boundary condition)
requires dynamically adding and removing of particles during simulation. To address this
issue, we adopt an efficient data management scheme based on a time-dependent space-filling curve (SFC)
induced indexing and hash table. The computational cost is further reduced by adjusting the simulation
domain adaptively.
The physical model of the plume is first presented in
Sect. and leads to a complete mathematical
description of the volcanic plume (governing equations and boundary
conditions). In Sect. , we briefly introduce the
numerical tool – the SPH method. Both the fundamental discretization
formulation and techniques that are used to handle specific issues involved
in plume modeling are discussed. Verification and validation with numerical
tests are presented in Sect. . In
Sect. , a discussion on future work is given following a
brief summary.
Physical model
Description of the model
During an explosive eruption, a volcanic jet erupts out from a vent with a
speed of several tens to more than 150 m s-1, driven by expanding gas.
The jet is initially denser than the surrounding atmosphere and begins to
decelerate through negative buoyancy and turbulent interaction with
surrounding air. Cauliflower-like vortices are generated along jet margins,
within which process, air is entrained and heated up, reducing the bulk
density of the entire jet, in many cases, to less than that of the
surrounding atmosphere. Once it becomes buoyant, such a jet develops into a
Plinian or sub-Plinian plume, rising up to several kilometers to tens of
kilometers until its heat is exhausted. Jets that lose their momentum before
becoming buoyant collapse back onto the ground and transform into pyroclastic
flows, surges and ignimbrites. During the process of plume rising up,
relatively larger particles might separate from main stream of the plume,
falling down onto the ground and possibly be re-entrained into the plume at a
lower height . Within this process, erupted
vapor condenses to liquid (droplet) and even further to ice. Latent heat
released from phase change of erupted vapor further heats up entrained air
and further dilutes the bulk density. The entrained vapor might also
experience a similar process and impact plume development. Particle
aggregation processes ,
either due to presence of liquid water, resulting from particle collision or
driven by electrostatic forces, might occur inside plume and thereby affect
the sedimentation.
All in all, the process of plume development is essentially a multiphase
turbulent mixing process coupled with heat transfer and other microphysical
and chemical reactions.
As an initial effort towards exploiting the feasibility and advantage of SPH in
plume modeling, our model is designed to describe an injection of well-mixed
solid and volcanic gas from a circular vent above a flat surface into a
stratified stationary atmosphere following SK-3D .
In this model, molecular viscosity and heat conduction are neglected since
turbulent exchange coefficients are dominant. Erupted material consisting of
solid materials with different size and mixture of gases is assumed to be well mixed
and behave like a single-phase fluid (phase 2) which is valid for eruptions
with fine particles and ash. Air (also a mixture of different gases) is
assumed to be another phase (phase 1). Thermodynamic equilibrium is assumed
so that no separate energy equation is needed for each phase. As a result,
there is only one energy equation for both phases (heat exchange term between
different phases does not show up under this assumption). Dynamic equilibrium
is assumed so that no separate momentum equation is needed for each phase. As
a result, there is only one vector momentum equation for both phases (drag
force term does not show up with this assumption).
Because of the above assumptions, all other microphysical processes (such as
the phase changes of H2O, aggregation, disaggregation, absorption
of gas on the surface of solids, solution of gas into a liquid) and chemical
processes are not considered in this model. These ignored microphysics
factors would play critical roles under particular eruption conditions.
Capturing of these processes needs a more comprehensive model. One critical
element in plume development, the effect of wind, is also not yet considered
in this model. Introducing wind effects in our model requires dynamic
pressure boundary conditions, which requires more numerical effort and
algorithm design. To summarize, our model is not valid for eruptions where
wind effect plays a significant role in its development, usually referred to as a
weak plume. Our model also lacks the ability in modeling plumes with large
particles or eruptions in which microphysics plays non-ignorable roles, such
as an eruption of the El Chichón volcano on 4 April 1982
. We are focused here on
developing the SPH-based methodology in the context of the more basic (and
more critical) aspects of the volcanic plume and therefore devote our effort to
this relative simpler model. It is worthwhile to mention here that because
SPH is adopted as our numerical method, adding these physics into our
model would require much less work in terms of programming compared to
mesh-based methods. Since our plan is an open-source distribution of the tool, we
believe some of these enhancements will rapidly ensue with community
participation.
Governing equations
Based on above assumptions, the governing equations of our model are given as
(which is the same as the governing equations of SK-3D)
∂ρ∂t+∇⋅ρv=0∂ρξ∂t+∇⋅ρξv=0∂ρv∂t+∇⋅ρvv+pI=ρg∂ρE∂t+∇⋅ρE+pv=ρg⋅v,
where ρ is the density, v is the velocity, ξ is the mass
fraction of ejected material, g is the gravitational acceleration,
and I is a unit tensor. E=e+K is the total energy which is a
summation of kinetic energy K and internal energy e. An additional
equation is required to close the system. In this model, the equation for
closing the system is the following equation of state (EOS):
p=γm-1ρe,
where
γm=Rm/Cvm+1Rm=ξgRg+ξaRaCvm=ξsCvs+ξgCvg+ξaCvaξa=1-ξξg=ξ⋅ξg0ξs=ξ-ξg,
where Cv is the specific heat with constant volume, and R is the gas
constant. ξ is the mass fraction of erupted material. The subscript “m”
represents mixture of ejected material and air, “s” represents solid
portion in the ejected material, “g” represents gas portion in the ejected
material, “a” represents air, and 0 represents physical properties of erupted
material. ξg0 is the mass fraction of vapor in the erupted
material.
In mesh-based methods, governing equations in Eulerian form,
Eqs. ()–(), are directly discretized.
For SPH, governing equations in Lagrangian form are needed. By deducting
kinetic energy from the energy equation, subtracting mass conservation from
the momentum equation, combining the transient term and advection term into the material
derivative term (for any function A, material derivative is defined as
dAdt=∂A∂t+v⋅∇A), the governing equations are put into the final form, in which
the advection term does not appear explicitly.
dρdt+ρ∇⋅v=0dρξdt+ρξ∇⋅v=0dvdt+∇Pρ=gdedt+P∇⋅vρ=0
Boundary conditions
In the current model, the initial domain is a box. The boundaries are
categorized as the velocity inlet (a circular area at the center of the
bottom of the box), wall boundary (box bottom) and pressure outlet (other
faces of the box).
Velocity inlet
At the vent, the temperature of erupted material T, eruption velocity
v, mass fraction of vapor in erupted material ξg0 and
mass discharge rate M˙ are given. The pressure of erupted material p
is assumed to be the same as ambient pressure for pressure-balanced eruption.
The radius of the vent is determined from ρ, M˙ and v.
Equations ()–() give the velocity inlet
boundary condition written in terms of primitive variables:
ρ=const=p/RmTξ=const=1v=const={u,v,w}T∂e∂n=M˙e0/πr2,
where r is the radius of the vent, n is the direction normal to the
boundary, and e0 is specific internal energy of erupted material which can be
calculated based on temperature and specific heat of erupted material.
Non-slip wall boundary
Velocity is zero for the non-slip wall boundary. If we assume the boundary to be
adiabatic, heat flux should be zero on the boundary. The flux of mass should
also be zero. As a result, internal energy flux, which consists of heat flux
and energy flux carried by mass flux, vanishes on the wall boundary.
Equations ()–() give the non-slip wall
boundary condition written in terms of primitive variables:
∂ρ∂n=const=0∂ξ∂n=const=0v=const={0,0,0}T∂e∂n=0.
Open outlet pressure boundary condition
The pressure of the surrounding atmosphere is given. Except for the pressure,
boundary values for density, velocity and energy on the outlet should depend
on the solution. As we ignore the viscosity, the shear stress is ignored and
normal stress (whose magnitude equals its pressure) balances the ambient
pressure.
p=pa(z)
SPH method
SPH is a mesh-free Lagrangian method. In SPH, the domain is discretized by a
set of particles or discretization points and the position of each particle
is updated at every time step based on the motion computed. Approximation of
all field variables (velocity, density and pressure, etc.) is obtained by
interpolation based on discretization points. The physical laws (such as
conservation laws of mass, momentum and energy) written in the form of partial differential equations
(PDEs) or ordinary differential equations (ODEs)
need to be transformed into the Lagrangian particle formalism of SPH. Using a
kernel function that provides the weighted estimation of the field variables
at any point, the integral equations are estimated as sums over particles in
a compact subdomain defined by the support of the kernel function associated
with the discretization points. Thus, field variables associated to the
particle are updated based on its neighbors. Each kernel function has a
compact support determined by the smoothing length of each particle. There are
several review papers by
, , ,
and ,
giving a pretty comprehensive view over SPH. We only refer here to the
representation of the constitutive equations in SPH and put more focus on
specific numerical techniques for plume modeling.
Fundamental principles
There are several procedures for discretizing governing equations (PDEs or
ODEs) with SPH. We present here one of them following
. The
starting point of approximating a function with SPH is the translation
property of the Dirac function δ(x), for an arbitrary function
A(x), the following equation holds.
A(x)=∫-∞∞Ax′δx′-xdx′
The Dirac function δ in Eq. () can be
approximated by a weighting function wx-x′,h (or
wx′-x,h which tends to a Dirac function when the
smoothing length h→0:
limh→0wx′-x,h=δx′-x.
The weighting function, as an approximate form of the Dirac function, should
satisfy the normalization condition:
∫wx-x′,hdx′=1.
Besides normalization, the weighting function of particle a has to be
symmetric with respect to a to ensure that neighbor particles located at
the same distance away from a contribute equally to the SPH summation equation;
see Eq. ().
wx-x′,h=wx′-x,h
The weighting function also needs to satisfy conditions such as positivity
and compact support. In addition, the kernel function must be monotonically
decreasing with the distance between particles.
There is a wide variety of possible weighting functions that can satisfy
these requirements, such as spline functions (with different orders) and
Gaussian functions. Generally, the accuracy increases with the order of the
polynomials of the kernel function, but the computational cost also increases
as the number of interactions increases. We adopt a truncated Gaussian
function as the weighting function in our simulation.
wx-x′=1hπdexp-x-x′h2|x-x′|≤3h0Otherwise,
where d is the number of dimensions. The derivative of the weighting function
is
∇wx-x′=-2x-x′h1hπdexp-x-x′h2|x-x′|≤3h0Otherwise.
By replacing the δ function in Eq. () with the
kernel function w, an arbitrary function A(x) can then be
approximated by
Ax≈<Ax>=∫ΩAx′wx-x′,hdx′+Oh2.
As the weighting function is symmetric (Eq. )
and satisfies the normalization condition
(Eq. ), odd error terms in
Eq. () vanish, leading to a second-order
approximation. However, in practice, a second order of accuracy cannot be
achieved because there is no guarantee of the symmetry of particle
distribution in real simulation . Recall that
dx′=dm(x′)ρ(x′); the
integration equation, Eq. (), can be
approximated by summation and can lead to an approximation of the function A:
<Ax>≈∑bmbAbρbwx-xb,h,
where the summation is over all the particles within the region of compact
support (see Eq. ) of the weighting function. Gradient
terms may be straightforwardly calculated by taking the derivative of
Eq. (), giving
<∇Ax>=∂∂x∫ΩAx′wx-x′,hdx′+Oh2≈∑bmbAbρb∇wx-xb,h.
For vector quantities, the expressions are similar, simply replacing A with
A in Eq. () and
Eq. (), giving
<Ax>≈∑bmbAbρbwx-xb,h<∇⋅Ax>≈∑bmbAbρb⋅∇wx-xb,h<∇×Ax>≈∑bmbAbρb×∇wx-xb,h<∇jAix>≈∑bmbAbiρb∇jwx-xb,h.
Artificial viscosity
In classical SPH, shock waves are handled by introducing artificial
viscosity, a term that is defined based on second derivatives of velocity, to
smear out discontinuities. As in the case of first-order derivatives,
second-order derivatives can be estimated by differentiating a SPH interpolation
twice. However, such a formulation has two disadvantages: first, it is very
sensitive to irregular distribution of particles; second, the second
derivative of the kernel can change sign and lead to unphysical
representations (for example, viscous dissipation causes decrease of the
entropy).
One of the most commonly used models of artificial viscosity
is
Πab=-νρ‾abvab⋅xabxab2+ηh2.
The coefficient ν is defined as
ν=αh‾abc‾ab,
where
c‾ab=ca+cb2ρ‾ab=ρa+ρb2vab=va-vbxab=xa-xb.
The artificial viscosity term Πab is a Galilean invariant and vanishes
for rigid rotation. It produces a repulsive force between two particles when
they are approaching each other and an attractive force when they are
receding from each other.
The SPH viscosity can be related to a continuum viscosity by converting the
summation to integrals . It has been shown that
shear viscosity coefficient η=ραhc8 and bulk
viscosity coefficient ζ=5η3 are appropriate for
two-dimensional flows. η=ραhc10 and ζ=5η3 are appropriate for three-dimensional flows. An extra term was
added to ν considering aspects of the dissipative term in shock solutions
based on Riemann solvers and led to a new formulation of artificial
viscosity. We adopt this new formulation in our simulation:
Πabβ=-αμabc‾ab+βμab2ρ‾abvab⋅xab<00vab⋅xab>0,
where
μab=hvab⋅xabxab2+(ηh)2.
α and β are two parameters that can be adjusted for different
cases. α=1 and β=2 are recommended by Monaghan for best
results. In our simulation, these two parameters are calibrated to α=0.3 and β=0.6. η is usually taken as 0.1 to prevent
singularities.
Discretization of governing equations and extensibility
The basic interpolation given in
Eqs. ()–()
provides a general way to obtain SPH expressions of governing equations. The
problem is that using these expressions “as is” in general leads to quite
poor gradient estimates. Various tricks can be used to conserve linear and
angular momentum and thermal energy . Special
treatments are also needed for second-order derivative terms
. We only refer here to one of these possible
discretizations of compressible Euler equations with SPH:
<ρa>=∑bmbwab(h)dvadt=-∑bmbpbρb2+paρa2+Πab∇awab(h)+gdeadt=0.5∑bmbvabpbρb2+paρa2+Πab⋅∇awab(h),
where a is the SPH particle index. Π is an artificial viscosity term,
which is discussed in Sect. . wab(h) is
a concise form of wxa-xb,h, and from here
on, we will use this concise form. As a Lagrangian method, particle position
is also updated at every time step.
dxadt=va
We highlight an important feature of the SPH methodology. Adding new physics
and new phases into the model is trivial in terms of discretization. For
example, adding a new source (or sink) into Eq. (), adding
a drag force into Eq. () and adding a heat exchange term
into Eq. () lead to the new discretization form:
<ρa>=∑mbwab(h)+ρ˙x,t
dvadt=-∑bmbpbρb2+paρa2+Πab∇awab(h)+g+D∑bmbvb-vaρbdeadt=0.5∑bmbvabpbρb2+paρa2+Πab⋅∇awab(h)+∑bmbρbκa+κbTa-Tbxa-xbwab(h),
where the source term ρ˙ can be a “sink” of erupted vapor due to
its phase change. D is a drag force coefficient. κ is the heat
conduction coefficient. T is the temperature. Other physics can be added
easily in a similar way. Adding these new terms leads to modification of
only a few lines in the source code. The drag force term should show up when
dynamic disequilibrium between different phases is considered. In that case,
each phase needs one set of governing equations of Navier–Stokes type.
Adding a new phase into SPH code only requires adding a few new lines for the new
phase besides interaction terms introduced by the new phase.
Time step
The physical quantities (velocity, density and pressure) and particle
position change every time step. The Courant condition, which is in spirit
similar to the Courant condition for the mesh-based methods, is used to
determine the time step Δt.
Δt=CFLminamaρa1dca,
where ca is sound speed at particle a and calculated based on
heat-specific ration of the mixture γm (see
Eq. ).
ca=γmpρ0.5
First-order Euler integration, with CFL=0.2, is used to advance in
time.
Tensile instability and corrected derivatives
The classical SPH method was known to suffer from tensile instability and
boundary deficiency. Tests of the standard SPH method indicate an instability
in the tensile regime, while the calculations are stable in compression. A
simple example of such a test calculation exhibiting the instability involves
a body which is subject to a uniform initial stress, either compressive or
tensile. If the initial stress is tensile, a very small velocity perturbation
on a single particle can lead to particles clumping together, forming large
voids and seriously corrupting density distribution. But if the initial
stress is compressive, the small velocity perturbation on a single particle
cannot lead to any changes in particle distribution
. To address these difficulties,
proposed a corrected SPH formulation. For the 1-D
case, employing a Taylor expansion for A(x) about xa, multiplying both
sides by kernel function and then doing an integration over the domain gives
∫ΩA(x)wx-xa,hdx=Aa∫Ωwx-xa,hdx+∂A∂x(xa)∫Ωx-xawx-xa,hdx+….
Ignoring derivative terms higher than first order and writing the integral
in particle approximation form leads to
Aa=∑bmbAbρbwab(h)∑bmb1ρbwab(h).
Notice that the denominator in Eq. ()
is actually the summation approximation of the left side of
Eq. (). That is to say,
Eqs. ()
and () are the same for particles far away from
boundaries as the denominator in Eq. ()
becomes 1 in that case. The first-order derivative term can be obtained in
a similar way:
∇Aa=∑bmbAb-Aaρb∇awab(h)∑bmbxb-xaρb∇awab(h).
For problems of higher dimension, the expressions for function approximation
are exactly the same as Eq. (), even
though the derivation is different. The first-order derivative can be
obtained by solving a system of equations explicitly or numerically
.
Mass fraction update
Air and erupted material are represented by two different sets of SPH
particles (or discretization points) in the model. Based on assumptions we
made in Sect. , only density needs to be updated
respectively for each phase. The updating of density is exactly the same as
Eq. () in spirit. Particles of phase 1 are not counted while
evaluating density of phase 2, and vice versa. Updating of density is then
based on the following discretized equations.
<ραa>=∑mbwαb(h)∑mbρbwαb(h)+∑mjρjwαj(h)
<ραsg>=∑mjwαj(h)∑mbρbwαb(h)+∑mjρjwαj(h),
where the subscripts a and b represent air particles (phase 1), while i
and j represent particles of erupted material. β and α
represent either erupted material particles or air particles.
ρaa is density of phase 1 (air). ρisg is
density of phase 2 (erupted material). ρ=ρa+ρsg
is density of mixture of air and erupted material. By definition, the mass
fraction is updated according to Eq. ().
<ξα>=ραsgρα
In areas far away from the interface, updating of density is exactly the same
as that for single-phase flow. For example, on the right side and left side
(or blue areas) in Fig. , where there are only
air particles, Eq. () evaluates to zero, and total density
is
ρα=ραa=∑mbwαb(h)∑mbρbwαbh,
which is a special case of Eq. (). For
these areas occupied by only particles of phase 2 and far away from the
interface, similarly, the equation for the density update becomes
ρα=ραsg=∑mjwαj(h)∑mjρjwαj(h).
That is to say, the same density updating equation can be applied for both
phases and no additional numerical treatment is needed to locate where the
interface is.
In panel (a), the blue particles (phase 1) represent
air particles; the red ones (phase 2) represent erupted material.
Panel (b) shows corresponding mass fraction. Mass fractions are evaluated
based on Eqs. () and () without any
other interface track or capture method.
Interface construction will become necessary and important when attempting to
include the effects of mixing by resolving the detailed interface structure
and dynamics of turbulence. As a Lagrangian method, interface tracking in SPH
is explicit through capturing of the locations of the particles, much
simpler than Eulerian methods. The existence of complex evolving interfaces
between phases presents severe challenges to conventional Eulerian grid-based
numerical methods. Either interface tracking (Lagrangian)
or
interface capturing (Eulerian)
methods are used to reconstruct the flow interface of free boundary flow.
High computational cost, a tendency to form numerical instabilities and the
inability to track complex topological changes are the significant drawbacks
of tracking techniques
. For the interface
capturing (Eulerian) method, the surrendering of surface detail before the
phase transport calculation means that interface reconstruction is required
between time steps to recover the interface information, which needs
additional numerical effort . Since SPH
is able to adaptively adjust the discretization and automatically construct
the interface, SPH requires less additional numerical effort for interface
construction and therefore is more suitable for volcanic plume simulation.
Turbulence modeling with SPH
For high-speed shearing flow, the momentum exchange and heat transfer are
dominated by turbulent fluctuations as turbulent exchange coefficients are
several magnitudes larger than corresponding physical coefficients (molecular
viscosity and heat conduction coefficient). In addition to momentum and energy
exchange, mixing between plume and air is important in plume modeling.
Quantifying these mixing processes in real implementation is challenging
because of the scale disparity between the large-scale fluid motion and the
diffusion processes on interface that ultimately lead to mixing. Ideally, one
would like to be able to include the effects of mixing on the large-scale
dynamics without resolving the detailed interface structure and dynamics of
turbulence to reduce computational cost. To resolve all turbulent exchange at
all different scales and sub-particle-scale mixing with relative coarse
resolution, a SPH sub-particle-scale (SPH-SPS) turbulence models should be included. Among existing
SPH-SPS turbulence models
, we adopt a LANS-type turbulence model, the
SPH-ε turbulence model .
However, the SPH-ε turbulence model was proposed only for
incompressible flow. In the following section, we will extend it for
compressible flow. It is necessary to mention that all other existing SPH-SPS
turbulence models
also only
focus on incompressible flow.
Lagrangian average in SPH-ε
constructed SPH-ε
turbulence model within the framework of SPH in such a way that general
principles such as conservation of energy, momentum and circulation are
satisfied using the ideas associated with the LANS turbulence modeling. The
basic idea of SPH-ε is to determine a smoothed (averaged
in space) velocity v^ by a linear operation on the
unsmoothed velocity v. The SPH particles move with this smoothed
velocity, and hence the average motion of the fluid is determined by the
averaged velocity v^:
dxadt=v^a.
The average of physical quantities over space introduces extra terms into the
governing equations. Once the form of the smoothing (average) is chosen, these
extra terms are determined. The typical LANS model uses a smoothed velocity
v^ defined in terms of the unsmoothed velocity v by
v^(x)=∫vx′G|x′-x|,ldx′,
where G satisfies
∫G|x′-x|,ldx′=1,
and is a member of a sequence of functions which tends to the δ
function in the limit when l→0. A typical example is Gaussian.
The length scale l determines the characteristic width of the kernel and
the distance over which the velocity is smoothed.
It is a common practice in LANS to use a differential equation for the
smoothing rather than the integral form and finally reach a system of
equations that need to be solved implicitly. In the SPH-ε
method, a XSPH smoothing is adopted which
conserves linear and angular momentum. In this way, solving a system of
equations is avoided, and it also makes the method simple to implement and
cheap for computation. The discretized form of the momentum equation is
obtained through lengthy derivation. Derivation and other discussions are
available in the literature; see, e.g., .
Here, we provide a brief summary of the key steps.
The smoothing adopted by is
v^(x)=v(x)+ϵ∫vx′-v(x)G|x′-x|,ldx′.
As function G has the same feature as kernel function w, SPH
approximation of the integration leads to
v^(x)=v(x)+ϵ∑bmbvb-vρbG|xb-x|,l.
By making the replacement,
G|xb-xa|,lρb→KabM,
where Kab=ldGab, M=ρ0ld in which ρ0 is initial
density. The SPH-ε turbulence model is obtained after lengthy
derivation:
dvadt=-∑bmbpbρb2+paρa2∇awab(h)+∑bmbε2vab⋅vabM∇aKab.
Notice that, if l is constant,
∇Kab=∇ldGab=ld∇Gab.
The discretized momentum equation with the SPH-ε turbulence
model can be written in terms of Gab instead of Kab:
dvadt=-∑bmbpbρb2+paρa2∇awab(h)+∑bmbΦab∇aGab(l),
where
Φab=ε2vab⋅vabρb,
which is the extra stress term induced by average. We take coefficient
ε as 0.8 following .
For compressible flow, the energy equation is coupled with the momentum
equation and mass conservation equation. Averaging of thermal energy over
space introduces some additional terms besides the stress term induced by
velocity average . The averaged momentum
equation for compressible flow is in the same form as that for
incompressible flow; all of the other additional terms, besides the corresponding
velocity-average-induced stress term, show up in the energy equation. Most
turbulence modeling focuses on the stress terms induced by the average of
velocity. These stress terms are usually either solved directly (for example,
LANS methods) or defined via a constitutive relation (for example, large eddy
simulation method). Less attention is typically given to the other terms.
Most commonly, a Reynolds analogy is used to model the turbulent exchange.
Simulations of heat transfer, or other scalar transfer, in turbulent flow
simply involve adding transport terms for thermal energy or species
concentration, at the expense of greater storage and longer computing times
but without other difficulties . We adopt this
strategy. The additional terms associated with molecular diffusion and
turbulent transport in the energy equation are either modeled in different
ways or neglected sometimes . We neglect
these terms in our simulation.
Turbulent heat transfer
We adopt the Reynolds analogy to get the heat transfer coefficient due to
turbulence. The Prandtl number is defined as
Pr=Cpμκ,
where μ is the dynamic viscosity, and κ is the thermal conductivity.
In addition, μ can be written in terms of the absolute viscosity (kinematic
viscosity) as
μ=ρν.
Then,
κ=CpμPr.
The typical value of Prt for air is 0.7–0.9. We take
Prt=0.85 for gases as recommended
from summarizing experimental results.
summarized the simulation of viscosity and heat
conduction in his review on SPH. We will refer to his summary in our
following discussion. The additional term in the discretized momentum equation,
Eq. (), is the turbulent shear stress term.
Recall that molecular viscosity can be discretized with SPH as shown in
Eq. (). It has been shown that the discretized
molecular viscosity has both bulk viscosity and shear viscosity, where shear
viscosity coefficient is
νt=Sν,
with
S=110ifd=318ifd=2.
The turbulent viscosity coefficient can be inferred from that formulation if
we can reformulate the turbulent shear stress term in a form which is similar
to the molecular shear term. Reformulating the turbulent shear stress term,
∑bε2mbρbvab⋅vab∇aGabla=∑bε2SmbvabρbSvab⋅xabxab2xab2xab∇aGabla,
the turbulent viscosity coefficient can be inferred from
Eq. ().
νt=ε2Svab⋅xab1
Please note that the turbulent viscosity term has the opposite sign of the molecular
viscosity term in the discretized momentum equation and there is a minus sign in
the expression of Πab, and they cancel out.
However, the above equation is correct only for the 1-D situations. For 2-D
or 3-D, it is not easy to get an explicit expression. We adopt an alternative
way: obtaining a value for each pair of particles instead of persisting on
getting an analytical expression. Choosing the smoothing function to be the same
as the SPH kernel and the smoothing length scale l to be the same as the smoothing
length h, the ratio between turbulent shear stress and physical shear
stress is
Υab=ε2vab⋅vabρbSνρabvab⋅xabxab2+η2hab2=εxab2+η2hab22Sνvab⋅vabvab⋅xab.
Υab is essentially equivalent to the ratio between the turbulent
viscous effect of particle b on particle a and molecular viscous effect
of particle b on particle a. Turbulent viscosity can be easily obtained
by
νt,ab=νΥab=εxab2+η2hab22Svab⋅vabvab⋅xab.
The corresponding turbulent thermal conductivity should be
κt,ab=εCp,ab‾ρab‾xab2+η2hab2vab⋅vab2SPrtvab⋅xab.
Cp,ab‾ and ρab‾ are simply the
arithmetic means of specific heat and density. The term used to prevent
singularity now can be removed.
κt,ab=εCp,ab‾ρab‾xab2vab⋅vab2SPrtvab⋅xab
We also need to prevent singularity, so
κt,ab=0ifvab=0orxab=0εCp,ab‾ρab‾xab2vab⋅vab2SPrtvab⋅xabotherwise.
A cross-section view of the simulation domain in the y-z plane at
66 s. Panel (a) shows the mass fraction. Panel (b) shows all boundary
conditions: the dark blue region is occupied by eruption ghost particles
with a “ghost particle ID” of 0; the light blue area is occupied by pressure
ghost particles with a “ghost particle ID” of 1; the gray area is filled with
wall ghost particles with a “ghost particle ID” of 2. The “ghost particle ID”
of all real particles is set to 100; they occupy the major portion of the
domain in panel (b). Panel (c) shows the cross-section view of domain
decomposition based on SFC. The simulation is conducted on 12 processors, so
there are 12 subdomains in total. The cross-section view shows a portion of
them.
The heat conduction equation without the source term is
CpdTdt=1ρ∇κ∇T.
The second spatial derivative can be approximated with SPH by following
:
CpdTdt=∑bmbρaρbκa+κbTa-TbFab(h),
where Fab(h) is short for Fxa-xb,h, whose definition is
Fab(h)xab=∇awab(h).
Fab is always non-positive, which guarantees that heat flux flows from
hot to cold. Plug the turbulent thermal conductivity into the heat conduction
equation:
CpdTdt=∑bmbρaρbκa+κbTa-TbFab(h)=2∑bmbρaρbCp,ab‾ρab‾εxab2vab⋅vab2PrtSvab⋅xabTa-TbFab(h).
Notice that the number “2” in the front of
Eq. () comes from integration approximation
of the second-order derivative . By further
simplification, we get
CpdTdt=εSPrt∑bmbρaρbCp,ab‾ρab‾xab2vab⋅vabvab⋅xabTa-TbFab(h).
Discretized governing equations with SPH-ε turbulence model
Plugging in the discretized turbulent stress term and turbulent heat transfer
term into the momentum and energy equation, we get new discretized governing
equations:
dvαdt=-∑bmbpbρb2+pαρα2+Παbβ-Φαb∇αwαb(h)-∑jmjpjρj2+pαρα2+Παjβ-Φαj∇αwαj(h)+g,
with
Φαβ=ε2vαβ⋅vαβρβ
deαdt=0.5∑bmbvαb^pbρb2+pαρα2+Παbβ-Φαb∇αwαb(h)+2∑bmbραρbκt,αbTα-TbFαb(h)+0.5∑jmjvαb^pjρj2+pαρα2+Παjβ-Φαj∇αwαj(h)+2∑jmjραρjκt,αjTα-TjFαj(h),
with κt,αβ given by
Eq. (). As the particle-scale movement of
flow is based on smoothed velocity, the velocity in the energy equation
should also be smoothed. The filtering process is done according to
Eq. (). Position of particles is updated
according to Eq. (). Smoothed velocity is
also used while computing artificial viscosity.
Boundary conditions
All boundary conditions are imposed by ghost particles.
Figure b shows how boundaries are deployed.
Wall boundary condition
Traditionally, either ghost particles that mirror real particles across the
boundary or boundary forces
have been used to impose the wall boundary conditions. One disadvantage of
the latter is that the boundary forces tend to corrupt the solution in the
local neighborhood. In addition, a natural way of imposing the eruption boundary
condition is using eruption ghost particles. To impose boundary conditions in
a consistent way, we adopt a modified version of the ghost particle method
for wall boundary conditions. Stationary wall ghost
particles are deployed in the same way as real particles. Instead of
enforcing symmetry particle by particle, a symmetric field across the
boundary is explicitly enforced. Ghost particles are reflected into the
domain and physical quantities are calculated at these reflected positions by
SPH interpolations. It should be noted that wall ghost particles should not
be counted when computing physical properties of these reflected positions.
All properties, except for velocity, are assigned at the corresponding reflected
position to the ghost particle. The velocity of each wall ghost particle is
set to have the same value but opposite direction of the interpolated
velocity at its corresponding reflection. This way, the non-slip wall
boundary condition (Eq. ) is imposed naturally. These wall
ghost particles serve as neighbors in momentum and energy update. More
implementation details about this method can be found in
. As these wall ghost particles are stationary,
there is no mass flux on the boundary (Eqs.
and ). In addition, as temperature is also symmetric with
respect to the boundary, the gradient of temperature vanishes, and hence there
is no internal energy flux on the wall boundary (Eq. ). In
our current model, the ground is assumed to be flat. For more complicated
topography, it has been shown in other work that
this method works as well. We do have concern regarding potential limitation
of this method of deployment of ghost particles for more complicated
boundaries in three dimensions. Fortunately, the current model does not involve
complicated wall boundary.
Eruption boundary condition
A natural way of imposing eruption boundary condition is using ghost
particles that move with the eruption velocity and bear the temperature of
the erupted material. A parabolic velocity profile that represents a fully
developed Hagen–Poiseuille flow is used to determine the inlet particle
velocity. The detailed shape of the parabolic profile is determined based on
an averaged eruption velocity (Eq. ). The mass of eruption
ghost particles is set to a value so that evaluation of
Eq. () can return a density that is consistent with the
value given in Eq. (). The internal energy associated
with these particles is set to a value so that Eq. () is
satisfied. The mass fraction of erupted material (Eq. )
is automatically satisfied as all particles in the eruption conduit are of
phase 2. The density, momentum and internal energy of these eruption ghost
particles are not updated before they move aboveground. As soon as they move
out from eruption conduit, these ghost particles will be shifted to real
particles and their physical quantities and position will be updated based on
discretized governing equations. New ghost particles need to be added at the
bottom of the eruption conduit as these existing ghost particles move
upwards.
Pressure boundary condition
Another boundary condition in our model is the pressure outlet boundary. For
flow in a straight channel, it is possible to treat the exit the same as the
entry with a prescribed velocity profile. For flow with a more complex channel,
an exit far downstream of the flow disturbance is also feasible. However, the
natural boundary condition (Eq. ) is more suitable for
plume simulation as the outlet is open atmosphere. The way we impose pressure
boundary condition is by adding several layers of pressure ghost particles
surrounding the real atmosphere particles. Pressure, density and temperature
are determined based on the elevation of particles. Velocity is set to zero
for static atmosphere. The physical quantities for pressure ghost particles
are not updated while these for real particles are updated at every time
step. As the position of all pressure ghost particles keeps constant, we
essentially impose a static pressure boundary condition. Real particles are
removed as soon as they move out of the pressure boundary.
As simulations progress, changes in position and physical quantities of real
particles near pressure boundaries might corrupt the pressure boundary condition
that was established initially. This shortcoming is relieved by choosing a
larger computational domain so that boundaries that might be corrupted are
far away from turbulent mixing area. In addition, to avoid enlarging
fluctuations, we add another constraint on the time step:
Δt≤CFLphv,
where CFLp is a safety coefficient which has similar function to
the normal Courant–Friedrichs–Lewy (CFL) number. Too small CFLp would slow down
simulation, while too large CFLp would lose its ability of mitigate numerical
fluctuation near the boundary. The proper CFLp is determined by a
series of simulation tests.
Parallelism and performance
One disadvantage of the 3-D model is that it usually takes a much longer time
than 1-D models to complete one simulation. This disadvantage further
prevents simulation with finer resolution and accounting for more
physics in one model. Non-intrusive uncertainty analysis, which is commonly
adopted in hazard forecasting, requires finishing multiple simulations within
a given time window. High-performance computing is therefore essential. Among
existing CPU parallel SPH schemes, most of them focus on the neighbor search
algorithm and dynamic load balancing
. Less attention has been paid
to developing more flexible data management schemes for more complicated
problems. Motivated by techniques developed for mesh-based methods, we
develop a complete framework for parallelizing SPH with distributed memory
parallelism (MPI) allowing flexible and efficient data access.
The time complexity for an SPH method is O(N2), where N is the total
number of SPH particles. Efficient neighbor searches and compact-supported
kernel functions can help to reduce computational cost. Of the many possible
choices, we adopt a background grid which was proposed by
and is quite popular in parallel SPH. Then, the
time complexity is reduced to O(MN+mN), where m is the average of number
of particles within the compact support of the kernel function, M is the
average number of particles in subdomain within which neighbor searches are
carried out. This background grid is also used for domain decomposition. We
refer to the elements of the background grid, namely squares for two dimensions
and cubes for three dimensions, as buckets. As for the actual storage of the
physical quantities, different strategies have been adopted in existing
implementations of SPH.
In DualSPHysics , the physical quantities of
each particle (position, velocity, density, etc.) are stored in arrays. The
particles (and the arrays with particle data) are reordered following the
order of the cells. This has two advantages: (1) the access pattern is more
regular and more efficient; (2) it is easy to identify the particles that
belong to a cell by using a range since the first particle of each cell is
known. But adding, deleting and especially accessing particles are
cumbersome. adopted linked lists using pointers so
that particles can be deleted or added during the simulation. Storage
problems caused by fixed-size arrays are thereby also eliminated. We define
C++ classes which contain all data of particles and buckets. As for the
management of data, we adopt hash tables to store pointers to particles and
buckets, which gives us not only flexibility of deleting and adding
elements but also quicker access compared with linked lists. Instead of
using the “natural order” to number particles, we adopt a SFC-based
index to give each particle and background bucket a unique
identifier – a strategy known to preserve data locality at minimal cost. The
SFC-based numbering strategy is further extended to include time step
information so that particles added at the same position but different time
have different identifiers.
The parallelization is achieved by splitting computational domain into
subdomains. Each subdomain is computed by a single processor. For any
subdomain, information from its neighboring subdomains is required when
updating physical quantities. To guarantee consistency, data are synchronized
if physical quantities are updated. Even though more complicated graph-based
partitioning tools might get higher quality
decomposition, they require much more effort in programming and computation.
So we adopt an easy programming scheme based on SFC
to decompose the computational domain.
Figure c shows a cross-section view of domain
decomposition.
More details about the data structure, domain decomposition, load balancing,
domain adjusting and performance benchmarking have been published separately
.
Strong scalability test result.
Weak scalability test result.
The effect of domain adjusting on simulation time.
Panel (a) shows execution time without domain adjusting;
panel (b) shows execution time with domain adjusting. Different bins represent
execution time up to specific physical time indicated by horizontal axis.
Performance tests have been carried out on the computational cluster of Center for Computational Research
(CCR) at the University at Buffalo. Intel Xeon
E5645 CPUs running at 2.40 GHz clock rate with 4 GB memory per core on a
Q-Logic Infiniband are used in these tests. Each node is comprised of two
sockets with six of these cores. Memory and level 3 cache are shared on each
node. The initial domain is [-4.8 km,
4.8 km] × [-4.8 km, 4.8 km] × [0 km, 6 km]. Almost
linear speed up is observed in our strong scalability test
(Fig. ).
The weak scalability test is conducted with the same initial domain and
various smoothing lengths. Each simulation runs for 400 time steps. The
average number of real particles of each process keeps constant at 25 900.
As shown in Fig. , simulation times increase around
one-third when the number of cores increases from 16 to 4208. For the test problem in this
section, the volcanic plume finally reaches a region of [-30 km,
30 km] × [-30 km, 30 km] × [1.5 km, 40 km] after
around 400 s of eruption. When numerical simulation goes up to 90 s, the
plume is still within a region of [-10 km, 10 km] × [-10 km,
10 km] × [0 km, 25 km]. This implies that adjusting of domain can
avoid computing a large number of uninfluenced air particles, especially for
the beginning stage of simulation. A domain-adjusting algorithm
is designed and implemented in our code.
Figure shows that simulation time of the test problem is
greatly reduced when we adopt the domain-adjusting strategy.
Verification and validation
We present a series of numerical simulations to verify and validate our model
in this section. Plume-SPH is first verified by 1-D shock tube tests, then by
a JPUE (jet or plume that is ejected from a nozzle into a uniform
environment) simulation. Velocity and mass fraction distribution both along
the central axis and cross transverse are compared with experimental results.
The pattern of ambient particle entrainment is also clearly shown. Then, a
simulation of representative strong volcanic plume is conducted. Integrated
local variable are comparable with simulation results from existing 3-D plume
models.
1-D shock tube tests
1-D shock tube tests are first conducted to verify our code. Input parameter
of each tests can be found in Table .
These tests can represent typical cases in 1-D. Test 1 consists of a left
rarefaction, a right traveling contact and a right shock. Density decreases
downwind of the contact wave. Test 2 also consists of a left rarefaction, a
right traveling contact and a right shock. Density increases downwind of
the contact wave. Test 3 is a double expansion test with different initial
density.
Input parameters of 1-D shock tube tests.
ρL
pL
vL
ρR
pR
vR
tf
Test 1
1.0
1.0
0
0.5
0.2
0
0.2
Test 2
1.0
1.0
0
0.25
0.1795
0
0.17
Test 3
2.0
1.95
1.0
1.0
1.95
-1.0
0.13
In Table , subscript “L” refers to the left
side and “R” to the right side; tf is the total simulation time. The
initial interval between two adjacent particles is 0.03. The computational
domain is [-0.4,0.4]. The specific internal energy is compared against
exact solutions. As shown in Fig. , the
position and magnitude of the waves are correctly predicted. The fluctuations
near the contact discontinuity are caused by sharp change of smoothing
length.
Comparison of specific internal energy of simulation results against
analytical results for shock tube tests. The plots from left to right
correspond to test 1, test 2 and test 3, respectively.
Simulation of JPUE
JPUE can be considered as a simplified volcanic plume. While the effect of
stratified atmosphere and the effect of expansion due to high temperature in
volcanic plume are not represented, JPUE reproduces the entrainment due to
turbulent mixing which is one of the key elements in the volcanic plume
development. There exist consistently good experimental data
that describe the JPUE flow field giving insight into details of JPUE,
such as transverse velocity and concentration profile. In this section, we
verify that our code and the SPH-ε turbulence model is
able to reproduce the features of turbulent entrainment by a JPUE simulation.
As many of these experiments were conducted with liquid, we replace the
original equation of state (Eq. ) with a weakly compressible Tait
equation of state (see Eq. ) to
avoid solving the Poisson equation:
p=Bρρ0γ-1,
with γ=7 and B is evaluated by
B=ρ0c2γ,
where c is the speed of sound in the liquid. The energy equation is
actually decoupled from the momentum conservation equation and the mass
conservation equation by using this EOS. In addition, the “atmosphere” is
assumed to be uniform and gravity is set to be zero. We set the temperature
and density of ejected material the same as surrounding ambient. This further
simplifies the scenario for the convenience of studying turbulent mixing.
One overall feature of JPUE is “self-similarity”, which means that the
evolution of the JPUE is determined solely by the local scale of length and
velocity, which theoretically account for the fact that the rate of
entrainment at the edge of JPUE is proportional to a characteristic velocity
at each height. As a result, physical and numerical experiments do not
necessarily have exactly the same setups and are compared on a
non-dimensional basis.
List of eruption condition for the test cases.
Parameters
Units
JPUE
Plume
Vent velocity
m s-1
500
275
Vent gas mass fraction
1.0
0.05
Vent temperature
K
273
1053
Vent height
m
0
1500
Mass discharge rate
kg s-1
5.47×107
1.5×109
Dimensionless concentration and velocity distribution across the
cross section.
Panel (a) shows normalized jet width (which determined based on
velocity) along the centerline. Panel (b) shows normalized concentration
along the centerline.
A three-dimensional axisymmetric JPUE which ejects from a round vent is
simulated with eruption parameters listed in
Table . Material properties of water are used as
material properties for both phases. The results are compared with
experiments for
verification purposes. Experimental data of concentration and velocity
distribution across the cross section were fit into a Gaussian profile (see
Eq. ) by
and even
though the actual profiles are slightly different from the Gaussian profile.
φφc=exp-coefrz2,
where φ is either velocity or concentration, the subscript “c”
represents the centerline. r is the distance from the centerline on any
cross section. z is the axial distance from the origin of the jet
transverse section under consideration. The coefficient “coef” for
concentration is 80 and 50, respectively, according to
and .
The “coef” for velocity is 90 and 55, respectively, according to
and .
also fit concentration distribution
and jet width based on velocity along the centerline into a straight line (see
Eq. ).
φ0φc=slopezD+intercept,
where subscript 0 represents the cross-sectionally averaged exit value; D
is the diameter of vent. The “slope” for jet width based on velocity is 0.104
and for concentration is 0.157. The “intercept” for jet width based on velocity
is 2.58, while that for the concentration is 4.35.
Although both velocity and concentration are found to be well matched with
experimental results, a small disparity in both velocity and concentration
is observed near the boundary of the jet, which is possibly caused by
overestimation of the drag effect by standard SPH
. also proposed an
alternative way to update density which relieved the overestimating of the
drag effect. However, how well this method conserves mass is not clear.
There are several other factors that might also attribute to such disparity.
The Reynolds number is not reported in many experiments assuming a high enough
Reynolds number. In addition, some details of the experiments, such as exit
velocity profile and viscosity of the experimental liquid, are not reported.
These factors prevent us from numerically reproducing these experiments in an
exact way as they were conducted. However, the features of JPUE are correctly
reproduced with our code.
We also investigated the mixing due to turbulence in the JPUE simulation by
checking the mixture of the two phases. It is shown in
Fig. that the ejected material and ambient fluids are
mixed through eddies at the outer shear region. Also, the inner dense core
dispersed gradually due to erosion of the outer shear region. Hence, our
confidence in the numerical correctness of our code is greatly reinforced.
Panel (a) shows particle distribution. Particles of phase 1
(blue) are gradually entrained and mixed with erupted particles (red) as jet
flows downstream. Panel (b) shows the mass fraction of erupted material
at the moment corresponding to panel (a).
Simulation of a volcanic plume
The development of a volcanic plume is more complicated than JPUE in several
aspects. Besides turbulent entrainment of ambient fluids, development of
the volcanic plume also involves heating of entrained air and expansion in a
stratified atmosphere. A strong eruption column without wind is tested in
this section for the purpose of further verification and validation. Both
global variables and local variables are compared with existing models.
Input parameters
Eruption parameters, material properties and atmosphere are chosen to be the
same as the strong plume no-wind case in a comparison study on eruptive
column models by . Eruption conditions are listed in
Table . As our model does not distinguish solid
particles of different sizes, only mass fraction of solids of all sizes is
used in simulation even though two size classes were provided by
. The density of erupted material at the vent and
radius of the vent can be computed from the given parameters. The eruption
pressure is assumed to be the same as the ambient pressure at the vent and
hence is not given in the table. The vertical profiles of atmospheric
properties were obtained based on the reanalysis data from the European
Centre for Medium-Range Weather Forecasts (ECMWF) for the period corresponding to
the climactic phase of the Mt. Pinatubo eruption (Philippines, 15 June 1991).
These conditions are more typical of a tropical atmosphere (see Fig. 1b in
). Vertical distribution of temperature, pressure
and density is used to establish stratified atmosphere. Wind velocity and
specific humidity are not used in our simulation even though they were also
provided by (see Fig. 1b). Material properties,
shown in Table , are selected based on
properties of the Pinatubo and Shinmoedake eruptions. Other material properties
not given in the table can be inferred from these given parameters based on
their relationships.
List of material properties.
Parameters
Units
Value
Specific heat of gas at constant volume
J kg-1 K-1
717
Specific heat of air at constant volume
J kg-1 K-1
1340
Specific heat of solid
J kg-1 K-1
1100
Specific heat of gas at constant pressure
J kg-1 K-1
1000
Specific heat of air at constant pressure
J kg-1 K-1
1810
Density of air at vent height
kg m-3
1.104
Pressure at vent height
Pa
84 363.4
Mass fraction for t=500 s after eruption. Panels (a, b) are visualizations of SPH simulation results.
Panel (b) shows visualization of a slice of the computational simulation, whose thickness is
around 10 000 m. The lowest portion of the plume represents erupted
material in the eruption vent (the underground portion).
Panels (c, d) show the contour of the mass fraction and velocity quiver on an x-z plane. These
panels are plotted utilizing postprocessed data (see
Appendix ). The contour levels in the plot are
0.00001, 0.0001, 0.001, 0.1, 0.3, 0.75 and 0.95. Panel (d) is a zoomed view of velocity quiver showing plume expansion and
entrainment of air.
Figure shows the mass fraction of
the simulated volcanic plume at 500 s after eruption, at which time the plume
starts spreading radially. A contour plot of the mass fraction on the
vertical cross section (x-z plane) was also provided. The zoomed view of
the quiver plot shows detailed entrainment of air at the margin of the
plume.
Global and local variables
One of the key global quantities of great interest is the altitude to which
the plume rises. The top height predicted by our model is around 40 km which
agrees with other plume models. For example, the height predicted by PDAC is
42 500 m, by SK-3D is 39 920 m, by ATHAM is 33 392 m and by ASHEE is
36 700 m. As for local variables, the profiles of integrated temperature,
density, mass fraction of entrained air, gas mass fraction, mass fraction of
solid materials and the radius of the plume as a function of height are compared with
existing 3-D models in
Figs. –. To get rid
of significant fluctuations in time and space, we conducted a time-averaging
and spatial integration of the dynamic 3-D flow fields by following
.
Temperature as a function of height.
The mass fraction of entrained air, gas and solids as a function of
height.
As particles distribute irregularly in the space in SPH simulation results,
we need to project simulation results (on irregular particles) onto a
predefined grid before doing time-average and spatial integration. See
Appendix A for more details of postprocessing.
The profiles of local variables match well with simulation results of
existing 3-D models in a general sense. The basic phenomena in volcanic plume
development are correctly captured by our model.
Density of the strong plume without wind after reaching its top
height.
As the height increases, the amount of entrained air also increases. Around
the neutral height, where the umbrella expands, the entrainment of air shows
a slight decrease due to lack of air surrounding the column at that height.
The profile for gas, which accounts for both air and vapor, shows a very
similar tendency to that of entrained air. Recall that vapor condensation is
not considered in our model. In addition, we assume that erupted material
behaves like a single-phase fluid. So the mass fraction of gas is simply a
function of entrained air (Eq. ).
ξa+ξg=ξa+1-ξaξg0
Among these 3-D models, ATHAM takes vapor condensation into account and
Eq. () does not hold for ATHAM. However, the profile
of entrained air and profile of gas predicted by ATHAM are still very close
to each other, which implies that ignoring water phase change is a valid
assumption for eruptions similar to this test case (strong plume with erupted
water fraction in erupted material less than 5 %). This observation can
be explained by the fact that air occupies a larger portion of the gas, and
ignoring phase change of vapor (which is only a small portion of gas)
has a slight influence on plume development. As for mass fraction of solids,
similarly, Eqs. ()
and () hold for our model.
ξs=1-ξa1-ξg0ξs=1-ξa+ξg
Radius of the strong plume without wind after reaching its top
height.
PDAC, which treats particles of two different sizes as two separate phases,
predicted a similar mass fraction profile. That implies that assumption of
dynamic equilibrium in our model is at least valid for eruptions similar to
the test case.
With more cool air entrained into the plume and mixing with the hot erupted
material, the temperature of the plume decreases as the height increases as
shown in Fig. . Meanwhile, bulk density
decreases due to entrainment and expansion
(Fig. ).
Our model adopts the same assumptions and governing equations as SK-3D.
However, there is still an obvious disparity between the profiles of local
variables of our model and SK-3D. One of the big differences between these
two models is that we adopt a LANS type of turbulence model while SK-3D
adopts a large eddy simulation (LES) turbulence model. This implies that
choice of turbulence model might play a critical role in plume simulation.
Conclusions
A new plume model was developed based on the SPH method. Extensions necessary
for Lagrangian methodology and compressible flow were made in the formulation
of the equations of motion and turbulence models. Advanced numerical
techniques in SPH were exploited and tailored for this model. High-performance
computing was used to mitigate the tradeoff between accuracy
(which depends on comprehensiveness of the model, resolution, order of accuracy of
numerical methods, scheme for time upgrading) and simulation time (which depends on
comprehensiveness of model, resolution, order of accuracy of numerical
methods, scheme for time upgrading, etc. and computational techniques).
The correctness of the code and model was verified and validated by a series
of test simulations. Typical 1-D shock tube problems were simulated and
compared against analytical results showing good agreement. Dimensionless
velocity and concentration distribution across the cross section and along
the jet axis match well with experimental results of JPUE. Top height and
integrated local variables simulated by our model are consistent with
simulation results of existing 3-D plume models. Comparison of our
results with those of SK-3D implies that the turbulence model plays a significant
role in plume modeling.
Currently existing 3-D models focus on certain aspects of the volcanic plume
(PDAC on pyroclastic flow, ATHAM on microphysics and SK-3D on entrainment
with higher resolution and higher order of accuracy), and hence, naturally,
different assumptions were made in these models. However, these different
aspects of volcanic plumes are not independent but are actually coupled. For
example, it has been illustrated by that
gas–particle non-equilibrium would introduce a previously unrecognized
jet-dragging effect, which has great influence on plume development,
especially for weak plumes. In addition, there is no absolute boundary to
determine which kind of hazard is dominant in certain eruptions. So it is
necessary to simulate all associated hazards in one model. Actually, effort
has already been put into developing more comprehensive plume models. For
example, a large-particle module (LPM) was added to ATHAM to track the paths
of rocky particles (pyroclastic or tephra) within the plume and predict where
these particles fall . We were also motivated by such
an evolution of plume modeling to choose SPH as our numerical tool. Besides
its ability to deal with interfaces for multiphase flows, as mentioned in
the introduction section, the SPH method has good extensibility and adding new
physics and phases requires much less modification of the code compared with
mesh-based methods. Last but not least, the dramatic development of
computational power makes it possible to establish a comprehensive model.
While current computational capacity may not allow us to have a fully
comprehensive model, the easy-extension feature of SPH makes it convenient to
keep adding new physics into the model when necessary and computationally
feasible.
We have presented in this paper an initial effort and results towards
developing a first principle-based plume model with comprehensive physics,
adopting proper numerical tools and high performance computing. More advanced
numerical techniques, such as adaptive particle size, Godunov-SPH,
semi-explicit time-advancing schemes and better data management strategies and
algorithms are on our list to exploit in the future. In the near future,
the effect of wind field will be taken into account. Our code will also be made
available in the open-source form for the community to enhance. Besides
improving the plume model, coupling the volcanic plume model with magma reservoir
models e.g.,, which could provide more accurate
eruption conditions, might improve accuracy of the volcanic plume simulation.