JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 112, B12205, doi:10.

1029/2007JB004985, 2007
Click
Here
for
Full
Article

Thermal effects during magma ascent in conduits

A. Costa,1,2 O. Melnik,2,3 and E. Vedeneeva3
Received 14 February 2007; revised 30 July 2007; accepted 4 October 2007; published 27 December 2007.

[1] Because of strong coupling between viscosity and temperature, the dynamics of
magma flows in conduits are drastically controlled by thermal effects due to heat
generation by viscous dissipation and loss to the walls by conduction. Here we present
analytical solutions and a practical procedure based on an order of magnitude analysis
that permits the characterization of the regime and estimation of the main features of the
flow. The ranges of validity of analytical and asymptotic solutions were bounded by
using results from fully two-dimensional (2-D) numerical solutions of mass, momentum,
and energy equations for magma flow inside a cylindrical conduit and the heat
conduction in the surrounding host rocks. The results permitted the identification of
three regimes: a conductive-heat-loss-dominated regime, an intermediate regime, and a
viscous-heating-dominated regime. Some useful analytical parameterizations are
proposed for estimating friction in simplified 1-D models. Temperature layering due to
heat loss by conduction can lead to local crystal growth and magma solidification
whereas heat generated by viscous dissipation can be responsible for crystal resorption
and remelting of wall rocks.
Citation: Costa, A., O. Melnik, and E. Vedeneeva (2007), Thermal effects during magma ascent in conduits, J. Geophys. Res., 112,
B12205, doi:10.1029/2007JB004985.

1. Introduction for the review of existing models). In order to model

conduit friction in a one-dimensional (1-D) approach, the
[2] The dynamics and thermodynamics of viscous fluids
velocity profile inside the conduit is assumed to be para-
with a strongly temperature-dependent viscosity, such as
bolic as in the classical Poiseuille solution for a pipe or a
magmas, cannot generally be described neglecting coupling
slab flow [e.g., Batchelor, 1967]. This solution is only valid
between the equations for momentum balance and the
for an incompressible laminar flow of a liquid with constant
equation describing energy conservation. Temperature var-
viscosity. In the case of magma, viscosity (even if magma is
iations in ascending magma can be produced by many
assumed to be a Newtonian liquid) depends on temperature,
physical processes including heat exchange with wall rocks
composition, volatile content, crystallinity, strain rate and
and hydrothermal systems, by viscous frictional heating, by
volume concentration of bubbles. Thus even for isothermal
release of latent heat of crystallization, and by cooling due to
flows the validity of a parabolic velocity profile needs to be
expansion of volatiles and chemical reactions. The amount
justified and may be violated in many flow conditions.
of heat that is generated or lost strongly depends on the flow
[4] There is increasing evidence for significant variations
conditions (velocity, pressure and magma composition dis-
in magma properties across volcanic conduits. For example,
tribution inside the conduit) and on the thermal state of the
the presence of grey and white pumices with different
surrounding rocks and hydrothermal systems. Changes in
thermal histories [Polacci et al., 2001; Rosi et al., 2004]
temperature result in changes in the rheological properties of
has been documented for many volcanic eruptions at
magma (by variations due to the temperature itself or
various different volcanoes. Tuffen et al. [2003] discovered,
through the change in crystallinity and chemical composi-
in addition, complicated layered structures in contact zones
tion of the magma). These changes lead to alteration in the
in rhyolitic dikes with evidence of shear melting and
velocity profile and may introduce positive feedbacks in the
fracture healing. Cooling manifests itself by the presence
system that can result in various instabilities.
of chilled margins commonly observed in eroded contacts.
[3] Most conduit flow models that have been developed
[5] A few conduit flow models incorporate cross-conduit
for description of magma flow are one dimensional and
temperature variations caused by solidification and heat
assume isothermal ascent conditions (see Sahagian [2005]
transfer to wall rocks [Delaney and Pollard, 1981, 1982;
Huppert and Sparks, 1989; Bruce and Huppert, 1990;
1
Osservatorio Vesuviano, Istituto Nazionale di Geofisica e Vulcanologia, Fialko and Rubin, 1999; Petcovic and Dufek, 2005] and
Naples, Italy. by shear heating [Hardee and Larson, 1977; Nelson, 1981;
2
Centre for Environmental and Geophysical Flows, Department of Costa and Macedonio, 2003, 2005; Vedeneeva et al., 2005;
Earth Sciences, University of Bristol, Bristol, UK. Mastin, 2005]. The range of assumptions in these models is
3
Institute of Mechanics, Moscow State University, Moscow, Russia.
wide, starting from constant viscosity flows, through quasi-
Copyright 2007 by the American Geophysical Union. 1-D models for variable viscosity fluids to fully 2-D
0148-0227/07/2007JB004985$09.00 simulations. These models show that temperature distribu-

B12205 1 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS
tion and history of the flow can be very complex. In the case
of temperature-dependent viscosity, the velocity profile can
evolve from parabolic in the entry region to plug-like as the
thermal boundary layers start to develop. This change has a
strong influence on the conduit friction. In the case of
isothermal flow, the friction increases as magma ascends
due to increase in magma velocity and viscosity due to
volatile exsolution. Formation of a thin zone of hot, low-
viscosity magma near conduit walls leads to a decrease in
friction during ascent. Overall friction decreases, and there-
fore the discharge rate predicted by nonisothermal models is
higher than for isothermal flow conditions. If heat flux to
wall rocks is taken into account, wall temperatures can be
nonmonotonic in space and time.
[6] From an experimental point of view, there have been
very few investigations of the effects of viscous heating on
magma flows. Recent results obtained by Hess et al. [2006],
using a uniaxial high-temperature, high-load apparatus with
calc-alkaline magmas, clearly showed that internal viscous
dissipation can be responsible for both temperature
increases and the apparent onset of non-Newtonian flow
behavior.
[7] Similar problems are encountered in industry for
polymer and oil flows. Most polymer processing methods
involve heating and cooling phases. As in silicate melts,
polymers have a low thermal conductivity and a strongly
temperature-dependent viscosity. Local temperature varia-
tions, even confined in very thin layers, can cause degra-
dation and many undesirable effects. For these reasons,
general processes involving heat transfer and heat genera-
tion by viscous friction have been extensively investigated
using experimental [e.g., Sieder and Tate, 1936; Sukanek
and Laurence, 1974; Skul’skiy et al., 1999], analytical [e.g.,
Pearson, 1978; Richardson, 1983, 1986a, 1986b] and
computational [e.g., Lin and Jaluria, 1996; Wichterle,
2004] approaches. Because high shear rate viscosity meas-
urements of engine oils can be affected by viscous heating,
some semiempirical corrections (valid in a limited range)
were proposed [Lodge and Ko, 1989; Ko and Lodge, 1991]
for viscosimeters. Recent studies also show that viscous
heating has a pivotal role in interpreting controversial
experimental data for liquid flows in microchannels [e.g.,
Morini, 2006].
[8] In this article we will study the thermal budget of
ascending magma resulting from convective and conductive
heat transfer and viscous heat production. We will specify
the controlling dimensionless parameters and describe as-
ymptotic behavior of the system at different end-members.
Wherever possible, we will present corresponding analytical
solutions and recipes to estimate main flow parameters
without the solution of fully 2-D transport and energy
equations. The primary dimensionless groups are estimated
for typical eruptive flow conditions. Several implications
for magma flows such as temperature and strain layering,
increases and decreases in friction factor, and crystal
resorption are also discussed.
2. Governing Equations and Dimensionless
Numbers
[9] As a first-order approximation, with the aim of
evaluating the role of heat transfer, magma can be modeled
2 of 16
B12205
as an incompressible and homogeneous fluid with constant
density, constant specific heat and constant thermal conduc-
tivity. We will also restrict our investigation to flows in
simple geometries such as cylinders or slabs. Let us con-
sider a vertical conduit of length L and diameter D (or
thickness H for a slab flow) with D/L
1.Under these
assumptions, fluid dynamics and heat transfer are described
by the following transport equations for mass, momentum
and energy, respectively:
rv¼0 ð1Þ
@v
r þ rv  rv ¼ rP þ rg þ r  t ð2Þ
@t
@T
rc þ rc v  rT ¼ kr2 T þ t : rv ð3Þ
@t
where r is the fluid density, v is the velocity vector, g
represents the gravitational force per unit of volume, P is the
pressure, t is the stress tensor, T is the temperature, c is
the specific heat, and k is the bulk thermal conductivity. The
term containing the stress tensor t in equation (3) represents
an internal heat generated by the viscous dissipation (the
symbol: indicates the inner product between two symmetric
second-rank tensors). Latent heat release due to crystal-
lization is here accounted for by assuming an equilibrium
crystallization process. An effective specific heat c = cp +
L*/(TL  Ts) accounts for the release of latent heat for the
case when the crystal fraction changes linearly with
temperature b = (TL  T) (TL  Ts) and temperatures of
liquidus TL and solidus Ts are constant (here cp denotes the
magma specific heat and L* is the latent heat of
crystallization). Thus we neglect here the decompression-
induced crystallization. For the sake of simplicity, we
adopt a Newtonian relationship between stress tensor and
strain rate although the results can be generalized:
h i
t ¼ m rv þ ðrvÞT ð4Þ
where the superscript T denotes the transposition.
Viscosities of silicate melts strongly depend on tempera-
ture. Temperature viscosity dependence is commonly
described by either using an Arrhenius relationship or
more adequately adopting semiempirical extensions of it
such as the Vogel-Tammann-Fulcher (VTF) parameteriza-
tion [e.g., Hess and Dingwell, 1996; Russell et al., 2003]:
 
A
m ¼ mVTF exp ð5Þ
T C
where m is the shear viscosity, mVTF is the viscosity value
at infinite temperature, A corresponds to a pseudoactiva-
tion energy parameter, and C is the temperature at which
viscosity tends to infinity. All adjustable parameters in (5)
depend on magma composition [e.g., Hess and Dingwell,
1996], and especially on volatile content. In the context of
an order of magnitude analysis, and when the typical range
of temperatures is not large, an exponential approximation
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

is more suitable for analytical manipulations [e.g., Richardson, 1986a; Costa and Macedonio, 2005]
Pearson, 1977]:
 2
@T k @ @T @vx
r c vx ¼ r þm ð9Þ
m ¼ mR exp½bðT  TR Þ
ð6Þ @x r @r @r @r

where mR = m(TR) denotes the shear viscosity value at the The overall pressure drop
reference temperature TR and Z L
dP*
DP ¼  dx ð10Þ
  0 dx
1 @m
b¼
m @T TR and the convected mean temperature

represents the rheological sensitivity to the temperature. Z R

[10] Equations (1) – (4) are solved with the following
boundary conditions: at the walls we assume a no-slip
velocity, whereas at the conduit inlet we consider that the
fluid enters with a mean velocity V and a parabolic profile,
although the shape of the velocity profile at the inlet has a
small influence on the flow conditions after a short (order of
a few D) entry length region. At the conduit outlet, a
reference pressure and zero radial velocity are specified.
For the temperature, we assume that fluid enters with a
uniform constant temperature T0. At the conduit walls, we
consider both a prescribed vertical temperature profile Tw
for a first-order analysis or more realistic boundary con-
ditions that are described in the section 4 where results of a
numerical solution of equations (1) –(4) are presented.
[11] Below, we present the approximated equations for an
order of magnitude analysis. In order to make preliminary
estimates on the thermal budget during magma ascent we
will employ an approach similar to that used by Pearson
[1978] for polymer flows. For sake of simplicity in this
2p
T ¼ rvx Tdr ð11Þ
Q 0
will be used in further analysis. The convected mean
temperature (11) is also commonly used as primary variable
in the energy equation of 1-D conduit models [e.g., Costa and
Macedonio, 2002; Melnik and Sparks, 2005; Costa et al.,
2007b].
[12] The only relevant velocity and length scales in
equations (7) and (8) are V and R, whereas equation (9)
involves two natural length scales, R for the conductive
term and L for the convective term, respectively. The wall
temperature Tw, the inlet temperature T 0 (with the
corresponding shear viscosity m0) and the following tem-
perature scales DT* are relevant in different flow regimes
[Pearson, 1978]:
DTbc ¼ jT0  Tw j ð12aÞ
 
1  m 
section we will restrict most of our considerations to a DTrheol  ¼ ð12bÞ
cylindrical conduit, but results can be easily reformulated to b @m=@T T ¼TR
the case of a slab geometry. Assuming that vertical varia-
tions of vertical velocity vx are much smaller than radial 
because L/R  1 (R = D/2, denotes the conduit radius) we mV 2 
DTvisc ¼ ð12cÞ
will find the solution with v = (vx(r),0)T where superscript T k T ¼TR
means transposition. Equation (1) for mass conservation is
written in an integral form The first temperature scale (12a) is given by the difference
between the inlet and wall temperatures; the second (12b) is
Z R the temperature scale at which rheological property changes
Q ¼ 2p rvx dr ð7Þ occur (TR is a reference temperature); the third temperature
0
scale (12c) is given by a balance between the viscous
dissipation term mV2 / R2 and the heat conduction term
and requires that volume flux Q  pR2V is constant. kDTvisc/R2.
Hereafter V = Q/(pR2) will denote the mean velocity. In the [13] Rewriting equations (7), (8), and (9) in dimension-
caseR of a slab flow with a slab thickness H, we have Q = less form and assuming for simplicity the relationship (6)
H
W 0 vxdy where W is the slab width and V = Q / (WH) for the viscosity, we obtain
Under typical conditions of low Reynolds number flow (Re =
rVR/m, based hereafter on the conduit radius R instead of the Z 1
diameter D = 2R) with Pr  1 and R/L
1 (here Pr = mRcp/k 2p u hdh ¼ 1 ð13Þ
0
is the Prandtl number), a lubrication approximation for
equation (2) is fully justified [Pearson, 1977; Costa and  
Macedonio, 2003, 2005; Vedeneeva et al., 2005]: 1 @ @u
heSq ¼ P* ð14Þ
  h @h @h
1 @ @vx @P*
rm ¼ ð8Þ
r @r @r @x
 2
@q 1 @ @q @u
where P* = (P  rgx) and m = m(T). Keeping only the leading uGz ¼ h þ Gr eSq ð15Þ
@x h @h @h @h
terms, equation (3) can be approximated by [Pearson, 1977;

3 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

where the dimensionless variables are introduced corresponding Nahme number based on this temperature
 2
x r vx ðT  T0 Þ R @P* =@x
x¼ ; h¼ ; u¼ ; q¼ ; P* ¼
L R V DT* mR V
ð16Þ
and the dimensionless numbers are
rcVR2 m V2
Gz ¼ ; Gr ¼ R ; S ¼ bDT* ð17Þ
kL kDT*
Gz denotes the Graetz number (i.e., the Péclet number times
the aspect ratio Gz = Pe R/L ), which represents the ratio of
heat convection in the longitudinal direction to heat
conduction in the radial direction. When Gz is very large
the temperature field is developing, that is axial convection
dominates over radial conduction everywhere in the flow
except for a thin thermal boundary layer near the conduit
walls. Here Gz is based here on the conduit radius R instead
of the diameter D = 2R. The viscous heating number Gr
represents a ratio between the viscous temperature (12c) and
the typical temperature scale DT*. Gr is known as Nahme-
Griffith number Na when DT* = 1/b i.e., Na = bmRV2/k,
and Brinkman number Br when DT* = DTbc, i.e., Br =
mRV2/(k(T0  Tw)). When Na is large, the temperature
increase due to heat generation by viscous dissipation is
sufficiently large to change the viscosity significantly. The
third dimensionless number S is given by the ratio between
the temperature scale DT* and the rheological temperature
(12b). When DT* = DTbc we will denote the number S
using the symbol B, i.e., B = bDTbc When B is small the
difference between the inlet and the wall temperatures is not
sufficiently large to cause a significant variation in viscosity.
The choice of the most convenient temperature scale and
reference temperature depends on the flow regime, i.e., on
the values of Gz and Na [Pearson, 1978].
3. Flow Regimes
[14] As we explain in the section 3.5, magma flows are
generally characterized by a low Reynolds number and a
very large Prandtl number. Typically, we have large Graetz
Na = Naw  bm(Tw)V2 / k. If the wall temperature is
assumed to be constant the flow remains isothermal.
3.1.2. Na = O(1)
[17] In this intermediate regime, viscous heating introdu-
ces a coupling between the energy and momentum equa-
tions but the choice of TR = Tw is still adequate. Because of
intensive heat conduction the temperature across the conduit
remains nearly constant.
3.1.3. Na  1
[18] Here the viscous heating effects are dominant and the
choice of Tw as a reference temperature is not valid since it
leads to an overestimation of mR. A better reference tem-
perature TR, determining a more representative shear vis-
cosity, consists of considering T* = Tw + Na*/b with Na* =
bm(T*)V2/k where b = bjT=T*. These relations implicitly
define TR = T*. For example, using the rheological relation-
ship (6) we obtain Na* = NaweNa*, which allows us to
estimate Na* in an iterative way by solving a nonlinear
equation where initially b can evaluated at T = Tw. In this
regime one can expect a large increase in magma temper-
ature in regions of high shear. Convective and conductive
heat transfers are not efficient so that the temperature
increase will be localized.
3.2. Gz = O(1)
[19] In this regime, the effects of conduction and convec-
tion are broadly comparable. All the temperature scales (12)
can be important. Viscous heating, convection and conduc-
tion effects can be equally important. Numerical solutions
are necessary.
3.3. Gz  1
[20] This regime is typical for most magma flows. In this
case the flow cannot be considered as thermally developed.
Heat conduction to the walls in this regime will only affect a
narrow boundary layer within the conduit. Depending on
the values of Gz and Na the heat generated by viscous
dissipation can be negligible, comparable, or dominant with
respect to the heat loss by conduction. A convenient choice
for this regime can be TR = T0 and DT* = DTrheol [Pearson,
1978], so that Na represents the so-called Nahme-Griffith
numbers and hence we could restrict our analysis to
regimes having Gz  1. However, for a complete descrip-
tion, we will briefly discuss below the cases with Gz
1
and Gz = O(1).
3.1. Gz
1
[15] In this regime the conduction term is dominant with
respect to the convective term. A fully developed thermal
profile and velocity flow will occur over most of the conduit
length. This condition is not typical for magma flow,
although in principle it can occur in the case of very slow
and very thin flows in dikes, and can thus be important in
the analysis of analogue experiments. For the temperature
scale, the most natural choice is DTrheol with m evaluated at
a suitable temperature TR.
3.1.1. Na
1
[ 16 ] In this case the dynamical equations can be
decoupled from the energy equation. Effects of viscous
heating are negligible. The most suitable reference temper-
number, Na = Na0  bm(T0)V2/k where b = bjT=T0. There are
three different cases that can now be relevant for magma
flows.
3.3.1. Na
1
[21] In this case, viscous heating can be neglected but if
the value of B is large enough, i.e., if the wall rocks were
significantly cooler than the magma, then equations (8) and
(9) cannot be decoupled, since cooling by conduction can
have an important role in the dynamics [e.g., Costa and
Macedonio, 2002, 2003, 2005].
[22] A simplified 1-D solution can be used to estimate the
wall temperature as a function of the far-field temperature
Tff, the average fluid temperature T , and the Nusselt number
NuD (see Appendix A for details):
Tff þ 2pNuD Ea ðt Þ T
Tw ¼ ð18Þ
1 þ 2pNuD Ea ðt Þ
where Ea is a universal function of the dimensionless time
ature is the wall temperature, i.e., TR = Tw and the t  kt/(rcpR2). Fluid temperatures near the wall can be

4 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Table 1. Parameters Used in the Simulations and Typical Ranges for Different Eruptive Stylesa
Values
Symbol Description Lava Flow Lava Dome Explosive Reference
R conduit equivalent radius (m) 0.1 – 15 10 – 50 (15) 10 – 50 1,3,5,7
L conduit length (km) 2 – 10 2 – 10 (5) 2 – 10 1,3,5,7
r density of magma (kg m3) 2500 – 2800 2200 – 2600 (2500) 2000 – 2300 8
rr density of host rocks (kg m3) 2000 – 2800 2000 – 2800 (2600) 2000 – 2800 8
Q discharge rate (m3 s1) 1 – 104 102 – 10 103 – 106 2,9
b0 initial crystal concentration (vol %) 0 – 10 20 – 60 (50 – 60) 0 – 60 5,6,8
w0 initial water concentration (wt %) 0.1 – 1 1 – 7 (2.5) 1 – 10 4,5,7,9
L* latent heat of crystallization (J kg1) 2 – 8  105 2 – 8  105 (3.5  105) 2 – 8  105 1,5,8
TL liquidus temperature (°C) 1200 – 1300 1000 – 1100 (1070) 1000 – 1100 5,6,8
T0 magma chamber temperature (°C) 1000 – 1200 700 – 1000 (850) 700 – 1000 1,5,6,8
m0 melt viscosity at T0 (Pa s) 10 – 105 106 – 1015 (2  106) 103 – 109 4,5,6,8
b rheological parameter (K1) 0.01 – 0.2 0.01 – 0.1 (0.03) 0.01 – 0.1 4,6,8
Ts solidus temperature (°C) 900 – 1000 600 – 800 (800) 600 – 800 5,6,8
cp specific heat (J kg1 K1) 1400 – 1500 1200 – 1400 (1200) 1200 – 1400 8
k thermal conductivity (W m1 K1) 1.0 – 3.0 1.0 – 3.0 (1.0) 1.0 – 3.0 8
a
Simulation values are given in parentheses. References: 1, Bruce and Huppert [1990]; 2, Carey and Sigurdsson [1989]; 3, Costa et al. [2007b]; 4, Hess
and Dingwell [1996]; 5, Melnik and Sparks [2005]; 6, Shaw [1969]; 7, Stasiuk et al. [1993]; 8, Touloukian et al. [1989]; 9, Wadge [1981].

significantly lower than in the core of the flow. If B is field. That means that the convective term will dominate
large, near the walls the viscosity is relatively large and over the conductive term.
may therefore drastically affect the velocity profiles and [26] In a way similar to that of Pearson [1978], we
thus the friction. This can also lead to significant propose a practical procedure that permits one to decide
corrections for the Nusselt number, i.e., heat exchange the relevant quantities and the appropriate regime for the
coefficient (see Appendix A). flow under study. The principal steps or stages are as
3.3.2. Na = O(1) follows:
[23] Viscous heating cannot be neglected in this case. The [27] 1. Estimate a transversal and a longitudinal length
energy and momentum equations now become strongly scale, R = D/2 and L, respectively. R will be the radius for a
coupled. For B
1, DTbc becomes irrelevant because the cylindrical geometry (or the half distance between the walls
flow temperature quickly deviates from the initial temper- H/2 for a slab), and L will be the conduit length.
ature. For larger B the solution is complex. Numerical [28] 2. Estimate the quantity
solutions are thus necessary. 8
3.3.3. Na  1 < rcp Q=ðpkLÞ if Q is specified
[24] Viscous heating effects are dominant. The choice of Gz ¼ ð20Þ
:
TR = T0 as a reference temperature can yield an overestima- rcp DPR4 =ð16m0 kL2 Þ if DP is specified
tion of mR. Similarly to the case when Gz
1, a better
estimation of TR can be obtained, considering T* = T0 + Na*/b This value allows one to decide whether the flow will be
with Na* = bm (T*)V2/k where b = bjT=T*. These relations likely a thermally developing (Gz  1) or fully developed
implicitly define TR = T*. For instance, using the rheological flow (Gz
1). This dimensionless group permits one to
relationship (6), we obtain Na* = Na0eNa* from which we understand the importance of the convection term with
estimate Na* then T* = T0 + Na*/b , where initially b can respect to the conduction term. This procedure can thus help
evaluated at T = T0. This implies that ln(Na) is the relevant to introduce considerable simplifications.
parameter to estimate heat generation DTgen = Tmax  T0. [29] 3. Calculate the quantity
Using asymptotic solutions [Pearson, 1978; Richardson,
1986a], we can estimate some relevant quantities: 8
< bmR Q2 =ðp2 R4 k Þ if Q is specified
Na ¼ ð21Þ
Tmax  T0  lnð NaÞ=b; :
bDP2 R4 =ð256mR kL2 Þ if DP is specified
 pffiffiffiffiffiffiffiffiffiffiffi
T  T0  glnð NaÞ= b 4pGz ; ð19Þ where mR and b can initially be estimated at some
 pffiffiffiffiffiffiffiffiffiffiffi representative temperature, for instance at T = 1/2(Tw +
DP  2g rc lnð NaÞ= b 4pGz
T0), then the estimation can be improved using a more
representative temperature. Values of the Na number permit
where g = 8 for a pipe and g = 4 for a slab geometry the user to decide whether there is strong heat generation
[Richardson, 1986a]. For typical values of b and Gz (see due to viscous heating or not. If Na is large enough, we can
Table 1), the maximum temperature may be much larger use analytical relationships (19) to estimate some relevant
than the mean temperature. quantities.
3.4. Order of Magnitude Analysis: A Procedure [30] 4. Finally, we estimate
Summary
[25] As mentioned above, for most magma flows we can B ¼ bðT0  Tw Þ ð22Þ
generally consider Gz  1 at least over most of the flow

5 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS
Table 2. Typical Dimensionless Parameter Ranges
Symbol Description Lava Flow
1 rcQ
Gz = Graetz number 10 – 106
p kL
2
Na0 = p12 bmkR0 Q4 Nahme number 10 4
– 10 8
Pr0 = mk0 c Prandtl number 3
10 – 10 8 8
Re0 = p1 mrQR Reynolds number 10 3
– 10 5
0
S = pNaffiffiGffi0z regime number 102 – 102
with b calculated at some representative temperature, such
as T = 1/2(Tw + T0). This allows one to understand whether
there will be a strong coupling between the momentum and
the energy equations. From a practical point of view, for
magma flows it can be difficult to estimate Tw, but if Na is
small enough, i.e., viscous heating is believed to be
negligibly small, a simplified approach can be adopted
and Tw can be estimated using the relationship (18) (see
Appendix A). In this case, if B is small there will be no
feedback influence of temperature variation on flow
profiles. The momentum equation must be solved first to
evaluate the velocity profile. Then the energy equation can
be solved to derive the temperature distribution.
3.5. Application to Magmatic Flows in Conduits
[31] In order to link the theory described above with
volcanic flows inside the conduit we present here typical
values of the dimensionless parameters. We now need to
distinguish the style of activity (explosive or effusive/
extrusive) and magma composition (mafic or silicic). In
the case of explosive eruptions, our analysis is applicable
only for the homogeneous or the bubbly liquid part of the
flow. In gas-particle dispersion flows, both the heat transfer
to surrounding rocks and the viscous heating are negligibly
small due to short times that magma parcels reside in the
conduit after fragmentation and the low viscosity of gas
phase. The temperature variation there is controlled by an
interplay between the cooling due to gas expansion and the
heat exchange with particles. Within an accuracy of several
percent, the flow can be considered isothermal. We will use
standardized values, suggested by Sahagian [2005] and
Carey and Sigurdsson [1989], for calculations of the typical
dimensionless parameters for explosive eruptions. For the
case of lava dome eruptions typical values are used follow-
ing Melnik and Sparks [1999, 2005] and Stasiuk et al.
[1993]. Magma discharge rates during basaltic eruptions are
mainly estimated from Wadge [1981]. Other property ranges
are estimated from Touloukian et al. [1989] and Shaw
[1969]. Ranges of dimensional parameters are listed in
Table 1. Calculated dimensionless numbers are listed in
Table 2.
[32] As one can see from Table 2, magma flows are
generally characterized by relatively low Reynolds numbers
(except for some cases of very intense eruptions) and very
large Prandtl numbers. Thus the main flow regime in a
volcanic conduit is laminar and lubrication approximation,
equations (8) and (9), can be appropriately applied. How-
ever, in the case of high-magnitude eruptions, Re can easily
B12205
reach values of a few orders of magnitude greater than 103,
Values which means that the flow cannot then be assumed laminar.
Lava Dome Explosive
[33] For most conditions, we have Gz  1.Thus heat
convection in the longitudinal direction dominates over heat
101 – 104 104 – 108
conduction in the radial direction. The Nahme number can
105
– 10 9
10 – 1013 vary from very small to very large values in both the lava
10 – 10 18
105 – 1012 flow and the lava dome eruptive styles whereas explosive
1015
– 10 3
10 – 105 eruptions are generally characterized by Na  1. This
104 – 105 1 – 1011 means that for both lava flow and lava dome eruptive styles
we can have both regimes dominated by conductive heat
loss and regimes dominated by viscous heating effects. On
the other hand, magma flows during explosive eruptions are
generally characterized by strong viscous dissipation
effects.
4. Results From 2-D Models: Validation and
Limits of the 1-D Approach
4.1. Two-Dimensional Simulation of Magma Flow
With Heat Transfer
[34] In this section we present simulation results obtained
from fully 2-D numerical solutions of the full equation
system (1) – (4) considering a cylindrical geometry, i.e., a
pipe flow (see Figure 1 for the schematic view of the
calculation domain and boundary conditions). The calcula-
tions were performed using Comsol Multiphysics1 finite
element solver. We investigate conditions typical for magma
flows during lava dome eruptions (see Table 1 for the list of
parameters). In order to compare results of fully 2-D
simulations with theoretical models we have approximated
the viscosity equation presented by Hess and Dingwell
[1996] in the form of equation (6). Because in the work
by Hess and Dingwell [1996], the viscosity depends not
only on temperature but on the magma water content we
have used a medium value of water content w0 of 2.5 wt %
because inside the conduit the water content typically
changes from 4 –6 wt % at depth to a very small value at
the surface. However, gas exsolution increases viscosity in
ascending magma, and that should be accounted for, espe-
cially for more silicic magmas. The assumption of constant
water content is made only for computational reasons in
order to make the problem tractable. Magma viscosity also
strongly depends on crystal content increasing by a factor of
100 at about 55% of crystals and up to 104 at 68% of crystals
[Costa, 2005; Costa et al., 2007a, 2007b] (further com-
ments available at http://arxiv.org/abs/physics/0512173).
We will use a parametric correction factor Q representing
an average crystal content in ascending magma and
neglecting, for comparative purposes, the vertical and
radial variations in viscosity due to changes in the crystal
content.
[35] In order to avoid imposing arbitrary thermal bound-
ary conditions at the conduit walls, we solve the mass,
momentum and energy equations for magma flow inside a
cylindrical conduit and the heat conduction equation in the
surrounding host rocks imposing local far-field conditions
for the rock temperature at a distance 10  R. The outer
boundary of the calculation domain is selected in a way that
variations of temperature near this boundary due to heating
of wall rocks are negligibly small for the whole time of the
simulation. The far-field temperature profile, Tff , is assumed
to vary linearly from the magma chamber temperature, Tch,
6 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 1. Sketch of the studied system. Inside the conduit of radius R and length L we solve the
transport equations for mass, momentum, and energy. At the inlet we assume a parabolic profile and a
constant temperature equal to the magma temperature Tch. At the outlet we assume free flow condition.
Along the wall we impose a no-slip condition for the velocity, whereas the wall temperature is given by
the solution of the conductive heat equation inside the host rock solved considering, as far-field
conditions (at a distance 10 R), a linear profile from Tch to a constant environmental temperature Tenv.

at a depth L to an environmental temperature, Tenv, at the
surface. Magma solidification and the melting of country
rock are neglected. We solve the Navier-Stokes equations
with the following boundary conditions: no-slip conditions
are assumed for the velocity at conduit walls, the pressure at
the top of the conduit is set equal to zero, whereas at the
bottom of the conduit a fixed pressure or discharge rate
(assuming parabolic velocity profile) are specified. In the
case of fixed discharge rate the pressure at the bottom of
the conduit can therefore vary according to the variation in
the friction along the entire conduit. For the fixed inlet
pressure, discharge rate will vary with time. The inlet
temperature is additionally assumed to be constant and
equal to the magma chamber temperature, i.e., T0 = Tch.
We assume at time t = 0 the temperature in the entire
conduit is equal to its magma chamber value, initial velocity
or pressure are determined by the Poiseuille solution. These
initial conditions are artificial because the temperature and
velocity distributions inside the conduit and in the wall
rocks depend on preeruptive processes (dike propagation or
the presence of a permanently open conduit, for example).
Results of simulations show that the system quickly devi-
ates from these artificial initial conditions to a state that is
governed by the overall thermal balances. Thus the influ-
ence of initial conditions is restricted to the initial stage of
the calculations (usually first 1 – 2 years, i.e., t < 0.05– 0.1).
The sensitivity of the system to initial conditions is outside
the scope of the paper.
4.2. A Fixed Pressure in the Magma Chamber Case
[36] Here we will present results of the simulations when
a fixed pressure is assumed as a boundary condition at the
conduit inlet. In this case discharge rate will be determined
by an overall friction of the conduit and can vary with time
as thermal conditions in ascending magma change.
[37] Figure 2a shows variations of discharge rate (nor-
malized by a discharge rate calculated from the Poiseuille
solution Qpois) with dimensional time t. Figure 2b repre-
sents cross-section-averaged temperature at the top of the
conduit hT i (more precisely (hT i  T0)/ T0) as a function of
time. Viscosity correction factor Q is equal to 10 for the
dashed lines and to 100 for the solid. At the initial stage of
the process when host rocks are cold and heat loss is
significant the temperature inside the conduit decreases
significantly leading to an increase in magma viscosity
and thus conduit resistance. Discharge rate consequently
decreases. In the case of low chamber pressure (curve 1)
decrease in discharge rate leads to a significant reduction of
heat advected by magma, progressive cooling and increase
in viscosity. Finally, discharge rate drops to very low values
and the upper part of the conduit cools down by conduction.
At higher chamber pressures, discharge rate reaches a
minimum and starts to increase. This leads to an increase
in heat advection and of the overall temperature increase
inside the conduit. Decrease in viscosity leads to increase in
discharge rate and viscous heating starts to contribute to a
temperature increase. Finally, host rock temperature in the
vicinity of the conduit increases to magmatic temperatures,
heat flux to wall rocks decreases and the system stabilizes
with a discharge rate that is higher than Qpois, and almost
adiabatic wall boundary conditions for the temperature. The
time and the absolute value of the minimum in discharge
rate depend on the intensity of advection of heat and viscous
dissipation. For higher-viscosity magmas, viscous dissipa-
tion will start to contribute to temperature increase at
smaller discharge rates.
[38] Calculations with the fixed pressure boundary con-
dition are computationally more difficult to perform in
comparison with the case when the influx rate into the
conduit is specified. Since after a relatively short transient
period (less than 10 years), discharge rate for pressure BC
tends to its asymptotic value, we performed most of
simulations for the case of fixed discharge rate BC.
Comparison of calculation results with theoretical predic-
tions will be done in section 5 for the case of fixed
discharge rate BC.

7 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 2. Variations of discharge rate (normalized by a discharge rate calculated from the Poiseuille
R R time t (Figure 2a). Cross-section-averaged normalized temperature at
solution Qpois) with dimensional
the top of the conduit hTi = 2 0 rTdr/R2 versus t (Figure 2b). Viscosity correction factor Q is equal to 10
for the dashed lines and to 100 for the solid. Qpois is calculated from Poiseuille solution for magma with
the initial temperature T0: Qpois=pDPR4 / 8m0QL The value of DP is 10, 30, 70, and 80 MPa for curves
1 – 4 and 10, 20, and 70 MPa for curves 5 – 7. The value of m0 is 2.1  106 Pa. Using reference
dimensional parameters, Qpois = 1.89 DP/Q.

4.3. Case of Fixed Influx Into the Conduit
[39] Below we will present results of fully 2-D simula-
tions for the case of fixed influx of magma into the conduit.
As described in section 3.5, for typical conditions and
parameters that characterize magma flows (see Tables 1
and 2), the Graetz number is very large, i.e., we can restrict
our considerations to regimes having Gz  1 only. Depend-
ing on the value of Na, viscous heating can either be
negligibly small or be an appreciable factor that determines
the temperature variations. Figure 3 shows the distribution
of dimensionless temperature ((T  T0)/T0) and dimension-
less velocity (vx /V) at the exit of the conduit for three
different typical regimes: (left) heat-loss-dominated regime,
i.e., low Na0, (middle) intermediate regime, and (right)
viscous-heating-dominated regime, i.e., large Na0. Curves
corresponding to dimensionless times t = 0.234, 0.467,
0.93, 2.34 and 4.67. There is a progressive increase in rock
temperature with increasepffiffiffiffiffiinffi t. Beside the number Na0 the
[41] In the case of large Nahme numbers, a thin layer of
hot magma is formed near the wall due to viscous dissipa-
tion effects. This layer is characterized by hot, low-viscosity
magma with high-velocity gradients. Instead of loosing
heat, magma in the central region of the conduit receives
heat from this hot layer. The surrounding rocks experience
temperatures that are higher than the initial temperature of
the magma; thus a melting of the wall rocks is now possible.
Because of the low viscosity of magma in the near-wall
zone, the profile of the velocity is flat with all shearing
occurring near the wall. The maximum velocity is smaller
than 2  V.
[42] In the intermediate regime, heat generated by viscous
dissipation is able to balance heat loss by conduction.
Temperature profiles inside the conduit are almost flat and
the velocity distribution very is similar to a parabolic
profile.
4.4. Temperature Corrections for 1-D Models
value of S  Na0 / Gz, useful for characterizing the
regimes (see below for details), is marked on the plots.
[40] In the case of low Na0 number, when cooling is
large due to a large temperature difference between the
magma and wall rock, the flow is localized in the central
part of the conduit. Near the walls the temperature is so
low that viscosity tends to very large values and the
velocity is almost zero. The maximum velocity is much
larger than 2  Va as required for a Poiseuille flow
solution. There is also a sharp transition between no-flow
and high-flow velocity domains where gradients of the
velocity are correspondingly large. These gradients lead to
high strain rates and therefore to a significant viscous
dissipation of heat. In the central part of the conduit, the
temperature is almost constant because the radial heat flux
is low due to the presence of a zone with elevated temper-
atures. As time progresses the temperature in the no-flow
region progressively increases, but at a very low rate
because in this region only conduction is responsible for
the temperature variations.
8 of 16
[43] We will use obtained results to validate and to define
the limits of validity of the asymptotic relationships pre-
sented in section 4.3 and also to estimate the deviation from
the classical Poiseuille solution for the conduit friction that
is usually adopted in 1-D models.
[44] As previously explained, when viscous heating is
dominant, i.e., large Na0, we use the inlet temperature as the
reference temperature, i.e., TR = T0, and we can adopt
formulas (19) to estimate maximum temperature and the
average temperature increase:
Tmax  T0  lnðNa0 Þ=b ð23aÞ
 pffiffiffiffiffiffiffiffiffiffiffiffiffi
T  T0  glnð Na0 Þ= b 4pGzx ð23bÞ
where (23a) expresses the upper limit for the maximum
temperature experienced in the domain (Na0 is based on the
values at the inlet and represents an overestimation of Na*)
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 3. (a) Temperature and (b) velocity radial profiles at the outlet for the three different regimes
(from p
left ffi right S  0.1 (Na0  1.6), S  0.6 (Na0  60), S  5.1 (Na0  125), respectively, where S 
ffiffiffiffiffito
Na0 / Gz). The intermediate case reported here has a large Gz number (about 8000). For lower values,
conductive loss is significantly important in the upper part of the conduit and velocity profiles are different
from the parabolic one. Curves 1 to 5 correspond to the values of t = 0.234, 0.467, 0.93, 2.34, and 4.67,
respectively (5, 10, 20, 50, and 100 years for the dimensional parameters used in simulations).

and in (23b) Gzx = Re0Pr0 Rx represents the local Graetz
number.
[45] Since equations (23a) and (23b) are based on the
inlet values (that are also used in 1-D models), in the limits
of their validity, these equations can be used for calculating
temperatures and estimating temperature variations the in
simplified 1-D approach. Solutions (19), (23a), and (23b)
were obtained in the limit Na/Gz  1 [Pearson, 1977] but
as we will discuss later, simulation results show that they
can be satisfactorily
pffiffiffiffiffiffi applied if the dimensionless quantity S
for characterizing the regime where Na numbers are rela-
tively low. However, from our simulation results we found
that we can distinguish the regime by evaluating S. We note
that the similarity variable introduced
pffiffiffiffiffiffi by Pearson [1977] is
also inversely proportional to Gz.
[46] On the other hand, when viscous heating is negligi-
ble, i.e., for low Nahme numbers, for large Graetz numbers,
we can still assume that the radial temperature variations are
much larger than the longitudinal and locally we can adopt
the 1-D model presented in Appendix A for estimating the
 Na0 / Gz is large enough. We found that values of the
dimensionless group S can be satisfactorily used to identify
the different regimes characterizing the flow. Obviously, a
parameter characterizing the flow regime can be a combi-
nation of Na and Gz, such as Na / Gz itself or other
combinations, but many features of the flow depend on
Na and Gz separately. For example, relationships (19) and
(23b)
pffiffiffiffiffiffi suggest the use of a number proportional to ln(Na) /
Gz. This will work well for very large Na numbers and for
the mean temperature estimation (see Figure 4b where two
different cases are shown). This approach will not be useful
9 of 16
wall temperature as function of the far-field temperature
profile and the local average temperature:
Tff ð xÞ þ 2pNuD ð xÞ Ea ðt ÞT ð xÞ
Tw ð x Þ ¼ ð24Þ
1 þ 2pNuD ð xÞ Ea ðt Þ
where NuD is the local Nusselt number based on the local
Graetz number Gzx and on the Sieder and Tate [1936]
variable viscosity correction L, and Ea is a universal
function of the dimensionless time t (see Appendix A).
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 4. Comparison between the analytical solutions and the 2-D computed. (a) Wall temperature
solution (24) valid in the heat-loss-dominated regimes (here case S  0.1 and Na  1.6. (b) Asymptotic
solution for the mean temperature (23b) reported for two different
pffiffiffiffiffifficases. Black line and black symbols
denote the case having S  5.1, Na0  125, Gz  60 ( In(Na)/ Gz  0.13), whereas pffiffiffiffiffiffi grey line and grey
symbols denote the case having S  3.6, Na0  60 and Gz  28, (with ln(Na) / Gz  0.24). Curves 1
to 5 correspond to the values of t = 0.234, 0.467, 0.93, 2.34, and 4.67, respectively (5, 10, 20, 50, and
100 years for the dimensional parameters used in simulations).

[47] As we mentioned above, results of simulations show
that we can distinguish three regimes that can be character-
ized by evaluating the dimensionless group
sffiffiffiffiffiffiffiffiffi
Na0 bm V 3L
S ¼ pffiffiffiffiffiffi ¼ 0
Gz R rkc
A first regime is viscous-heating-dominated and corre-
sponds to S ^ 2. In this case the relationships (23) can be
satisfactory applied and, although wall temperature cannot
be estimated analytically, we expect that the more important
the viscous heating, the less important the wall rock
conditions. A second regime is conductive-heat-loss-
dominated and corresponds to S ^ 0.2. In this regime we
can reasonably estimate the local wall temperature (24) and
heat exchange coefficient by the method described in
Appendix A. An intermediate regime corresponds to 0.2 ]
S ] 2. In this case the conduction is approximately
balanced by viscous heating and heat loss. In the
intermediate case, for practical purposes, one can assume
isothermal flow conditions.
[48] Figure 4 shows the comparison between the analytical
solutions (23) and (24) and the numerically computed values
in the two end-member cases, respectively, i.e., S
1 and
S  1. In Figure 4a, computed values of wall temperature
are represented by symbols, whereas the line corresponds
to the analytical solution for S
1 . In Figure 4b the
computed mean temperature is compared with the
corresponding asymptotic solution (line). In both cases
we have a reasonable agreement that it is expected to
improve for values of S much larger or much smaller than
those used in the simulations.
4.5. Conduit Resistance Corrections for 1-D Models
[49] One of the most important implications for 1-D
models is related to an estimation of the friction factor
because overall conduit resistance controls the discharge
rate for a given imposed driving pressure drop. The friction
factor is defined as
R dp
l¼ ð25Þ
rV 2 dx
Here V is the average velocity calculated from the discharge
rate. The pressure gradient is calculated from the results of
the simulations and, depending on the regime, can
dramatically deviate from the isothermal Poiseuille solution
in which the friction factor is lp = 8 / Re0. The ratio l/lp
between the calculated friction (25) and the corresponding
Poiseuille solution is shown in Figure 5 at different values
of t from the beginning of the eruption for three different
regimes. Examples reported in Figure 5 clearly show that
depending on the regime, the classical Poiseuille solution
can underestimate or overestimate the real friction factor by
up to an order of magnitude. In the low-S regime, friction
factor is drastically underestimated, whereas, in the large-S
regime, it is highly overestimated. We found that in the
conductive loss-dominated regimes, i.e., low-S, l can be
quite satisfactorily estimated using the Sieder and Tate
[1936] correction:
" #c
8 mðTw Þ
lc ¼  ð26Þ
Re0 m T
where c is an empirical coefficient with a value of about
0.15– 0.25 (suggested values are 0.14 and 0.25 [Sieder and
Tate, 1936]). Generally, c could depend on S too. In this
case both Tw and T can be calculated following the method
described in Appendix A (for practical purposes, T0 can be
used instead of T ). On the other hand, for large S, the
estimation of the friction factor is much more difficult
because the scarcity of data due to the highly demanding

10 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 5. Longitudinal profiles along the conduit of the ratios between the computed friction factor and
the corresponding Poiseuille solutions in the three different regimes (from left to right S  0.1, (Na0 
1.6), S  0.6 (Na0  56), S  5.1 (Na0  125), respectively). Different symbols correspond to the
values of t = 0.234, 0.467, 0.93, 2.34 and 4.67 as reported in the inset (5, 10, 20, 50, and 100 years for
the dimensional parameters used in simulations).

computational costs. We found that an empirical correction
valid in the range of large S for Gz > O (103) can be given
by
8 1
lc ¼ ð27Þ
Re0 1 þ AS1:5
with A  0.3 (Gz > 500 can be considered as a practical
threshold for the validity of the parameterization).
[50] Usage of corrections (26) and (27) significantly
improves the fit for the friction factor for high-S or low-
S regimes, respectively (Figure 6). It is not possible to
apply corrections for the intermediate regime because the
effects of cooling and viscous heating are both of about the
same order. However, in this case the deviation from the
classical Poiseuille solution is not very large, especially in
the case of very large Graetz numbers, when conduction
plays a minor role. Since both relationships (26) and (27)
represent empirical corrections, its application outside the
validity range can lead to incorrect results. Further calcu-
lations for a wider range of parameters will help to improve
the parameterizations (26) and (27) and extend their validity
limits.
5. Discussion
[51] Asymptotic and numerical results show that, for
large discharge rates and/or high magma viscosity, i.e.,

Figure 6. Longitudinal profiles along the conduit of the ratios between the computed friction factor and
the analytical corrections in the three different regimes (from left to right S  0.1, (Na0  1.6), S  0.6
(Na0  56), S  5.1 (Na0  125), respectively). Different symbols correspond to the values of t = 0.234,
0.467, 0.93, 2.34, and 4.67 as reported in the inset (5, 10, 20, 50, and 100 years for the dimensional
parameters used in simulations).

11 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS
Figure 7. Comparison between radial temperature profiles
obtained by fully 2-D (dashed grey lines) and quasi-2-D
(black lines) numerical models for the case of Re0  107,
Pr0  1012, Gz = 560, Na0  125, and S  5.1.
large Na numbers, very high temperatures can be reached
due to viscous dissipation. Typically, this may happen
during explosive eruptions of both basaltic and silicic
magmas (characteristic values of Na range from 102 to
1012) or for lava dome eruptions for highly crystalline
magmas. Our calculations show a temperature increase with
respect to inlet temperature by up to 150 K for S = 5.1 and
up to 45 K for S = 1.05. For typical parameters of explosive
eruptions, Vedeneeva et al. [2005] calculated temperature
rises of order of 1000 K.
[52] We have not considered several mechanisms that can
potentially reduce these very large temperature rises. The
first mechanism that can dampen temperature increase is
related to thermal erosion of conduit walls. Once the
country rock liquidus temperature (usually much higher
than that of the flowing magma because anhydrous con-
ditions) is reached, a consequent thermal erosion can
happen within a narrow layer close to the conduit walls.
The second mechanism is related to crystal content reduc-
tions in the layer where temperature is higher because of the
viscous dissipation. In fact a local crystal content decrease,
which contributes in reducing viscosity, is expected because
of the temperature increases. As a consequence, the regions
that experience high shear and temperatures will be char-
acterized by crystal resorption, transiently or permanently
elongated bubbles and by compositional differences. The
third mechanism is due to the magma viscosity strain rate
dependence [Caricchi et al., 2007] which can reduce
magma viscosity by up to 2 – 3 orders of magnitude in the
high-shear, high-crystallinity regions. Further reductions in
viscosity can be due to the effects of bubble deformation in
the high-shear regions [Llewellin and Manga, 2005]. As
shown by results of fully 2-D simulations, the effective heat
exchange can be enhanced by the effects of thermal
instabilities which may lead to vortex formation [Costa
and Macedonio, 2005]. Another mechanism mentioned by
Mastin [2005] is stress redistribution, followed by brittle
failure in nearby magma or host rock. Like fracture growth,
12 of 16
B12205
the process of viscous dissipation involves strain localiza-
tion along planar features of finite size. Slip along those
planes tends to concentrate stress at the ends of the slip
patches, which then fail by brittle mechanisms because they
are at lower temperature and cannot accommodate the high
strains. This process would be analogous to the growth and
linkage of fractures associated with fault formation [e.g.,
Segall and Pollard, 1983; Martel, 1990]. From a computa-
tional point of view, considering all these processes together
is an extremely complicated and expensive undertaking.
[53] Many mechanisms discussed above can be taken into
account using a quasi-2-D model developed by Vedeneeva et
al. [2005] because the approach is less computationally
expensive. Two significant limitations of the model narrow
its applicability to magmatic flows with heat loss to host
rocks. First, the model is steady state and therefore cannot be
used to explore the transient evolution of typical eruptions.
Because the hydrodynamic timescale is much shorter than
the conductive timescale in host rocks, changes in the
temperature and velocity fields occur quasi-statically. Thus
quasi-static approach can be used. Second, there are only
two types of boundary conditions implemented in the model:
an isothermal BC with wall temperatures equal to the inlet
temperature or adiabatic BC. Therefore the influence of rock
temperature variations on conduit flow dynamics cannot be
treated within the framework of current model. When heat
flux to wall rocks is negligibly small (at large values of t and
large flow rates), the adiabatic boundary condition becomes
a good approximation. Figure 7 shows comparison of
temperature profiles for the case of Re0  107, Pr0 
1012, Gz = 560, Na0 = 125 and S = 5.1. Because of the
presence of some heat flux to the wall rocks in the case of
fully 2-D model the quasi-2-D model overestimates the
temperature increase due to viscous dissipation although
the difference in model predictions is less than 20%. In the
case of fully 2-D model the temperature maxima occur at
some distance from the wall because of the cooling of the
layer of magma attached to the wall. In the case of adiabatic
conditions the maximum temperature increase occurs direct-
ly at the conduit wall.
[54] The advantage of the quasi-2-D model is that it is
computationally very efficient and allows fine meshes for
highly nonlinear problems. As an example, we show cal-
culation results for the case when viscosity is a function of
crystal content as by Costa [2005] and Costa et al. [2007a,
2007b], and the crystal content is a linear function of
temperature b = (TL  T)/(TL  Ts), where Ts and TL are
assumed constant (Figure 8). As a reference, temperature
profiles obtained in the fully 2-D model for the case of
constant crystal content are shown with crosses (Figure 8a).
The temperature increase due to viscous heating is smaller
in the case of variable crystal contents due to the formation
of the layer near the wall where crystal content, and
therefore viscosity of magma, is lower (Figure 8b). Crystal
content of the magma decreases from 52 down to 33%
leading to a viscosity decrease by a factor of 160.
[55] At its current stage the quasi-2-D model gives an
oversimplified representation of the conduit flow dynamics
in comparison with existing 1-D models (e.g., Melnik and
Sparks [2005] for the case of lava dome eruptions) because
it does not account for volatile exsolution, bubble growth,
degassing-induced crystallization, and gas filtration through
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Figure 8. (a) Influence of crystal resorption and crystal content-dependent viscosity on temperature
increase inside the conduit. Dashed grey lines are calculations for Q = 500, and black lines are
calculations for Q = Q(b) following Costa [2005] and Costa et al. [2007a, 2007b]. (b) Radial variation of
crystal content across the conduit.

the magma. However, the approach used in the model estimated from the Poiseuille solution, which is widely used
allows one to build all these processes into the model in 1-D flow models can be orders of magnitude larger for
keeping the code computationally efficient. More effort will the case of significant magma cooling during ascent or
be required to make this model transient. Even with such orders of magnitude smaller when viscous heat generation is
complications the model will not be able to account for all significant. Correction to the Poiseuille solution makes it
aspects of magma-rock interactions (including melting and possible to estimate the friction within ‘ order of magnitude
gas exchange with the surrounding hydrothermal system). A for both cooling- and heating-dominated regimes. Large
fully 2-(3)-D code must be developed in the future in order cross-conduit variations in temperature, estimated both from
to study this complicated phenomenon. It will require asymptotic and numerical analysis, must be considered in
intensive usage of parallel processing and collaboration the interpretation of crystal distribution, textural and com-
between scientists with mathematical, computer science positional diversity of volcanic products.
and geological backgrounds. Successful realization of such
projects in volcanology already exists [Esposti Ongaro et Appendix A: One-Dimensional Model for
al., 2007]. Estimating the Wall Temperature and Nusselt
[56] A major issue is that account of cross-conduit
variations in magma properties is essential for better under-
Number for Conductive-Heat-Loss-Dominated
standing of the dynamics of volcanic eruptions. Preliminary Regimes
estimates of the eruption intensity based on 1-D isothermal [58] Here we provide the derivation of the wall temper-
ascent models can be orders of magnitude wrong in both ature Tw, the Nusselt number NuD, and the universal time
directions. Discharge rate can be greatly overestimated for function Ea which appear in equations (18) and (24). Let us
the case of cooling-dominated flows or highly underesti- consider a cylindrical conduit with a radius R, filled
R R with a
mated in the case of heating-dominated flows. Models that Rfluid flowing at a bulk average temperature T = 0 rVTdr/
R
take into account parameter variations across the conduit 0 rVTdr, embedded in a host rock having a far-field
will also be very useful for interpretation of textural and temperature Tff = Tff(x). Assuming that the transversal
compositional variations in eruptive products. thermal gradients are much larger than the longitudinal, at
a given depth we can write a 1-D equation for the temper-
6. Conclusion ature inside the country rock (R  r < 1), as [Ingersoll and
Plass, 1948]
[57] Results from calculations of coupled nonisothermal
2-D conduit flow and heat transfer in wall rocks show  2 
@T @ T 1 @ T
complicated nonlinear variations of magma temperature due ¼k þ ; t>0 ðA1Þ
@t @r2 r @r
to interactions between viscous heating and heat conduction
and advection. Magma flow dynamics change drastically as
a consequence of these couplings with respect to the with the following initial and boundary conditions:
isothermal case. When one of the processes is dominant,
T ¼ Tff ; t¼0 ðA2Þ
i.e., heat generation by viscous dissipation or heat loss by
conduction, analytical approximations are found. These
approximations are based on the parameters of the magma @ T
k j ¼ qw ; t>0 ðA3Þ
at the inlet of the conduit and therefore do not require any @r r¼R
knowledge of special and temporal variations of parameters
inside the conduit. Results of simulations show that friction, T ¼ Tff ; r ! 1; t>0 ðA4Þ

13 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

where qw represents the heat flux from the wall per unit section shapes developed by Muzychka and Yovanovich
length and k = k/(r cp). We can express the heat flux qw [2004]:
through the heat transfer coefficient h(x) or its dimension- h
less form, i.e., NuD = 2hR/k: in  Po n
 1=n
NuRH ¼ c1 ð Po  GzRH x Þ1=3 þ c2 pffiffiffi k ðA10Þ
8 pe
qw ð xÞ ¼ 2phð xÞðT  Tw ð xÞÞ ðA5Þ
12
Po    p  ðA11Þ
Equation (A1) with the conditions (A2), (A3), and (A4) has pffiffiffi 192e
an analytical solution for the wall temperature [e.g., Yu et eð1 þ eÞ 1  5 tanh
p 2e
al., 2004]:
Muzychka and Yovanovich [2004] based the dimensionless
Tw ¼ Tff þ 2pNuD ðT  Tw Þ  Ec ðt Þ with number pffiffiffion the square root of duct section RH (for a cylinder
Z 1  RH = pR. Therefore, in the relationships (A10) and (A11),
1 2 J0 ðx ÞY1 ðxÞ  Y0 ðxÞJ1 ðx Þ
Ec ð t Þ ¼ 2 etx  1   dx ðA6Þ GzRHx = ReRH Pr RxH denotes the local Graetz number (where
p 0 x 2 J12 ðxÞ þ Y12 ðxÞ x is the distance from the inlet must be larger then a few
times RH, ReRH = VRH/n Reynolds number, and Pr = n/k
where J0, J1, Y0 and Y1 are the Bessel functions in the Prandtl number). Po indicates a dimensionless factor
standard notation, x = r/R is the dimensionless distance and (sometimes called Poiseuille number) depending on the
t = kt/(rcpR2 ) the dimensionless conductive time. Previous friction factor-Reynolds number products (Po = f Re). The
studies [Yu et al., 2004] have shown that for large values of symbol e = b/a represents the aspect ratio where a and b are
t, the function Ec(t) can be approximated as the major and minor axis of ellipse, respectively. However,
since we consider a cylindrical conduit (a = b = R) we have
  e = 1. Concerning the numerical parameters, Muzychka and
1 1
Ec ðt Þ  Ei t >> 1 ðA7Þ Yovanovich [2004] gave c1 = 0.409 and c2 = 3.24 for
4p 4t
uniform wall temperature boundary condition and c1 =
0.501 and c2 = 3.86 for uniform wall flux boundary
where condition, n ffi 5 and k ffi 0.1 (for ellipse).
[60] In estimating the Nusselt number of fluids with a
Z 1 strong temperature-dependent viscosity, such as magmas, it
Eið zÞ ¼ ev =vdv is necessary to account for effects of viscosity variability
z
[Sieder and Tate, 1936; Wichterle, 2004]. To account for
that, in the relationship (A10), we multiply c1 times L
is the exponential integral function. In general, for magma
following the classical correction proposed by Sieder and
flows, we have regimes where tc  R2/k  1 and we cannot
Tate [1936]:
assume t  1, i.e., we cannot use relationship (A7).
Making a parameterization for the difference between Ec
L ¼ ðm=mw Þax ðA12Þ
and its approximation (A7), we found a reasonable
approximation valid for t > 107 is given by
where m is the viscosity corresponding to the bulk
temperature, mw is the viscosity at T = Tw, and the exponent
   ax is an empirical parameter of the order of 0.1 ranging
1 t 1=2 p 1
Ec ðt Þ  Ea ðt Þ  þ Ei ðA8Þ from 0.05 to 0.25 [Wichterle, 2004] with a recommended
p2 a1 þ a2 t 3=4 þ a3 t 5=4 4 4t
average value of 0.14 [Sieder and Tate, 1936; Wichterle,
2004]. After incorporating all the features described above,
where a1 = 0.5386, a2 = 0.4072, and a3 = 1.2977. Using the Nusselt number for a cylindrical conduit is calculate
(A6) and (A8), we obtain finally a relationship for the finally as
conduit wall temperature Tw as
(   )1=5
2 h  pffiffiffi 1=3 i5 Po 5
NuD ¼ pffiffiffi C1 L pPo  Gzx þ C2 pffiffiffi ðA13Þ
Tff þ 2pNuD Ea ðt Þ T p 8 p
Tw ¼ ðA9Þ
1 þ 2pNuD Ea ðt Þ
with C1 ffi 3 and C2 ffi 0.4. With these parameters we have
For magma flows the thermal relaxation length is typically verified that the model is able to reproduce the computa-
very large [Costa and Macedonio, 2003]. For this reason we tional solution of the fully 2-D thermofluid problems in the
need to adopt a parameterization for the local Nusselt regime ofpvery
ffiffiffi low Re, high Pr and pffiffiffivery low Na (note that
number valid in regimes characterized by a fully developed NuD = 2/ p  NuRH and GzRH x = p Gzx).
velocity distribution and a developing temperature distribu-
tion. Here we restrict our considerations to the case where Notation
viscous heating effects are negligible.
[59] For estimating the Nusselt number Nu we adopt a b rheological parameter in equation (6)
semiempirical model valid over the entire range of dimen- B dimensionless temperature scale for DT* =
sionless duct length for very large Pr and different cross DTbc

14 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS B12205

Br Brinkman number DTvisc temperature scale in equation (12c)

c bulk specific heat l friction factor in equation (25)
cp magma specific heat lp friction factor from Poiseuille solution
D conduit diameter lc corrected friction factor in equations (26) and
Ea approximate conductive universal function (27)
Ec conductive universal function L factor correction for Nusselt number
Ei exponential integral function m dimensional magma viscosity variable
g gravity acceleration vector mR magma viscosity at T = TR
Gr viscous dissipation number in equations (15) m0 magma viscosity at T = T0
and (17) P* dimensionless pressure in equation (16)
Gz bulk Graetz number in equations (15) and (17) Q correction to viscosity due to crystals content
Gzx local Graetz number in equation (23b) q dimensionless temperature variable
H slab thickness r dimensional magma density
k magma thermal conductivity rg density of host rocks
L conduit length S dimensionless regime number
L* latent heat of crystallization t dimensionless time
Na Nahme-Griffith number t stress tensor in equations (3) and (4)
Na0 inlet Nahme-Griffith number x dimensionless longitudinal variable
Na* auxiliary Nahme-Griffith number
NuD Nusselt number based on D [61] Acknowledgments. This work was supported by NERC research
NuRH Nusselt number based on RH grant reference NE/C509958/1, the Russian Foundation for Basic Research
(05-01-00228), and President of Russian Federation program (NCH-
P dimensional pressure 4710.2006.1). Authors would like to thank R. S. J. Sparks for useful
P* auxiliary pressure variable in equation (8) comments. L. Mastin, T. Koyaguchi, and M. P. Ryan are warmly acknowl-
Po Poiseuille number edged for their insightful reviews that helped improve and clarify this work.
Pr Prandtl number
Pr0 inlet Prandtl number References
Batchelor, G. K. (1967), An Introduction to Fluid Dynamics, 615 pp.,
Q volume discharge rate Cambridge Univ. Press, Cambridge, U.K.
Qpois volume discharge rate from Poiseuille solution Bruce, P. M., and H. E. Huppert (1990), Solidification and melting along
r radial direction variable dykes by the laminar flow of basaltic magma, in Magma Transport and
R conduit radius Storage, edited by M. P. Ryan, pp. 87 – 101, John Wiley, Chichester, U.K.
Carey, S., and H. Sigurdsson (1989), The intensity of Plinian eruptions,
RH conduit square root section Bull. Volcanol., 51, 28 – 40.
Re Reynolds number based on R Caricchi, L., L. Burlini, P. Ulmer, T. Gerya, M. Vassalli, and P. Papale
Re0 inlet Reynolds number based on R (2007), Non-Newtonian rheology of crystal-bearing magmas and impli-
cations for magma ascent dynamics, Earth Planet. Sci. Lett, 204, 402 –
S dimensionless temperature scale in equation 419, doi:10.1016/j.epsl.2007.09.032.
(17) Costa, A. (2005), Viscosity of high crystal content melts: Dependence on
t dimensional time solid fraction, Geophys. Res. Lett., 32, L22308, doi:10.1029/
T dimensional temperature variable 2005GL024303.
Costa, A., and G. Macedonio (2002), Nonlinear phenomena in fluids
Tch magma chamber temperature with temperature-dependent viscosity: An hysteresis model for magma
Tenv environmental surface temperature flow in conduits, Geophys. Res. Lett., 29(10), 1402, doi:10.1029/
Tff rock far-field temperature 2001GL014493.
Costa, A., and G. Macedonio (2003), Viscous heating in fluids with tem-
TL liquidus temperature perature-dependent viscosity: Implications for magma flows, Nonlinear
TR reference temperature Processes Geophys., 10, 545 – 555.
TS solidus temperature Costa, A., and G. Macedonio (2005), Viscous heating in fluids with tem-
perature-dependent viscosity: Triggering of secondary flows, J. Fluid
Tw conduit wall temperature Mech., 540, 21 – 38.
T0 conduit inlet temperature Costa, A., O. Melnik, R. S. J. Sparks, and B. Voight (2007a), The control of
T* auxiliary temperature variable magma flow in dykes on cyclic lava dome extrusion, Geophys. Res. Lett.,
T convected mean temperature in equation (11) 34, L02303, doi:10.1029/2006GL027466.
Costa, A., O. Melnik, and R. S. J. Sparks (2007b), Controls of conduit
hTi outlet average temperature in Figure 2 geometry and wallrock elasticity on lava dome, Earth Planet. Sci. Lett.,
u dimensionless longitudinal velocity 260, 137 – 151, doi:10.1016/j.epsl.2007.05.024.
v = (vx, vr) velocity vector Delaney, P. T., and D. D. Pollard (1981), Deformation of host rocks and
flow of magma during growth of minette dikes and breccia-bearing in-
V average velocity trusions near Ship Rock, New Mexico, U. S. Geol. Surv. Prof. Pap., 1202,
w0 concentration of dissolved gas 61 pp.
W slab width Delaney, P. T., and D. D. Pollard (1982), Solidification of basaltic magma
during flow in a dike, Am. J. Sci., 282, 856 – 885.
x longitudinal direction variable Esposti Ongaro, T., C. Cavazzoni, G. Erbacci, A. Neri, and M. V. Salvetti
b crystal fraction (2007), A parallel multiphase flow code for the 3D simulation of explo-
g geometry parameter in equations (19) and sive volcanic eruptions, Parallel Comput., 33(7 – 8), 541 – 560.
(23b) Fialko, Y. A., and A. M. Rubin (1999), Thermal and mechanical aspects of
magma emplacement in giant dike swarms, J. Geophys. Res., 104,
h dimensionless radial variable 23,033 – 23,049.
DP overall pressure drop in equation (10) Hardee, H. C., and D. W. Larson (1977), Viscous dissipation effects in
DT* generic temperature scale magma conduits, J. Volcanol. Geotherm. Res., 2, 299 – 308.
Hess, K. U., and D. B. Dingwell (1996), Viscosities of hydrous leucogranite
DTbc temperature scale in equation (12a) melts: A non-Arrhenian model, Am. Mineral., 81, 1297 – 1300.
DTrheol temperature scale in equation (12b)

15 of 16
B12205 COSTA ET AL.: THERMAL EFFECTS ON MAGMA CONDUITS
Hess, K. U., B. Cordonnier, Y. Lavallée, and D. B. Dingwell (2006), Ex-
perimental confirmation of viscous heating effects in calc-alkaline mag-
mas, paper presented at European Geosciences Union General Assembly
2006, Vienna, Austria, 2 – 7 April.
Huppert, H. E., and R. S. J. Sparks (1989), Chilled margins in igneous
rocks, Earth Planet. Sci. Lett., 92, 397 – 405.
Ingersoll, R., and H. Plass (1948), Theory of the ground pipe heat source
for the heat pump, Heat. Piping Air Cond., 20, 150 – 167.
Ko, Y. S., and A. S. Lodge (1991), Viscous heating correction for thermally
developing flows in slit die viscosimetry, Rheol. Acta, 30, 357 – 368.
Llewellin, E. W., and M. Manga (2005), Bubble suspension rheology
and implications for conduit flow, J. Volcanol. Geotherm. Res., 143,
205 – 217.
Lin, P., and Y. Jaluria (1996), Numerical approach to model heat transfer in
polymer melts flowing in constricted channels, Numer. Heat Transfer,
Part A, 30, 103 – 123.
Lodge, A. S., and Y. S. Ko (1989), Slit die viscosimeter at shear rates up to
5  106 s1: An analytical correction for small viscous heating errors,
Rheol. Acta, 28, 464 – 472.
Martel, S. J. (1990), Formation of compound strike-slip fault zones, Mount
Abbot quadrangle, California, J. Struct. Geol., 12, 869 – 882.
Mastin, L. G. (2005), The controlling effect of viscous dissipation on
magma flow in silicic conduits, J. Volcanol. Geotherm. Res., 143, 17 – 28.
Melnik, O., and R. S. J. Sparks (1999), Nonlinear dynamics of lava dome
extrusion, Nature, 402, 37 – 41.
Melnik, O., and R. S. J. Sparks (2005), Controls on conduit magma flow
dynamics during lava dome building eruptions, J. Geophys. Res., 110,
B02209, doi:10.1029/2004JB003183.
Morini, G. (2006), Scaling effects for liquid flows in microchannels, Heat
Transfer Eng., 27(4), 64 – 73.
Muzychka, Y. S., and M. M. Yovanovich (2004), Laminar forced con-
vection heat transfer in the combined entry region of non-circular
ducts, J. Heat Transfer, 126, 54 – 61.
Nelson, S. A. (1981), The possible role of thermal feedback in the eruption
of siliciceous magmas, J. Volcanol. Geotherm. Res., 11, 127 – 137.
Pearson, J. R. A. (1977), Variable-viscosity flows in channels with high
heat generation, J. Fluid Mech., 83, 191 – 206.
Pearson, J. R. A. (1978), Polymer flows dominated by high heat generation
and low heat transfer, Polymer Eng. Sci., 78, 222 – 229.
Petcovic, H. L., and J. D. Dufek (2005), Modeling magma flow and cooling
in dikes: Implications for emplacement of Columbia River flood basalts,
J. Geophys. Res., 110, B10201, doi:10.1029/2004JB003432.
Polacci, M., P. Papale, and M. Rosi (2001), Textural heterogeneities in
pumices from the climactic eruption of Mount Pinatubo, 15 June 1991,
and implication for magma ascent dynamics, Bull. Volcanol., 63, 83 – 97.
Richardson, S. M. (1983), Injection moulding of thermoplastics: Freezing
during mould filling, Rheol. Acta, 22, 223 – 236.
Richardson, S. M. (1986a), Injection moulding of thermoplastics: Freezing
of variable-viscosity fluids. I. Developing flows with very high heat
generation, Rheol. Acta, 25, 180 – 190.
16 of 16
B12205
Richardson, S. M. (1986b), Injection moulding of thermoplastics: Freezing
of variable-viscosity fluids. II. Developing flows with very low heat
generation, Rheol. Acta, 25, 308 – 318.
Rosi, M., P. Landi, M. Polacci, A. Di Muro, and D. Zandomeneghi (2004),
Role of conduit shear on ascent of the crystal-rich magma feeding the
800-year-B.P. Plinian eruption of Quilotoa Volcano (Ecuador), Bull.
Volcanol., 66(4), 307 – 321.
Russell, J. K., D. Giordano, and D. B. Dingwell (2003), High-temperature
limits on viscosity of non-Arrhenian silicate melts, Am. Mineral., 88,
1390 – 1394.
Sahagian, D. (2005), Volcanic eruption mechanisms: Insights from inter-
comparison of models of conduit processes, J. Volcanol. Geotherm. Res.,
143(1 – 3), 1 – 15.
Shaw, H. R. (1969), Rheology of basalt in the melting range, J. Petrol., 10,
510 – 535.
Segall, P., and D. D. Pollard (1983), Joint formation in granitic rock of the
Sierra Nevada, Geol. Soc. Am. Bull., 94, 563 – 575.
Sieder, E. N., and G. E. Tate (1936), Heat transfer and pressure drop in
liquids in tubes, Ind. Eng. Chem., 28, 1429 – 1436.
Skul’skiy, O. I., Y. V. Slavnov, and N. V. Shakirov (1999), The hysteresis
phenomenon in nonisothermal channel flow of a non-Newtonian liquid,
J. Non-Newtonian Fluid. Mech., 81, 17 – 26.
Stasiuk, M. V., C. Jaupart, and R. S. J. Sparks (1993), On the variations of
flow rate in non-explosive lava eruptions, Earth Planet. Sci. Lett., 114,
505 – 516.
Sukanek, P. C., and R. L. Laurence (1974), An experimental investigation
of viscous heating in some simple shear flow, AIChE J., 20(3), 474 – 484.
Touloukian, Y. S., W. R. Judd and R. F. Roy (1989), Physical Properties of
Rocks and Minerals, 548 pp., Hemisphere, New York.
Tuffen, H., D. B. Dingwell, and H. Pinkerton (2003), Repeated fracture and
healing of silicic magma generate flow banding and earthquakes?,
Geology, 31(12), 1089 – 1092.
Vedeneeva, E. A., O. E. Melnik, A. A. Barmin, and R. S. J. Sparks (2005),
Viscous dissipation in explosive volcanic flows, Geophys. Res. Lett., 32,
L05303, doi:10.1029/2004GL020954.
Yu, M. Z., F. Peng, X. D. Li, and Z. H. Fang (2004), A simplified model for
measuring thermal properties of deep ground soil, Exp. Heat Transfer, 17,
119 – 130.
Wadge, G. (1981), The variation of magma discharge during basaltic erup-
tions, J. Volcanol. Geotherm. Res., 11, 139 – 168.
Wichterle, K. (2004), Transfer in temperature-dependent non-Newtonian
flow, Chem. Eng. Processes, 43, 1223 – 1230.

A. Costa, Osservatorio Vesuviano, Istituto Nazionale di Geofisica e
Vulcanologia, Sezione di Napoli, Via Diocleziano 328, I-80124 Naples,
Italy. (costa@ov.ingv.it)
O. Melnik and E. Vedeneeva, Institute of Mechanics, Moscow State
University, 1-Michurinskii prosp., Moscow 119192, Russia. (melnik@
imec.msu.ru)