Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Journal of Volcanology and Geothermal Research 275 (2014) 71–84

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research


journal homepage: www.elsevier.com/locate/jvolgeores

Assessing lava flow evolution from post-eruption field data using


Herschel–Bulkley rheology
Angelo Castruccio a,b,⁎, A.C. Rust c,1, R.S.J. Sparks c,1
a
Departamento de Geología, Universidad de Chile, Plaza Ercilla 803, Casilla 13518, Santiago, Chile
b
Centro de Excelencia en Geotermia de los Andes (CEGA), Chile
c
Department of Earth Sciences, University of Bristol, Wills Memorial Building, Queen's Road, Bristol BS8 1RJ, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: We present a method to estimate the rheology of a lava flow and reconstruct variations of eruption parameters
Received 8 October 2013 from measurements of flow dimensions together with petrological analysis of samples from the solidified flow.
Accepted 6 February 2014 The model assumes a Herschel–Bulkley lava rheology, which we demonstrate is a reasonable approximation
Available online 20 February 2014
from new analyses of samples and rheological measurements from past eruptions of Etna volcano, as well as
published data for other lava compositions and crystal contents. We present a simplified 2-D model for the
Keywords:
Lava flow modelling
flow of a Herschel–Bulkley fluid over an inclined plane with a constant flux from a vent, which we validate
Etna volcano with analogue experiments scaled to typical conditions of lava flows. Further analogue experiments with two
Lava rheology fluids of different rheologies demonstrate the dominance of the rheology at the flow front in controlling the ad-
Lava morphology vance rate of the flow. The Herschel–Bulkley flow model was applied to a lava flow of the 2002 eruption on Etna
volcano. Flow velocity and effusion rate were calculated from the post-eruption flow dimensions and rheology
estimates along the length of the flow were based on the crystal content and glass compositions of samples.
The results compare well with the observations during the eruption, with a mean flow rate of 19.5 m3/s, which
is in the range of the 15–20 m3/s calculated by previous authors.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction Another complication in modelling flow advance rate is the variation


of the rheology with distance from the vent and between the interior
The evolution of lava flows is a complex process that is controlled by and the crust, because the lava cools, degasses and crystallises as it
the properties of the lava (rheology, density), extrusion rate, cooling flows. Additionally, previous field observations of active lava flows
and degassing effects and topography. Crystal-bearing lava flows have (Borgia et al., 1983) and theoretical and numerical studies (Kilburn
been described for many years as having a non-Newtonian rheology. and Lopes, 1991; Wadge et al., 1994) indicate that the front zone of
Early studies (Robson, 1967; Shaw, 1972; Hulme, 1974) proposed a lava flows is responsible for the advance rate, levee formation and
Bingham rheological law for lavas and many models simulate lava debris formation.
flows with this rheology (e.g. Ishihara et al., 1992; Harris and One question that emerges is: can we reconstruct the evolution of a
Rowland, 2001; Del Negro et al., 2008). However, the non-Newtonian lava flow in terms of the front position and flow rate through time using
behaviour is not only due to the yield strength, but also to strain-rate the lava properties left by the eruption? Hulme (1974) in his pioneering
dependence of the viscosity or consistency, usually in the form of work developed a theory to interpret the morphology of lava flows, as-
shear-thinning behaviour (Pinkerton and Norton, 1995; Caricchi et al., suming a Bingham rheology. From this he extracted the mean flow rate
2007; Lavallée et al., 2007; Ishibashi, 2009). The Herschel–Bulkley and yield strength for terrestrial and extra-terrestrial flows. Subsequent
rheological model incorporates both yield strength and shear-thinning works attempting to infer the lava rheology from the flow morphology
effects, but its application to lava flows has not been widely used, (Moore et al., 1978; Bastero et al., 2006) or to model flow advance and
although there has been some theoretical work on the dynamics of final length (Wilson and Head, 1983; Pinkerton and Wilson, 1994)
Herschel–Bulkley fluids in connection with lava flows (Balmforth and have been based on, or used part of his work. Recently, Deardoff and
Craster, 2000; Balmforth et al., 2004). Cashman (2012) inferred the effusion rate for a prehistoric andesitic
lava flow using a formulation proposed by Kerr et al. (2006) that relates
effusion rate to channel width, lava rheology (internal viscosity and
⁎ Corresponding author at: Departamento de Geología, Universidad de Chile, Plaza
Ercilla 803, Casilla 13518, Santiago, Chile. yield strength), slope, and thermal diffusivity. This approach would
E-mail address: acastruc@ing.uchile.cl (A. Castruccio). work only in cases where the flow width is controlled by its rheology
1
Tel.: +44 117 9706156; fax: +44 117 9253385. and is not significantly affected by channel walls (i.e. the flow width is

http://dx.doi.org/10.1016/j.jvolgeores.2014.02.004
0377-0273/© 2014 Elsevier B.V. All rights reserved.
72 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

smaller than the channel width). Another limitation of Kerr et al. (2006) This is valid when Hρgsinβ N τy, with H = thickness of the flow.
is the assumption of constant rheology along the lava flow. These two Integrating with respect to z and with boundary condition u(H) = 0
limitations are addressed in the method presented in this paper. (no slip at the base), the velocity profile inside the flow is given by:
We refine the method of Hulme, firstly by introducing a viscoplastic 8 " 9
> nþ1  nþ1 # >
Herschel–Bulkley rheology that takes into account changes in viscosity >
> Kn Hρg sinβ−τ y n zρg sinβ−τy n >
>
< ρg sinβðn þ 1Þ − for hcbz≤ H >
>
=
due to variations in the strain rate that the Bingham model does not K K
μ ðzÞ ¼  nþ1
capture. We present a simple, 2-D model for the advance of a flow >
> Hρg sinβ−τ y n >
>
>
> Kn >
: for 0≤ z≤hc >
;
over an inclined surface with this rheology and a constant flux, which ρg sinβðn þ 1Þ K
we validate with a series of laboratory experiments scaled to lava ð5Þ
flows. Additional experiments with flows with two fluids test the hy-
τ
pothesis that the rheology of the fluid at the front of the flow dominates y
where hc ρg sinβ is the thickness of the plug region (Dragoni et al., 1986).
its advance rate. We applied the model to the 2002–03 eruption on Etna It can be shown that the average velocity of the flow is:
volcano, to test if we can reconstruct the observed advance rate with
input parameters based on the lava morphology and petrography. We  1−n !
H 2 ρg sinβ 3n ρg sinβ n
used lava samples and levee dimensions at different distances from u¼
3K H2 ðn þ 1Þ K
the vent to estimate changes in rheology that occurred as the lava  
nþ1 n 2nþ1

advanced, using the melt (glass) composition and crystal content.  H ðH−hc Þ − n
ðH−hc Þ n : ð6Þ
2n þ 1
Then, we investigated if it is feasible to use the dimensions of the flow
(length, width and thickness) together with the inferred rheological
 (with units of volume/
If we assume that the average flow rate Q
properties to estimate the variations in extrusion rate and the advance
of the flow during the eruption. time) is equal to:

 ¼u
Q  HW ð7Þ
2. Theory of Herschel–Bulkley flows over an inclined plane

In this section, we derive a simple 2-D relationship for the front ve- with W the width of the flow. Combining Eqs. (6) and (7):
locity of a fluid over an inclined surface as a function of the volumetric 3  1−n !
flow rate, slope and rheological parameters. This approach is valid  ¼ H Wρg sinβ 3n ρg sin β n
Q
when the aspect ratio (width/thickness) of the flow is large enough to 3K H 3 ðn þ 1Þ K
 
make wall effects negligible. According to Tallarico and Dragoni nþ1 n 2nþ1
 H ðH−hc Þ n − ðH−hc Þ n : ð8Þ
(1999), 2-D models of a Newtonian fluid flowing over an inclined 2n þ 1
plane give similar solutions to 3-D models with wall effects when the
aspect ratio is greater than 3.9. Long lava flows (from a few to tens of With this equation we can find the thickness of the flow H (and con-
kilometres in length) usually have widths N 100 m (e.g. Naranjo et al., sequently the velocity of the flow) for any given Q  . The equation has
1992; Navarro-Ochoa et al., 2002; Vicari et al., 2009) with thicknesses multiple solutions, but only one with physical significance, which is
rarely exceeding some tens of metres. Consequently, a 2-D model is real and greater than hc. The solution and details of the procedure to
appropriate to describe the evolution of the front of a long lava flow obtain it can be found in the supplementary material.
providing that the aspect ratio at the front is large enough. Conse-
quently, the model can be applied to both lavas self-channelized by 3. Experiments with Herschel–Bulkley fluids
levee formation and lavas confined in valleys. We assumed that the
fluid has a rheology that can be modelled with the Herschel–Bulkley As a first test of the validity of Eqs. (6) and (8) to describe the
relationship: advance of lava flows, we performed a series of experiments using fluids
with a rheology that can be adequately modelled with the Herschel–
n
τ ¼ τy þ K y ð1Þ Bulkley relationship (Eq. (1)). The apparatus is a Perspex channel,
10 cm wide, 5 cm in height and 100 cm long with an adjustable slope
where τ is the applied stress, τy is the yield strength, K is the consis- set at 5°, 10° or 15° (Fig. 1). The channel has two holes for the fluid to
tency, y is the strain rate and n is a power law exponent. Note that be injected: a 1 cm diameter circle and a 1 cm wide, 10 cm long rectan-
when τ y = 0 and n = 1, K becomes the Newtonian viscosity. We gle in order to have point or linear sources (Fig. 1). The fluid pump
made the following assumptions and simplifications: no slip at the consists of a Perspex cylinder, 10 cm in diameter and 45 cm long and
base of the flow and the flow has a constant thickness. a piston controlled by an electric motor. The piston moves with veloci-
The applied stress in the fluid is ties between 1 and 295 μm/s, which translates into flow rates between
7.85 × 10− 9 and 2.32 × 10−6 m3/s. The base of the channel was
τ ¼ zpg sin β ð2Þ covered with sandpaper in order to avoid basal slip.
We used syrup, suspensions of syrup plus solid particles (sugar
where z is the depth from the top of the lava flow, ρ is the lava density crystals or nylon cubes; grain size ~ 1 mm) and hair gel as the flow-
and β is the terrain slope. The shear rate is defined as: ing materials. The syrup is a Newtonian material with a viscosity
of ~50 Pa s at 20 °C. The syrup plus crystal suspension is a shear thinning
du material with a low yield strength/consistency ratio. Hair gel is also a
γ̇¼ ð3Þ
dz shear thinning material but with a high yield strength/consistency
ratio. Table 1 indicates the main rheological parameters of the fluids.
where u is the flow velocity and z is the coordinate perpendicular to the Syrup and hair gel rheologies were determined with a concentric
surface, with the origin at the top of the flow. Combining Eqs. (1), (2) rheometer, using incremental steps with strain rates between 0.05
and (3) we get: and 50 s−1. Syrup plus crystal suspension rheology was determined
using the method developed in Castruccio et al. (2010).
 1 In order to demonstrate similarity between our experiments and
du zρg sin β−τ y n lava flows, we used dimensional analysis similar to Applegarth et al.
¼ : ð4Þ
dz K (2010). The main parameters of the experiments and lava flows and
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 73

Fig. 1. A) Experimental setup with the sloping channel and the electrical piston pump. Below the sketch there is a plan view of the upper section of the channel, showing the linear and
point sources. B) Run of an experiment in a 10° slope using syrup with 43% (by volume) of crystals with a flow rate of 1.57 × 10−7 m3/s. Note the bulge that formed at the vent. Black lines
indicate 10 cm.

the non-dimensional numbers derived from these quantities are listed Flows of hair gel did not reach a constant thickness but instead the
in Table 2. The dimensional analysis shows that our experiments thickness decreased continuously from the source to the front; this
compare favourably with real lavas and consequently the results can effect was most prominent for the lesser slopes. Measured velocities
be applied to make inferences about the evolution of lava flows. are slightly higher than predicted, probably reflecting that Eq. (6) does
Additional experiments with two concurrent fluids were carried out not include pressure gradients related to differences in thickness along
to test the hypothesis that the rheology of the fluid at the flow front the flow. Despite this, both measured velocities and mean thicknesses
dominates the advance rate of the flow front. Each experiment began are within the predicted range taking into account the error in the mea-
with a fixed volume of a suspension of syrup plus nylon particles behind sured rheology. Interestingly, the velocity and thickness of the flows can
a removable wall 10 cm from the back of the channel. The wall was be predicted from scaling analysis that assumes that the resistance to
lifted synchronously with the initiation of pumping of pure syrup (with- flow is entirely due to the yield strength of the fluid (see derivation of
out particles), injecting syrup through the point source at the base of the yield strength regime in Castruccio et al., 2013):
syrup–nylon particle suspension (Fig. 2a, b and c). Nylon cubes were
chosen rather than sugar crystals (as used in the single fluid experi- Qρg sinβ
u¼ ð9Þ
ments described above) so that the suspension at the front was less Wτy
dense than the syrup alone injected from up-slope, thus preventing
complications due to gravitational (Rayleigh–Taylor) instabilities
during flow. τy
H¼ : ð10Þ
ρg sinβ

3.1. Results The hair gel data are compared to Eqs. (9) and (10) as well as the full
2-D Herchel–Bulkely model (Eqs. (6) and (8)) in Fig. 3c and d.
In all the experiments the flow achieved a steady state with respect
to velocity and thickness after an advance of 5–8 cm from the source. Table 2
The point source experiments reached the same velocities and thick- Variables and non-dimensional parameters of lava flow systems and experiments.
nesses as the linear source experiments once the flow touched the
Parameters Basalt–andesite Dacite–rhyolite Experiments
walls of the channel and we measured velocity and thickness after
this. The aspect ratio of the flows in our experiments are N5, thus wall K (Pa sn) 1–107 108–1010 45–1000
n 0.4–1 0.4–1 0.3–1
effects can be neglected as shown by Tallarico and Dragoni (1999).
τy (Pa) 0–105 104–106 0–60
We plotted the measured velocity of the flow front and the mean thick- ρ (kg/m3) 2500 2500 1000–1500
ness of the whole flow against the flow rate in a manner similar to Width (m) 10–500 50–1000 0.1
Dragoni et al. (1986) and compared the results with the expected values Height (m) 1–10 10–100 0.005–0.03
using Eqs. (6) and (8). Slope β 1–35 1–35 5–15
Velocity (m/s) 10−4–10−1 10−5–10−3 10−5–10−3
The results of the experiments with pure syrup are plotted in Fig. 3. Strain rate (~V/H) (1/s) 10−5–10−1 10−7–10−4 10−4–2
Both thickness and velocity are in excellent agreement with the theory. g (m/s2) 9.8 9.8 9.8
The thickness was constant to within measurement precision all along Flow rate (m3/s) 0.1–1000 0.1–100 1 × 10−8–2 × 10−6
the flow and the velocity became constant after the flow touched the Non-dimensional numbers
sides of the channel. Reynolds number 10−9–10−2 10−10–10−3 10−8–5 × 10−3
Bingham number 0–102 10–103 0–5 × 10
Table 1 Gravitational/viscous 102–1010 10−1–106 10−1–103
Rheological parameters of the fluids used in the experiments. forces
Aspect ratio (H/W) 2 × 10−2–10−1 10−1–2 × 10−1 5 × 10−2–3 × 10−1
Material K (Pa sn) n τy (Pa)
Notes:
Syrup 55 ± 8 1 ± 0.01 0 Reynolds number: ρVDh / μapp, with Dh = WH / (2H + W) and μapp = app. Viscosity =
Hair gel 27 ± 5 0.38 ± 0.03 61 ± 10 n−1
τ y =γ̇þ Kγ̇ .
Syrup (30% crystals) 97 ± 10 0.53 ± 0.05 12 ± 3 Bingham number = τ y =Kγ : n .
 
Syrup (43% crystals) 800 ± 150 0.63 ± 0.04 4±3 Gravitational/viscous forces = Hρ sinβ= τ y þ Kγ: n .
74 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

Fig. 2. A), B) and C) Time sequence of a two-fluids experiment. The low viscosity fluid (dark colour fluid) is injected at the back of the flow. The high viscosity fluid (43% crystals) is the light
coloured fluid at the front (and later front and sized) of the flow. D) Plot of the flow evolution of a two-fluid experiment, showing the three dynamical regimes that develop during the flow.

The results with the mixtures of syrup plus sugar crystals are plotted in in excellent agreement with theory, whereas the measured thicknesses
Figs. 3 and 4. The flows usually advanced 7–8 cm before achieving a steady (where not affected by the near-vent bulge) were systematically slightly
state. An interesting phenomenon occurred at the source region where a lower than expected but still usually within the experimental error.
bulge formed, usually with its peak 1–1.5 cm higher than the rest of the Our experiments show that the approximation we used of a 2-D
flow (Fig. 1). The measured flow front velocities ahead of the bulge were model without considering pressure gradients due to differences in

A Syrup B Syrup
0.01
Thickness (m)
Velocity (m/s)

10-3

K = 55 Pa s 5°
n=1
10°
τy = 0 Pa
15°
0.001
10-4
1.0×10-15 1.0×10 -14
1.0×10 -13
1.0×10-12 1.0×10-07 1.0×10-06 1.0×10-05 1.0×10-04
Q2sinβ (m6/s2) Q/sinβ (m3/s)
0.1
10-3
C Hair gel D Hair gel
Thickness (m)
Velocity (m/s)

10-4

K=26.96 Pa sn
10-5 n=0.38
τy =61 Pa
slope = 15°
0.01
1.0×10-08 1.0×10-07 1.0×10-06 1.0×10-08 1.0×10-07 1.0×10-06
Q (m3/s) Q (m3/s)
10-3 0.1
E Syrup + 30% crystals F Syrup + 30% crystals
Thickness (m)
Velocity (m/s)

10-4

K=97 Pa sn
n=0.53
τy=12 Pa
slope=5°
10-5 0.01
1.0×10-07 1.0×10-06 1.0×10-07 1.0×10-06
Q (m3/s) Q (m3/s)

Fig. 3. Experimental results for golden syrup for A) velocity and B) thickness. The x-axis was normalised with sinβ to collapse the results with different slopes. The solid lines are the the-
oretical values of the velocity and thickness. Dashed lines are the upper and lower bounds in the calculated values due to uncertainties in the measured rheology. C) and D) Experimental
results with the hair gel, showing velocity and thickness respectively, with a 15° slope. The continuous black line shows the theoretical values expected with the measured rheology with
dashed lines reflecting uncertainties. The red lines are the theoretical values using only the yield strength (K = 0). E) and F) Experimental (velocity and thickness respectively) results for
syrup plus 30% crystals. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 75

0.001 0.1
A Syrup + 43% crystals B Syrup + 43% crystals

Velocity (m/s)

Thickness (m)
0.0001
K = 925 Pa sn
n = 0.64
τy = 4 Pa
slope = 10° 0.01
0.00001
-08 -07 -06
1.0×10 1.0×10 1.0×10 1.0×10-08 1.0×10-07 1.0×10-06

Q (m3/s) Q (m3/s)

0.001 C 2 fluids D 2 fluids

Thickness (m)
Velocity (m/s)

syrup
alone
0.01
0.0001
syrup
K = 733 Pa sn alone
n = 0.62
τy = 3 Pa
slope = 10°
0.00001
0.001
1.0×10-08 1.0×10-07 1.0×10-06 1.0×10-08 1.0×10-07 1.0×10-06

Q (m3/s) Q (m3/s)

Fig. 4. A) and B) Experimental results with syrup plus 43% of crystals (by volume). C) and D) experimental results of experiments with two fluids (syrup plus crystals and syrup alone)
experiments. The thick solid line far away from the experiments is the theoretical curve for syrup alone. The thin line is the theoretical curve for syrup plus 43% crystals and the dashed
lines are associated to the uncertainties of the rheology measurements. Open diamonds are the same results of A) and B). Filled squares are the experiments with pure syrup and syrup plus
43% crystals. The results show that in the two-fluid experiments the velocity and thickness are largely determined by the rheology of the front.

thickness or wall effects is appropriate in the range of conditions we test- of the three regimes, as the measured velocity and thickness are very
ed. Dimensional analysis suggests that these experiments can be com- close to the values expected for a material with the front rheology.
pared with lava flows (Table 2). Measured thicknesses are not as well The experiments with two fluids show that for the tested conditions
described by the 2-D theory as are front velocities, probably because the the flow will be controlled by the rheology of the front until the injected
flows were not constant in thickness. Thickness deviations were most material reaches to within a certain distance from the flow front (5 cm
pronounced at the source, where a bulge developed in experiments from the front under the conditions we used). Our experimental flow
with gel or syrup plus crystals. The yield strength measured by rotational conditions with a fixed volume are not expected to occur in a lava
rheometer and by inversion of constant-volume gravity current flow flow as the viscous front will increase its proportion relative to the
(Castruccio et al., 2013) is too low for it alone to explain the thickness of rest of the flow due to cooling and/or degassing effects. We believe
the bulge (i.e., τy b ρgZ, where Z is the thickness of the bulge). This phe- that the third regime developed in our experiments because the propor-
nomenon could be because our 2-D approximation is not valid near the tion of the more viscous fluid at the front decreased in relation with the
source. Balmforth et al. (2004) showed theoretically similar bulges at more fluid pure syrup, which is unlikely to happen in a real lava flow
the origin for viscoplastic flows. Another contribution to the formation unless there is a sudden increase in the effusion rate and new hot lava
of these bulges at the origin could be some degree of compression and is arriving faster at the frontal zone. The experiments support the infer-
suspension jamming inside the pump and hose system, followed by a re- ence that if the eruption rate is low compared to the growth rate of the
laxation as it is extruded, but this hypothesis has not been tested. front during the waning part of an eruption, then the front rheology will
The hair gel experiments demonstrate that if the yield strength/con- dictate the advance. However, if the eruption rate is high enough, then
sistency ratio is sufficiently high, then the flow can be successfully the properties of the extruded lava could dominate.
modelled by taking into account only the yield strength of the fluid.
This validates the approach of Castruccio et al. (2013) who model lava 4. Viscoplastic rheology of lavas
flow advance with Eqs. (9) and (10) when yield strength forces domi-
nate over viscous forces. Many lavas show non-Newtonian behaviour such as the presence of
The experiments with two fluids, where syrup was injected into the a yield strength and strain-rate dependence on viscosity. Various rheol-
channel with a fixed volume of syrup plus particles initially contained be- ogies and relationships between crystal content and viscosity have been
hind a gate, developed three sequential flow regimes. (1) The flow in the proposed (Pinkerton and Stevenson, 1992; Harris and Rowland, 2001;
first 5 cm of advance was controlled by the collapse of the contained vol- Lavallée et al., 2007; Costa et al., 2009), but a unifying relationship for
ume of syrup–particle suspension. (2) The flow then reached a steady lavas of different compositions, crystal shape and size distribution
state where the front velocity and thickness were constant but the remains elusive. In this section we consider whether the rheology of
front of the injected syrup gradually approached the overall flow front. lavas can be adequately described by the Herschel–Bulkley relationship
(3) When the front of the injected syrup was within 5 cm of the front and test if the parameterization obtained in Castruccio et al. (2010) for
of the flow, the flow front accelerated until the syrup busted through K, n and τy (Eq. (1)) is valid when applied to real lava samples. We
the syrup–particle suspension and became the front of the flow analysed two samples collected by Harry Pinkerton during eruptions
(Fig. 2d). The data plotted in Fig. 4c and d correspond to the steady re- of Etna volcano in 1975 and 1983, and compared our rheology estimates
gime. It is clear that at least for the conditions of these experiments, the based on crystal content and glass compositions with field rheological
overall flow is controlled by the more viscous front material in the second data that have been previously published for these lavas (Pinkerton
76 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

and Sparks, 1978; Pinkerton and Norton, 1995). Additionally, we chemical composition of the glass was determined using the Cameca
re-analysed rheological data from lava samples collected by Lavallée SX100 microprobe at the University of Bristol and the temperature
et al. (2007) and Ishibashi (2009) to investigate if our equations are was estimated using the Mg content in glass geothermometer described
still valid at different conditions of strain rate — stress and different by Putirka (2008). The viscosity of the melt was then calculated using
crystals shapes. the composition and temperature with the formulae of Giordano et al.
We assume that the rheology of lavas is described by Eq. (1). In (2008). Finally, we fit the rheological data with the Herschel–Bulkley re-
Castruccio et al. (2010) we obtained the following relationships lationship to solve for the maximum packing fraction of the suspension
between the crystal content ϕ and the parameters K, n and τy of the and the yield strength fitting parameter D, minimising the squared
Herschel–Bulkley equation from experiments with sugar crystals in residuals in the applied stress.
syrup:
4.1. Results
 −2:3
ϕ
K ðϕÞ ¼ K o 1− ð11Þ 4.1.1. Sample E21
ϕm
Sample E21 was collected by H. Pinkerton and R.S.J. Sparks from a
lava flow during the 1975 eruption of Etna volcano. The sample was
8 partially water cooled from Bocca 1 (Pinkerton and Sparks, 1976). The
< 1
  ϕ≤ ϕc sample has a total crystal content (volume fraction) of 0.46, with 0.33
nðϕÞ ¼ ϕc −ϕ ð12Þ
: 1 þ 1:3 ϕNϕc of phenocrysts and 0.13 of microphenocrysts (we defined phenocrysts
ϕm
N1 mm, 0.1 b microphenocrysts b 1 mm and microlites b0.1 mm).
The phenocrysts are mainly plagioclase with length/width ratios of
 3–4 on average and maximum lengths of 4 mm, with minor amounts
0 ϕ≤ϕc
τy ðϕÞ ¼ 8 ð13Þ of olivine and pyroxene. Temperatures were estimated using glass Mg
Dðϕ−ϕc Þ ϕNϕc
content, olivine-glass and plagioclase-glass thermometers (Putirka,
2008), with temperatures in the 1078–1107 °C range, in agreement
where ϕm is the maximum packing fraction, ϕc is the crystal fraction with the field measurement of 1086 ± 3 °C reported by Pinkerton and
when non-Newtonian behaviour starts to occur and was empirically Sparks (1978). We assume melt water contents of b 0.5% based on pub-
determined to be: lished values for Etna lavas (e.g. Pinkerton and Norton, 1995; Andronico
et al., 2005), which is consistent with H2O-by-difference from micro-
ϕc ¼ 0:44ϕm ð14Þ probe totals for our samples. Using the method of Giordano et al.
(2008) we obtained a liquid viscosity of 823 ± 110 Pa s taking into
and D is a fitting parameter. account the error associated with temperature estimates and water
These equations are very similar to those obtained by Mueller et al. content.
(2010) using rotational rheometer measurements of mixtures of Fig. 5a shows the stress–strain rate data for this lava flow collected by
silicone oil and particles of different shapes, and the results given in Pinkerton and Sparks (1978) using a Mark 2 field viscometer at Bocca 1.
this section are in close agreement with results using their equations. The best fit to a Herschel–Bulkley rheology (Eqs. (1) and (11)–(13))
We determined the crystal content of the Etna samples from point was obtained with values of K = 7996 Pa sn, n = 0.76 and τy =
counting with images taken from optical microscope and SEM. The 181 Pa, using ϕm = 0.73 and D = 1.4 × 109 Pa.

2500 10000
A B
2000
stress (Pa)

K (Pa sn)

1500
1000
1000

500
MPF (Eqs. 11-13): 0.73
0 100
0.00 0.05 0.10 0.15 1080 1100 1120
strain rate (1/s) temperature (°C)
100
1.0 C D
yield strength (Pa)

80

0.8 60
n

40
0.6
20

0.4 0
1080 1100 1120 1140 1080 1100 1120
temperature (°C) temperature (°C)

Fig. 5. A) Strain rate–stress data (Pinkerton and Sparks, 1978) for the Etna sample E21. The line is the fit using Eqs. (11)–(13) with a maximum packing fraction ϕm = 0.73. B), C) and D)
Rheological parameters K, n and τy, respectively of the GPW83/2 sample taken from Pinkerton and Norton (1995). The filled diamonds with errors bars are the calculated parameters using
Eqs. (11)–(13) with a ϕm = 0.73. Circles are lavas with higher crystal content than triangles. See Pinkerton and Norton (1995) for discussion.
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 77

4.1.2. Sample GPW 83/2 a parallel plate uniaxial press to investigate the stress–strain rate behav-
Sample GPW 83/2 was collected by H. Pinkerton from an overflow iour of the samples. The temperature of the experiments ranged from
from the main channel at the 1983 Etna eruption. The sample was air 940 to 1010 °C. We selected the samples from Colima, Unzen and
quenched. The sample has a total crystal content of 0.43, with 0.33 of Krakatau as the sample from Bezymianny was too crystalline (ϕ =
phenocrysts and 0.1 of microphenocrysts. The phenocrysts are mainly 0.8) and lies outside the range (ϕ b 0.7) of the experiments with which
plagioclase with length/width ratios of 3–4 on average and maximum we determined Eqs. (11)–(13). The rheological data of Lavallée et al.
length of 4 mm, similar to the E21 sample. Temperatures estimates, (2007) are plotted in Fig. 6a and b. The best fits with Eqs. (11)–(13)
based on Putirka (2008) give a range between 1078 and 1096 °C, just are using a maximum packing fraction of 0.69 (Colima) and 0.77
bracketing the reported temperature of 1095 °C from where it was (Krakatau).
collected (Pinkerton and Norton, 1995). The lower calculated tempera- Ishibashi (2009) studied the rheology of a Fuji volcano basalt sam-
tures may reflect the fact that the sample was air quenched. Fig. 5b, c ple, erupted in 1707, using a concentric cylinder rotational viscometer
and d shows the calculated K, n and τy by Pinkerton and Norton at temperatures between 1157 and 1297 °C. The crystal content varied
(1995), using a rotational viscometer. The filled diamonds with error from 0 up to 0.25 and was mainly composed of plagioclases with very
bars are the calculated values of the sample using Eqs. (11)–(13) with high length/width ratios (mean ratio of 8.5). The data of K and n obtain-
a maximum packing fraction similar to sample E21 (0.73). Both calcu- ed by Ishibashi (2009) are plotted in Fig. 6c and d. The best fit of the data
lated K (3312) and n (0.77) fall well into the range of the measured with Eqs. (11)–(13) was obtained using a maximum packing factor of
values. The calculated yield strength is 43 Pa, also inside the bounds of 0.44. Ishibashi (2009) did not report any yield strength from the
the estimate of Pinkerton and Norton (1995). samples. More recent concentric cylinder rheology measurement by
Our samples from the 2002 Etna eruption that we analyse in the next Vona et al. (2011) on re-melted Stromboli and Etna magmas with crys-
section, are very similar to both E21 and GPW 83/2 in terms of crystal tal content up to 0.27, had very similar results, with ϕm = 0.37–0.42
content and glass composition, except that microlites (b0.1 mm) are (Fig. 6c, d), using the parametrization of Mueller et al. (2010). This
more abundant in the 2002 samples. This difference is likely due to maximum packing fraction is lower than the values we use for Etnean
the fact that the 2002 samples were collected from solidified deposits lavas in this study because of differences in crystal shapes reflecting
and were affected by post-emplacement crystallisation in contrast to differences in crystallization conditions: the crystals that formed in
the quenched samples E21 and GPW 83/2. If we consider only the sam- the rotational viscometer from samples remelted at atmospheric
ple closest to vent from the 2002 eruption (sample AET-6), and consider pressure were mainly very elongated plagioclases, opposed to the
only the phenocrysts and microphenocrysts (i.e. crystals N0.1 mm), more equant crystal shapes of the natural samples analysed here that
the crystal content (0.45) is very similar to both quenched samples must have dominantly crystallised during ascent given the crystallinity
analysed here. of lava close to the vent.

4.1.3. Other examples from literature 4.2. Discussion


There are other published lava rheology datasets that have been fit
to a power-law stress–strain rate relationship, which is a special case For all of the samples described in Section 4.1, the rheological data
of the Herschel–Bulkley constitutive equation with τy = 0. Lavallée are well-represented by the Herschel–Bulkley relationship with the pa-
et al. (2007) analysed four highly-crystalline samples ranging from an- rameterization of Castruccio et al. (2010) (Eqs. (11)–(13)) for a given
desite to dacite, from different volcanoes (Colima, Unzen, Bezymianny crystal content ϕ. The calculated K and n are in excellent agreement
and Krakatau) with crystal contents ranging from 0.5 to 0.8. They used with the data for all the analysed samples, but the yield strength is not

5.0×10 7 8.0×10 07
A B
4.0×10 7 07
6.0×10
stress (Pa)

stress (Pa)

3.0×10 7
4.0×10 07
2.0×10 7

Colima 2.0×10 07 Krakatau


1.0×10 7
φ = 0.55 φ = 0.7
0 0.0
0.000 0.001 0.002 0.003 0.004 0.0000 0.0005 0.0010 0.0015
strain rate (1/s) strain rate (1/s)

15
C 1.0
D
Relative K

10 Fuji
0.8
n

Etna
5 Stromboli
0.6

0 0.4
0.0 0.1 0.2 0.3 0.0 0.1 0.2 0.3
crystal content crystal content

Fig. 6. Strain rate–stress data from Lavalle et al. (2007) for samples from Colima A) and Krakatau B). The lines are the best fits using Eqs. (11)–(12) with a ϕm = 0.69 for the Colima and 0.77
for the Krakatau cases. C) and D) Relative K and n taken from the experiments of Ishibashi (2009) on Fuji and Vona et al. (2011) on Etna and Stromboli basalts using a rotational viscometer.
The best fits were obtained with ϕm = 0.44 for the Fuji case, 0.38 for Etna and 0.41 for the Stromboli sample.
78 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

well constrained by the crystal content in natural samples and in some 5.1. Eruption chronology and field observations
cases there is no apparent yield strength. For the examples of Lavallée
et al. (2007) the stresses applied when determining rheology were According to Andronico et al. (2005), on 26 October 2002 an earth-
very high (in the order of tens of MPa) and any possible yield strength quake swarm preceded and accompanied the formation of fissures
of much less of 1 MPa will be impossible to detect and could be omitted over the NE and S flanks. At about 4:00 GMT on 27 October, a third fis-
in any chosen rheological model. Ishibashi (2009) did not report (or sure opened along the NE-Rift. The fissure was 3.7 km long and trended
note the lack of) any yield strength. An apparent yield strength of SW–NE. The eruptive fissure propagated down-slope accompanied by
order 100 Pa fits the field and laboratory rheology data for Etna lava violent Strombolian and phreatomagmatic explosions at the vent. Two
(Pinkerton and Sparks, 1978), however the fitting parameter D main flows were generated. One went NE and stopped 2.8 km from
(Eq. (13)) is three orders of magnitudes greater than the D values for an- the vent on 31 October. The E flow reached 6.7 km on 3 November
alogue suspensions of syrup and crystals from which Eqs. (11)–(13) are (Fig. 7); after the lava supply from the main vent stopped, 20 m of
derived. Consequently, the equations for yield strength developed from down-slope movement was observed at the flow front on 5 November.
analogue experiments cannot be tested rigorously by comparison with The activity at the S fissure occurred between 27 October 2002 and
natural samples and more investigation is needed to determine the 28 January 2003 showing different eruptive styles with a two-week
causes of the discrepancies between analogue materials and lava pause in lava effusion. The activity ranged from lava fountains, with
samples in the estimated values of the yield strength and to see if crystal lava jets reaching up to 600 m, to Strombolian explosions, with columns
content is really controlling its magnitude. reaching up to 6 km a.s.l. From 10 December the effusive activity de-
The determined maximum packing fraction ranged from 0.42 up to creased slowly until the end of the eruption on 28 January 2003.
0.77. Numerous studies (Chong et al., 1971; Chang and Powell, 1993; Andronico et al. (2005) calculated a mean volumetric flow rate of
Sudduth, 1993; Mueller et al., 2010) showed that this parameter is 15–20 m3/s and a total volume of 7.8–11.8 × 106 m3 based on the area
affected by the size distribution and the shape of the solid particles. covered by the flows and a mean thickness of 4–6 m.
Despite difficulties in determining the size distribution and shape of For this study, data of the front position were compiled from the re-
particles of natural samples and the lack of definitive formulae to deter- ports of the Instituto Nazionale di Geofisica e Vulcanologia (INGV, web
mine the maximum packing fraction from these parameters, our results page: http://193.206.223.22/Etna2002/Main.htm). During the first
show very good agreement with previously published data. The lowest stages (approx. first 4 km) only approximate estimates of front position
maximum packing fraction obtained was 0.44 in the Fuji sample. could be made due to the very rapid advance (first 4 km in a few hours)
Ishibashi (2009) described the sample crystals mainly as very elongated and the development of the eruption through different vents through
plagioclases, with mean length/width ratios between 7.4 and 11.6 and time. The evolution of the flow at these first stages was based mainly
with sizes that varied over 2 orders of magnitude. Mueller et al. on the reported velocities of the flow and photographs rather than
(2010) determined maximum packing fractions of 0.339–0.433 for position data (Fig. 11a) and consequently has a higher error.
elongated particles (prolate shape with mean length/width of 4.69– The NE flank lava flow field, which is the focus of our study, is com-
9.17) but with more unimodal distributions (mean length of 353 μm), posed of many flows generated by the Sections 3a, b, c and d of the erup-
showing that very low maximum packing fractions can be attained tive fissure (Fig. 7). The observations were made at the two main flows
with very elongated particles. The rest of the samples have maximum and some minor flows generated at 3c. The lava flows generated at the
packing fractions ranging between 0.65 and 0.77. Following Marsh 3c segment reached the Piano Provenzana tourist complex and were
(1981), most volcanological studies have assumed a packing fraction later covered by the flows generated by 3d fissure. Individual flows
of 0.6 to calculate the viscosity of crystal-bearing magmas using the Ein- have widths of 13–25 m. The flows show very well formed channels
stein–Roscoe formulation. However, this formulation does not consider bounded by levees with a width of 0.8–1.2 m and thicknesses up to
non-Newtonian effects such as shear thinning or yield strength or size 0.7 m. The levees have slopes between 40 and 50° and are composed
distribution effects. Sudduth (1993) showed that for bimodal distribu- of loose rubble, with very irregular borders and composed by blocks
tions for which the loose random packing for an unimodal distribution with a maximum size of 50 cm in diameter. The central channel is usu-
is 0.59, the maximum packing fraction can be up to 0.83, depending ally lower (0.4 m) than the levees crests, with a variable width between
on the size ratio of the particles and the volume fraction of each size 12 and 22 m.
population. If we approximate the crystal size distribution of the Etna In the NE flow, the cross section of a small branch was observed. The
samples as bimodal (composed of phenocrysts and microphenocrysts) thickness of this branch was 2.5 m, with the typical 'a'a flow sequence of
with modes at 2 and 0.6 mm (size ratio of 3.3) and with a phenocrysts rough clinker at the top and base and a medial massive zone. The clinker
content of 0.63 versus 0.37 of microphenocrysts (from data of the E21 at the top and bottom are very similar with thicknesses between 0.5 and
and GPW83/2 samples), the maximum packing fraction is around 0.7 1 m and together are up to 70% of the total thickness. The clinker is com-
according to Sudduth (1993). This value is a minimum for the studied posed of very angular and irregular fragments with a mean size of 20 cm
samples as the crystal size distribution is more complex than bimodal with blocks up to 40 cm. The massive core could be up to 1 m thick with
and the size range of crystals is greater, and shows that maximum up to 20% of vesicles that are oriented horizontally with maximum
packing fractions greater than 0.6 and in the range we obtained are lengths of up to 4 cm.
achievable. The medial sector (2–5 km from the vent) is larger both in thickness
and width and is more complex than the proximal sector. In the N flow,
5. Application to the 2002 Etna eruption lava flow the central channel–levee morphology is very well developed (Fig. 8b).
In this zone, the flow is slightly channelized by the previous topography
In this section we analyse a lava flow generated during the and the inner levee walls are higher (~5 m) than the flow margin (max.
2002–2003 eruption on Etna volcano. We use its morphometric and 2 m). The widths of the levees are up to 2 m and the pre-existing slope is
petrographic characteristics to estimate the advance and effusion rate close to 30°. The levees consist of blocks 20–30 cm to 1 m across and are
of the lava flow, applying the 2-D advance model and Herschel–Bulkley angular to subrounded. The inner walls of the levees are almost vertical
rheology presented in the previous sections. We focus on the 6.7 km and very complex. Usually the top is composed of loose fragments. The
length lava flow generated on the NE flank, as this flow is simple in char- medial sector could be made either of massive lava, welded fragments
acter (using the nomenclature of Walker, 1972) compared with S flank or matrix-dominated loose rubble or a combination of these. The basal
flows, thus it is easier to determine parameters such as advance and part is usually covered by avalanching of the inner walls of the levee.
volume rate and also the flow is better exposed as it is not covered by The E flow is up to 10 m thick, and individual flows are up to 50 m
subsequent eruptions. wide. Levee slopes usually are close to 30° with very angular blocks
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 79

Fig. 7. Map showing the extension of the lava flows of the 2002 Etna eruption in the NE flank. 3a–3d are the eruptive fissures. Also shown are locations where samples were taken (red
circles) and field measurements were made (green circles) Locations of Fig. 7 are also indicated. (For interpretation of the references to colour in this figure legend, the reader is referred to
the web version of this article.)
Modified from Andronico et al. (2005).

bigger than 1 m. The maximum thicknesses are at the front of individual according to the nomenclature of Lipman and Banks (1987). Thick-
lobes that are composed of blocks bigger than 2 m and have a frontal nesses can be up to 15 m (Fig. 8d). In this sector, the flow is composed
slope of about 45°. mainly of flat and angular massive blocks up to 2 m in diameter. At
Distal sections (5–6.7 km from the vent) do not present the the very end of the NE flow there was a 50 m long outbreak, 8 m thick
channel–levee morphology but are similar to the dispersed zone with more thermally immature blocks, similar to proximal deposits. In

Fig. 8. A) Proximal sector of the 2002 E lava flow, Etna volcano. The arrow indicates the direction of the flow from the 3d fissure. Dashed lines indicate the 40 m wide central channel.
B) Medial part of the NE flow, 3 km from the vent, showing very well developed channel–levee structure. C) Imbricated, curved slabs in the frontal overflow. Flow direction was to the
left. D) 15 m thick flow front of the E flow.
80 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

places it has a toothpaste texture, with curved slabs with wave-like un- petrography and lava dimensions of the margin of the flow best repre-
dulations that can be imbricated (Fig. 8c), similar to lava flows described sent the conditions of the flow front as it passed through.
by Rowland and Walker (1987) and Sheth et al. (2011). Fig. 10 indicates our estimates for the crystallinity, liquid viscosity
and yield strength along the lava flow. Rheology parameters K and n
were obtained using Eqs. (11) and (12) with the crystal content obtain-
5.2. Petrography of the NE flank lavas
ed from SEM images and liquid viscosity from glass composition of the
samples as described in Section 4. We did not include microlites
Three samples were chosen from the E flow generated from the NE
(b0.1 mm) in the crystallinity which we assume formed after emplace-
fissure to analyse variations in crystal content and glass composition
ment as the chilled Etna lava samples from 1975 and 1981 are very
with distance from the vent. Table 3 shows the main parameters
similar to the 2002 samples, except for having a much lower content
analysed. Sample AET-6 was collected 700 m from the vent. The total
of microlites. The yield strength of the lava was calculated from the
bubble-free crystal volume fraction is 0.6, where 0.09 are phenocrysts
levee dimensions measured in the field, following Hulme (1974):
of plagioclase, olivine and clinopyroxene, 0.36 are microphenocrysts
and 0.15 are microlites. Plagioclases phenocrysts are elongated with a 2
τy ¼ 2wl ρg sin β ð15Þ
maximum length of 6 mm and a length/width ratio between 2 and 5
in thin section. They are euhedral to subeuhedral, with zoning from
core to the rim and sieve texture. Olivine and clinopyroxene pheno- where wi is levee width, ρ is lava density and β is terrain slope. We are
crysts are up to 3 mm long and more equidimensional with length/ assuming the levees are formed during the front advance, thus
width ratios between 1 and 2, and usually forming small clots of 3 to 5 reflecting the yield strength of the frontal part of the flow, which is
crystals. Microphenocrysts are mainly plagioclase with length/width ra- cooler and more viscous than the lava at the back that is feeding the
tios between 2 and 3 and zoning similar to the phenocrysts. Microlites front. The flow thickness, H, was estimated at nine field locations with
are also dominated by plagioclase, but very acicular, with length/ a spacing of 500–2000 m depending on ease of access. Lava flow widths,
width ratios greater than 5. Magnetite is the main accessory mineral measured at the same nine locations, were required to convert velocity
and its size ranges from 0.3 mm down to 10 μm and usually presents determinations (Eq. (6)) into effusion rate (Eq. (8)).
cubic euhedral forms. The required parameters were introduced in the right side of Eq. (6)
Sample AET-4 was collected 1.7 km from the vent and is very similar to estimate the flow front velocity at each point. These estimates were
to sample AET-6, but with greater a bubble-free crystal volume fraction used to calculate flow front position and effusion rate (Eq. (8)) as a
of 0.73, with 0.11 of phenocrysts, 0.44 of microphenocrysts and 0.18 of function of time, with the time ti to reach each point xi calculated as:
microlites. Sample AET-2 was collected at the front of the lava flow, X ðx −x
i i−1 Þ
6.6 km from the vent with a total bubble-free crystal content of 0.74, ti ¼ ; t0 ¼ 0 ð16Þ
ui
with 0.13 of phenocrysts, 0.48 of microphenocrysts and 0.13 of
microlites. The more elongated crystals show a preferred orientation
where ui is the calculated velocity at the location xi. The results for flow
and there is an absence of very acicular microlites as found in the
front position are compared to actual data from the time of the eruption
sample AET-6 (Fig. 9).
in Fig. 11a. The dashed lines show the sensitivity of the modelled evolu-
Summarising, the main variation observed in the samples is an in-
tion to variations in the thicknesses measured by us (the central line
crease in total crystal content with distance (Fig. 10a), reflected mostly
was done with an average thickness at each of our nine field locations).
by the microphenocrysts, as phenocrysts and microlites do not show a
Proximal sectors of the flow show the greatest range in measured thick-
clear pattern. Crystals tend to be more aligned with increasing crystal
nesses and consequently, the modelled advance shows large variations.
content and samples close to the vent show higher bubble content
At medial and distal sections, the measured thicknesses of the flow
(0.15).
show a smaller dispersion and the flow does not show large discrepan-
cies in the modelled advance. The modelled flow rate is consistent with
5.3. Modelling of the advance of the lava flow field observations (Fig. 11b) with a mean value of 19.5 m3/s, which is
within the range of 15–20 m3/s estimated by Andronico et al. (2005)
We applied theoretical two-dimensional Herschel–Bulkley flow using the area covered by the lava flow and mean thickness.
over an inclined surface to estimate the frontal velocity (Eq. (6)) and To assess whether the use of the Herschel–Bulkley rheology is a sig-
volumetric flow rate (Eq. (8)) of the lava flow. The lava density was nificant improvement over the Bingham rheology (which is Herschel–
assumed to be constant with a value of 2500 kg/m3, and terrain slope Bulkley with n = 1) we have also modelled the evolution of the lava
was assessed from pre-eruption topographic maps. Other input param- flow with the equivalent two-dimensional theory with a Bingham
eters were determined from petrography and field measurements rheology (Hulme, 1974; Dragoni et al., 1986). The yield strength values
(Table 3). Rheology changes as the flow advanced were estimated by were taken from field measurements (as in the Herschel–Bulkley case)
measuring the variations of the liquid composition (to obtain the liquid and the Bingham viscosity was calculated using the Einstein–Roscoe
viscosity) and crystal content along the flow. We assume that the equation as done by others (e.g., Ishihara et al., 1992; Harris and

Table 3
Parameters used in the modelling of the 2002 lava flow at Etna volcano.

Distance from vent Slope Mean thickness Mean width Crystal content K n τy (Pa)
(m) (°) (m) (m) (Pa sn)

3200 8.0 5.5 200 0.53 149,200 0.58 6700


5510 5.3 10 200 0.56 310,700 0.51 8200
6600 7.1 11 70 0.58 443,800 0.48 9000
6750 7.1 11 90 0.58 481,800 0.47 9100
6840 8.9 11 80 0.59 569,100 0.46 9400
6970 8.0 12 50 0.59 609,800 0.46 9600
7180 4.4 11 40 0.60 721,400 0.45 9800
7250 4.1 15 70 0.60 745,200 0.44 9900
7280 3.8 15 70 0.60 890,700 0.43 10,200
7320 3.2 18 60 0.62 1,541,800 0.40 11,100
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 81

Fig. 9. Samples from the 2002 NE lava flow, Etna volcano. Images of samples AET-6 (A, C and E) and AET-2 (B, D and F). A) and B) are scans of the polished Sections. C) and D) are optical
microscope images and E) and F) are SEM images. Notice the orientation of elongated plagioclases in sample AET-2 (B and F).

Rowland, 2001) using the same melt viscosity (based on petrology) as the Newtonian regime the calculated flow thicknesses are significantly
for the Herschel–Bulkley case. The modelled front position versus time greater than those of actual flows observed in these, or comparable
for the Bingham rheology (Fig. 11d) does not fit the data as closely as flows at Etna (Fig. 12a). With the YS regime, where the core of the
for the Hershel–Bulkley rheology in the initial stages (Fig. 11a) and flow and not just the crust has strength, the lava yield strength inferred
the final position is overestimated by about 40%, compared with less from the flow model is much higher than the value obtained with field
than 5% for the Hershel–Bulkley rheology. measurements.
We also tested the applicability of three simpler lava flow models The YSC scaling model gives similar results for the front advance to
based on scaling arguments that we developed in Castruccio et al. the two-dimensional HB model proposed here, despite moderate
(2013): Newtonian, yield strength in a growing crust (YSC) and a core discrepancies in the thicknesses of the lava flow predicted by the YSC
yield strength (YS) regimes (Fig. 11c). The fit of the YSC model to data model compared to the actual thickness of the lava flow (Fig. 12a).
on flow front position with time is particularly good and was done Both models include an increase in resistance to flow with time but in
with a single yield strength for the crust of 3.5 × 105 Pa, which is the the YSC model this resistance is related to a thickening crust whereas
same order of magnitude we calculated for several eruptions at differ- in the two-dimensional HB model, the increased resistance is related
ent volcanoes (Castruccio et al., 2013). Castruccio et al. (2013) found to changes in crystallinity, and thus rheology, of the lava at the flow
that YSC regime was also the most likely of these three regimes for front with time. We favour the HB model for the 2002 eruption because
the 2001 and 2006 Etna eruptions. In all three Etna case studies, the the force balance that is the basis for the YSC model requires that the
Newtonian and YS regimes are rejected for the same reasons: With yield strength of the crust dominates over the internal viscosity in
82 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

0.8 for yield strength are the same order of magnitude but still higher than
A the 3.5 × 105 Pa value.
crystal content φ 0.6
6. Discussion and conclusions
0.4
In this paper we developed a simple 2-D model for the flow of a fluid
φ = 0.32 + 0.28(1-e(-x/1.21)) with a Herschel–Bulkley rheology over a slope with a constant effusion
0.2
R 2 = 0.99 rate and applied this model to the evolution of the 2002 eruption at Etna
volcano. This was then applied to extracting key eruption parameters
0.0
0 2 4 6 8
such as effusion rate with time and front velocity from lava flow deposit
dimensions and petrography.
distance (km)
The 2-D model assumes that border effects due to side walls are
10000 negligible when the flow is much wider than its thickness, which is
Liquid viscosity (Pa s)

B true for our experiments and most long lava flows. We conducted a se-
8000 ries of analogue laboratory experiments scaled to conditions of typical
lava flows eruptions and we found a very good match with theory in
6000
terms of velocity and thickness of the flows as a function of effusion
4000 rate. The rheology of a wide range of lava samples can be modelled
with the Hershel–Bulkley relationship over a spectrum of crystal con-
2000 ul = 1747(1-φ/0.69)-0.68 tent, liquid composition and applied stress–strain rate conditions. Anal-
R2 = 0.92 ysis of data from past Etna lavas as well as reanalysis of other data from
0 Lavallée et al. (2007), Ishibashi (2009) and Vona et al. (2011) show that
0.4 0.5 0.6
the maximum packing fraction, which depends on the size distribution
crystal content φ and shape of the crystals, is a key parameter to calculate flow rheology
from textures and petrography of natural samples.
15000
In our modelling, cooling of the lava was considered indirectly by
C
yield strength (Pa)

changing crystal content and liquid viscosity with time. We did not
attempt to relate these changes to temperature variations because the
10000
lava crystallinities probably do not reflect equilibrium conditions and
the crystal nucleation and growth rates for lava flows are still poorly
5000
constrained (Crisp et al., 1994) with estimates ranging by orders of
τy =50160φ3.16 magnitude.
R2 = 0.99 In modelling lava evolution we have applied the Herschel–Bulkley
0 rheology to the frontal zone fed by lava flowing from a central channel.
0.4 0.5 0.6 0.7 Many previous simulations of lava flows have used the Bingham rheol-
crystal content φ ogy (e.g. Ishihara et al., 1992; Miyamoto and Sasaki, 1997; Harris and
Rowland, 2001; Del Negro et al., 2008) and have obtained very good
Fig. 10. Empirical fits of A) crystal content versus distance, B) liquid viscosity versus crystal results. A material that exhibits Herschel–Bulkley behaviour can be
content and C) yield strength versus crystal content, using analysed samples collected and approximated as a Bingham fluid but only over a limited strain rate
measurements in the field. range, as explained by Chester et al. (1985) and Pinkerton and Norton
(1995). Thus this approximation will be valid only over limited condi-
controlling the advance rate of the flow (Castruccio et al., 2013). This tions and the parameterization will not be valid over different strain-
condition is not satisfied for the 2002 eruption, especially in the first rate–stress conditions. The front evolution of the 2002 lava flow can be
two days, when most of the lengthening of the flow occurred. fit (within uncertainties related to flow dimension measurements)
In particular, according to Castruccio et al. (2013) the YSC regime with the Bingham flow model but the rheological parameters (yield
will dominate when: strength and Bingham viscosity) that correspond to this fit are not in
agreement with rheology measurements of Etna lavas.
 
u H Our experiments with two fluids also confirm (in the steady regime)
σ c ≫μ : ð17Þ
δ H−δ that the evolution of the flow is controlled mainly by the rheology at the
front with little influence of the properties of the fluid at the back of the
Where δ is the thickness of the external crust, σc is the crustal yield flow. There is evidence from field observations that lava flow advance is
strength, μ is the internal effective viscosity and u is the velocity of the mainly controlled by the front of the flow. A division between stable
flow. For a Herschel–Bulkley rheology, we can define Ω as: channel and frontal zone has been made at lava flows in volcanoes
    such as Arenal (Borgia et al., 1983), Mauna Loa (Lipman and Banks,
H  n
u 1987) and Etna (Favalli et al., 2009) and it has been argued that the dy-
Ω¼ σy þ K ð18Þ
δ H−δ namics that occur at the front controls the advance of the flow: Borgia
et al. (1983) observed that the frontal section was responsible for the
and the condition for the YSC regime to dominate will be: formation of levees, debris and the selection of the flow path. Kilburn
and Lopes (1991) also pointed out that the lengthening of a lava flow
σ c ≫Ω: ð19Þ
is controlled by the lava properties at the flow front and the numerical
If we assume that the internal rheology is given by the liquid viscos- simulations of Wadge et al. (1994) showed that it is possible to get
ity and crystal content estimated from petrography and we take velocity good results simulating only the frontal part of the flow. Our laboratory
and thickness values from field measurements, then a yield strength of experiments where a constant flow of syrup is injected into a more
3.5 × 105 Pa is needed to fit the evolution of the flow with YSC model. viscous front mixture of syrup plus crystals further support the view
This is noticeably lower than the values necessary for the YSC regime that the flow advance is controlled by the rheology of the front rather
to dominate (Fig. 12b) for the first 2 days. After that, the required values than the rheology of the injected syrup when the flow is in the steady
A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84 83

volumetric flow rate (m3/s)


1000
A B
10000
100

distance (m)
10
5000

0 0.1
0 1.0×105 2.0×105 3.0×10 5 4.0×105 5.0×105 0 1 2 3 4 5 6
time (sec) days

10000
C 10000
D

distance (m)
distance (m)

5000 5000
Newtonian
YSC
YS Bingham model
0 0
0 1.0×10 5 2.0×105 3.0×105 4.0×105 5.0×105 0 1.0×10 5 2.0×10 5 3.0×105 4.0×105 5.0×105
time (sec) time (sec)

Fig. 11. A) Modelled flow evolution using Eq. (6) of the flow in the NE flank during the 2002 Etna eruption. B) Data (dots) and calculated (line) flow rates. C) Fitting of the evolution of the
flow obtained using the three models discussed in Castruccio et al. (2013). D) Modelled flow evolution of the 2002 Etna flow using a Bingham model (Eq. (6) but with n = 1). The thin solid
line is the modelled evolution using a rheology adapted to the Etna lavas (thin line in b). The only fitting parameter is the ϕm from which the Bingham viscosity was calculated with the
Einstein–Roscoe equation. Dashed lines in panels A and D mark the variations associated with uncertainties in field measurements of flow thickness.

regime. The field observations at the 2002 Etna eruption also suggest volume of lava was contained by the more mature flow front. The over-
that the main control on lava advance is exerted by the rheology and flow occurred 5 days after the front stopped (Andronico et al., 2005),
flow at the front. The presence of a toothpaste-like overflow at the when the pressure built up and was large enough to break and over-
very front of the NE flow implies that a relatively less viscous, hotter come the front. This corresponds to a late event, but suggests that if
the flow rate from the channel zone is large enough compared with
40 the advance and growth of the front, it will overcome the front and
A will dominate the advance, consistent with analysis of Hawaiian flows
by Castruccio et al. (2013), with typical effusion rates N100 m3/s, but
Thickness (m)

30 Newtonian
also could be related to waxing stage of a flow. If the flow rate from
YSC
the channel is low enough, as in the present case studied, then the
20 YS
front rheology will dominate, because its proportion is not decreasing
compared with the hotter lava that is being effused, comparable to the
10
analogue laboratory experiments of gravity currents with two fluids of
different rheology presented here.
0 The method presented here offers the prospect of extracting key
2000 4000 6000 8000
eruption parameters such as eruption rate and front evolution from
distance (m) the deposit dimensions and petrography. Measured thicknesses and
widths of the flow are assumed to represent the dimensions of the
1.0×10 07
B front flow during its passage and are used together with the rheology
estimates to obtain velocities and flow rates. The rheology of the flow
can be inferred from samples taken from the flow; analysing the glass
composition we obtain the liquid viscosity and together with the choos-
Ω

1.0×10 06 ing of an appropriate ϕm (based in crystal shape and distribution) K and


n are obtained. We have assumed that the most reliable estimate of τy
for flow advance modelling is obtained from the levee dimension,
which also reflects large-scale flow behaviour of the lava.
1.0×10 05 In order to apply our method confidently to determine flow advance
0 2 4 6 and effusion rates from lavas, it is necessary to establish with more cer-
days tainty what fraction of the crystals formed after flow emplacement in
order to refine the estimation of the crystal content at the flow front
Fig. 12. A) Measured (solid circles) and calculated thicknesses for the 2002 Etna lava flow during flow advance. Also, flow thicknesses should be measured care-
using the Newtonian, YSC and YS models. B) Ω values (Eq. (18)) against time for the 2002 fully during and after the eruption to confirm that post-eruption flow
lava flow. The dashed line indicates the yield strength value necessary to fit the evolution
thicknesses are related to the flowing conditions, especially at proximal
of the front with the YSC model. Values of Ω higher than this yield strength mean that the
internal rheology is more important than the yield strength of the external crust for the zones where the velocity is high and the modelled evolution is very sen-
dynamics of the lava. sitive to the input thickness. Finally, the parameterization of the yield
84 A. Castruccio et al. / Journal of Volcanology and Geothermal Research 275 (2014) 71–84

strength as a function of crystal content still is not robust enough and Harris, A.J.L., Rowland, S.K., 2001. FLOWGO: a kinematic thermo-rheological model for
lava flowing in a channel. Bull. Volcanol. 63, 20–44.
more data from others volcanoes are necessary to achieve more reliable Hulme, G., 1974. The interpretation of lava flow morphology. Geophys. J. R. Astron. Soc.
and broadly applicable yield strength estimates. 39, 361–383.
Ishibashi, H., 2009. Non-Newtonian behaviour of plagioclase-bearing basaltic magma:
subliquidus viscosity measurement of the 1707 basalt of Fuji volcano, Japan.
Acknowledgements J. Volcanol. Geotherm. Res. 181, 78–88.
Ishihara, K., Iguchi, M., Kamo, K., 1992. Numerical simulation of lava flows on some volca-
AC acknowledges a travel grant from the Vicerrectoria de noes in Japan. In: Fink, J.H. (Ed.), Lava Flows and Domes: Emplacement Mechanisms
and Hazard Implications. Springer-Verlag, Berlin, pp. 174–207.
Investigacion y Desarrollo (VID) from the University of Chile and Kerr, R.C., Griffiths, R.W., Cashman, K.V., 2006. Formation of channelized lava flows on an
FONDAP project 15090013. ACR acknowledges a Royal Society URF unconfined slope. J. Geophys. Res. 11. http://dx.doi.org/10.1029/2005JB004225
and RSJS acknowledges a European Research Council grant. We thank (B10206).
Kilburn, C.R.J., Lopes, R., 1991. General patterns of flow field growth: aa and blocky lavas.
Harry Pinkerton for kindly giving us samples from Etna volcano to ana-
J. Geophys. Res. 96, 19721–19732.
lyse their crystal content and glass composition. We also thank the help Lavallée, Y., Hess, K.-U., Cordonnier, B., Dingwell, D.B., 2007. Non-Newtonian rheological
and support from the people from the INGV at Catania, Italy during law for highly crystalline dome lavas. Geology 35, 843–846.
fieldwork in 2009. The authors thank Scott Rowland and Jess Robertson Lipman, P.W., Banks, N.G., 1987. Aa flow dynamics, Mauna Loa. In: Decker, W., Wright,
T.L., Stauffer, P.H. (Eds.), Volcanism in Hawaii. US Geol. Surv. Prof. Pap., 1350,
for their insightful reviews that greatly improved the quality of the pp. 1527–1567.
manuscript. Marsh, B., 1981. On the crystallinity, probability of occurrence, and rheology of lava and
magma. Contrib. Mineral. Petrol. 78, 85–98.
Miyamoto, H., Sasaki, S., 1997. Simulating lava flows by an improved cellular automata
Appendix A. Supplementary data method. Comput. Geosci. 23, 283–292.
Moore, H.J., Arthur, D.W.G., Schaber, G.G., 1978. Yield strengths of flows on the earth,
Supplementary data to this article can be found online at http://dx. Mars and the Moon. Proc. Lunar Planet. Sci. Conf. 9th, pp. 3351–3378.
Mueller, S., Llewellin, E.W., Mader, H.M., 2010. The rheology of suspensions of solid par-
doi.org/10.1016/j.jvolgeores.2014.02.004. ticles. Proc. R. Soc. 466, 1201–1228.
Naranjo, J.A., Sparks, R.S.J., Stasiuk, M.V., Moreno, H., Ablay, G.J., 1992. Morphological,
References structural and textural variations in the 1988–1990 andesite lava of Lonquimay
Volcano, Chile. Geol. Mag. 129, 657–678.
Andronico, D., Branca, S., Calvari, S., Burton, M.R., Caltabiano, T., Corsaro, R.A., Del Carlo, P., Navarro-Ochoa, C., Gavilanes-Ruíz, J.C., Cortés-Cortés, A., 2002. Movement and emplace-
Garfì, G., Lodato, L., Miraglia, L., Murè, F., Neri, M., Pecora, E., Pompilio, M., Salerno, G., ment of lava flows at Volcán de Colima, México: November 1998–February 1999.
Spampinato, L., 2005. A multi-disciplinary study of the 2002–03 Etna eruption: in- J. Volcanol. Geotherm. Res. 117, 153–167.
sights for a complex plumbing system. Bull. Volcanol. 67, 314–330. Pinkerton, H., Norton, G.E., 1995. Rheological properties of basaltic lavas at sub-liquidus
Applegarth, L., James, M., Wyk, Van, de Vries, B., Pinkerton, H., 2010. Influence of surface temperatures: laboratory and field measurements on lavas from Mount Etna.
clinker on the crustal structures and dynamics of 'a'a lava flows. J. Geophys. Res. 115. J. Volcanol. Geotherm. Res. 68, 307–323.
http://dx.doi.org/10.1029/2009JB006965 (B07210). Pinkerton, H., Sparks, R.S.J., 1976. The 1975 sub-terminal lavas, Mount Etna: a case history
Balmforth, N.J., Craster, R.V., 2000. Dynamics of cooling domes of viscoplastic fluid. J. Fluid of the formation of a compound lava field. J. Volcanol. Geotherm. Res. 1, 167–182.
Mech. 422, 225–248. Pinkerton, H., Sparks, R.S.J., 1978. Field measurements of the rheology of lava. Nature 276,
Balmforth, N.J., Craster, R.V., Sassi, R., 2004. Dynamics of cooling viscoplastic domes. 383–385.
J. Fluid Mech. 499, 149–182. Pinkerton, H., Stevenson, R.J., 1992. Methods of determining the theological properties of
Bastero, C.F., Mahar, A., Lagmay, F.A., 2006. Estimating SiO2 content of lava deposits in the lavas from their physicochemical properties. J. Volcanol. Geotherm. Res. 53, 47–66.
humid tropics using remotely sensed imagery. J. Volcanol. Geotherm. Res. 151, Pinkerton, H., Wilson, L., 1994. Factors controlling the lengths of channel-fed lava flows.
357–364. Bull. Volcanol. 56, 108–120.
Borgia, A., Linneman, S., Spencer, D., Diego, M.L., Brenes, A.J., 1983. Dynamics of lava flow Putirka, K., 2008. Thermometers and barometers for volcanic systems. Rev. Mineral.
fronts, Arenal Volcano, Costa Rica. J. Volcanol. Geotherm. Res. 19, 303–329. Geochem. 69, 61–120.
Caricchi, L., Burlini, L., Ulmer, P., Gerya, T., Vassalli, M., Papale, P., 2007. Non-Newtonian Robson, G.R., 1967. Thickness of Etnean lavas. Nature 216, 251–252.
rheology of crystal-bearing magmas and implications for magma ascent dynamics, Rowland, S.K., Walker, G.P.L., 1987. Toothpaste lava: characteristics and origin of a lava
Earth Planet. Sci. Lett. 264, 402–419. structural type transitional between pahoehoe and aa. Bull. Volcanol. 49, 631–641.
Castruccio, A., Rust, A.C., Sparks, R.S.J., 2010. Rheology and flow of crystal-bearing lavas: Shaw, H.R., 1972. Viscosities of magmatic silicate liquids: an empirical method of predic-
insights from analogue gravity currents. Earth Planet. Sci. Lett. 297, 471–480. tion. Am. J. Sci. 272, 870–893.
Castruccio, A., Rust, A.C., Sparks, R.S.J., 2013. Evolution of crust- and core-dominated lava Sheth, H.C., Ray, J.S., Kumar, A., Bhutani, R., Awasthi, N., 2011. Toothpaste lava from Barren
flows using scaling analysis. Bull. Volcanol. 75, 681. http://dx.doi.org/10.1007/ Island volcano (Andaman Sea). J. Volcanol. Geotherm. Res. 202, 73–82.
s00445-012-0681-2. Sudduth, R.D., 1993. A new method to predict the maximum packing fraction and the vis-
Chang, C., Powell, R.L., 1993. Dynamic simulation of bimodal suspensions of hydrodynam- cosity of solutions with a size distribution of suspended particles II. J. Appl. Polym. Sci.
ically interacting spherical particles. J. Fluid Mech. 253, 173–209. 48, 37–55.
Chester, D.K., Duncan, A.M., Guest, J.E., Kilburn, C.R.J., 1985. Mount Etna: The Anatomy of a Tallarico, A., Dragoni, M., 1999. Viscous Newtonian laminar flow in a rectangular channel:
VolcanoChapman and Hall, London. application to Etna lava flows. Bull. Volcanol. 61, 40–47.
Chong, J.S., Christiansen, E.B., Baer, A.D., 1971. Rheology of concentrated suspensions. Vicari, A., Ciraudo, A., Del Negro, C., Herault, A., Fortuna, L., 2009. Lava flow simulations
J. Appl. Polym. Sci. 15, 2007–2021. using discharge rates from thermal infrared satellite imagery during the 2006 Etna
Costa, A., Caricchi, L., Bagdassarov, N., 2009. A model for the rheology of particle-bearing eruption. Nat. Hazards 50, 539–550.
suspensions and partially molten rocks. Geochem. Geophys. Geosyst. 10, N3. Vona, A., Romano, C., Dingwell, D.B., Giordano, D., 2011. The rheology of crystal-bearing
Crisp, J., Chashman, K.V., Bonini, J.A., Hougen, S.B., Pieri, D.C., 1994. Crystallization history basaltic magmas from Stromboli and Etna. Geochim. Cosmochim. Acta 75,
of the 1984 Mauna Loa lava flow. J. Geophys. Res. 99 (B4), 7177–7198. 3214–3236.
Deardoff, N., Cashman, K.V., 2012. Emplacement conditions of the c. 1600-year BP Collier Wadge, G., Young, P., McKendrick, I., 1994. Mapping lava flow hazards using computer
Cone lava flow, Oregon: a LiDAR investigation. Bull. Volcanol. 74, 2051–2066. simulation. J. Geophys. Res. 99, 489–504.
Del Negro, C., Fortuna, L., Herault, A., Vicari, A., 2008. Simulations of the 2004 lava flow at Walker, G.P.L., 1972. Compound and simple lava flows and flood basalts. Bull. Volcanol.
Etna volcano using the magflow cellular automata model. Bull. Volcanol. 70, 805–812. 35, 579–590.
Dragoni, M., Bonafede, M., Boschi, E., 1986. Downslope flow models of a Bingham liquid: Wilson, L., Head, J.W., 1983. A comparison of the volcanic eruption processes on Earth,
implications for lava flows. J. Volcanol. Geotherm. Res. 30, 305–325. Moon, Io and Venus. Nature 302, 663–669.
Favalli, M., Harris, A.J.L., Fornaciai, A., Pareschi, M.T., Mazzarini, F., 2009. The distal
segment of Etna's 2001 basaltic lava flow. Bull. Volcanol. 72 (1), 119–127.
Giordano, D., Russell, J.K., Dingwell, D.B., 2008. Viscosity of magmatic liquids: a model.
Earth Planet. Sci. Lett. 271, 123–134.

You might also like