Earth and Planetary Science Letters 178 (2000) 59^72

www.elsevier.com/locate/epsl

Integrated two-dimensional lithospheric conductivity

modelling in the pyrenees using ¢eld-scale and laboratory
measurements
Paul W.J. Glover a , Jaume Pous b; *, Pilar Queralt b , Josep-Anton Mun¬oz b ,
Montserrat Liesa c , Malcolm J. Hole a
a
Department of Geology and Petroleum Geology, University of Aberdeen, Aberdeen AB24 3UE, UK
b
Departament de Geodina©mica i Geof|¨sica, Universitat de Barcelona, Mart|¨ Franquës s/n, 08028 Barcelona, Spain
c
Departament de Petrologia, Geoqu|¨mica i Prospecciö Geolo©gica, Universitat de Barcelona, Mart|¨ Franquës s/n,
08028 Barcelona, Spain
Received 19 July 1999; received in revised form 3 February 2000; accepted 19 February 2000

Abstract

Recent magnetotelluric (MT) studies have shown that the lower crust in the Pyrenees contains a high conductivity
zone consistent with a subducting continental slab, whose conductivity is 0.33 S/m. Partial melting has been interpreted
to be the most plausible explanation for this high conductivity. Here we report a two-dimensional conductivity model of
the lithosphere by integrating field-scale and laboratory determinations of the conductivity of continental crustal and
mantle rocks. The laboratory data provide empirical formulas which allow us to determine the fluid saturated rock and
melt conductivity when temperature, pressure and lithology are known. Consequently, we have also calculated the
density, lithostatic pressure, and several alternative temperature profiles for use in the model from gravity, seismic and
thermal field data. These can be used with a prescribed melt fraction to predict the electrical conductivity at depth,
which can be compared with the MT conductivity data. Alternatively, the laboratory data can be combined with the
MT conductivity data to predict the melt fraction at depth. The primary outputs of the modelling are conductivity and
melt fraction prediction profiles for six mixing models; (i) Waff's model/Hashin^Shtrikman (HS) upper bound, (ii) HS
lower bound, (iii) parallel layers, (iv) perpendicular layers, (v) random melt areas, and (vi) a modified Archie's law that
takes account of the presence of two conducting phases. The modelling results indicate that a good match to the MT
data can be obtained along the whole profile by the influence of pressure, temperature and the fluid phase with the only
exception being the subducted slab, where a minimum of 4.7% melt fraction is necessary to explain the data. ß 2000
Elsevier Science B.V. All rights reserved.

Keywords: Pyrenees; electrical conductivity; partial melting; two-dimensional models; magnetotelluric methods; conductivity;
laboratory studies

1. Introduction

* Corresponding author. Fax: +34-3-402-1340; Magnetotelluric (MT) studies in the Pyrenees

E-mail: jaume@natura.geo.ub.es (Fig. 1) have recently noted the presence of a

0012-821X / 00 / $ ^ see front matter ß 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 1 2 - 8 2 1 X ( 0 0 ) 0 0 0 6 6 - 2

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

60 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

Fig. 1. Geological map of the Pyrenees showing the location of the MT line (along which the modelling has been carried out)
and the ECORS seismic line [4].

high conductivity zone (0.33 S/m) that has been geophysical data [4,5]. However, the combination
interpreted as partially molten subducted Iberian of MT data with lithospheric conductivity model-
continental lower crust [1,2]. The subduction is a ling from well controlled laboratory data o¡ers a
consequence of the north^south collision between route whereby certain conduction mechanisms
the Iberian and European plates from late Creta- may be eliminated. This paper reports a new
ceous to early Miocene times. Subduction of the two-dimensional (2D) model of the lithosphere
lower crust was inferred initially from the con- along the existing MT section that uses new and
struction of a balanced cross-section along the existing electrical conductivity data derived from
ECORS^Pyrenees deep re£ection seismic pro¢le laboratory measurements on a range of rock types
[3], which showed an upper crust 100 km longer saturated with aqueous £uids and melts subjected
than the lower crust. The high conductivity slab to lithospheric pressure and temperature condi-
begins in the lower crust at approximately 30 km tions. This model has been used to predict the
depth and at 160 km along the modelling pro¢le electrical conductivity at depth, and the results
(Fig. 2) and subducts towards the north retaining have been compared with conductivities observed
its high conductivity until it joins a high conduc- by MT (Fig. 2a).
tivity deep basement interpreted as aestheno-
sphere at 80 km depth at approximately 90 km
along the modelling pro¢le. The central Pyrenees 2. Model design
are characterised by an asymmetric tectonic dou-
ble wedge of upper crustal rocks above the sub- The modelling calculations have been carried
ducted slab, which incorporates the North Pyre- out on a 2D grid 300 km long by 120 km depth,
nean fault (NPF). This is a major strike-slip fault with a 10 km lateral resolution and a 5 km depth.
which developed during the sinistral displacement The modelled pro¢le (Fig. 1) is approximately co-
of Iberia during the Middle Cretaceous [4]. Fur- incident with the ECORS seismic pro¢le [4], and
ther details of the geology can be found in [3]. the MT data pro¢le [1,2], and has been tied to
Discrimination of the conduction mechanism these laterally using the NPF as a reference
responsible for the observed high MT conductiv- (Fig. 2b). The NPF occurs at a distance of
ities in the subducted slab is currently impossible x = 100 km in the model, where x = 0 km is in
from MT data alone [1,2], or from other indirect the Aquitaine basin in France and x = 300 km at

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 61

Fig. 2. (a) Conductivity pro¢le based on MT ¢eld measurements from Pous et al. [1,2]. (b) Structure and lithology.

Fig. 3. (a) Calculated temperature pro¢le, and (b) the temperature pro¢le from the model of Zeyen and Fernandez [17].

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

62 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

the Catalan coastal range in Spain. Six mixing guidance from the materials cropping out along
models have been used, each of which requires the modelled pro¢le and in di¡erent cross-sections
the input of the relevant rock matrix and melt from neighbouring massifs of the Pyrenees (Fig.
conductivities. These are all functions of litho- 2b) [6]. These are: (i) the sedimentary cover (lime-
static pressure and temperature, so the tempera- stones and sandstones), (ii) low metamorphic
ture and pressure at each point must also be cal- grade rocks, mainly metapelites, (iii) granodiorites
culated. and medium/high grade metapelites, (iv) amphib-
olites, (v) granulite facies rocks, either felsic or
2.1. Structure basic, and (vi) olivine (Fo90 ) representing mantle
peridotite.
The structure of the crust (Fig. 2b) has been
taken from a seismic re£ection section [4] that 2.3. Rock matrix conductivities
extends to a depth of 60 km. Interfaces between
the surface sediments, upper crust, lower crust, Rock matrix conductivities for the surface lith-
mantle and subducting slab have been transferred ologies have been measured in the laboratory with
from the seismic interpretation to the model grid 1 M NaCl (5.84 wt%) aqueous pore £uid, and
to a depth of 60 km. Below this depth only mantle have been considered constant to a depth of
material is expected to occur, except between lat- 5 km for the purposes of the model. Approxi-
eral locations 100 and 120 km at a depth of 60^ mately 15 fresh samples of each rock type (sand-
80 km, where subducting slab is indicated by the stone, limestone and metapelites) were collected
MT data [1,2]. The base of the lithosphere has along the pro¢le. Each has been subjected to he-
been delineated in accordance with the MT mod- lium porosity measurement, and the mean values
el. These last two constraints represent the only are given in Table 1. Each sample has been satu-
occurrences where the conductivity model input rated with 1 M NaCl solution using vacuum and
data are not independent of the MT data against raised pressures, and has been subjected to im-
which they will be tested. pedance spectroscopy measurements from 50 Hz
to 1 MHz at 20³C, a hydrostatic pressure of
2.2. Lithology 4.08 MPa, and a pore £uid pressure of 0.68 MPa
using a four electrode arrangement to remove
Lithological groups have been de¢ned with electrode polarisation e¡ects. All aqueous solu-

Table 1
Saturated rock matrix and melt conductivities
Rock or Source Sample Porosity Fitted conductivity equation
melt type code (%) (S/m)
Rock matrix
Sandstone Laboratory measurement Mean 8.2 þ 0.2 crock = 0.006 (constant)
Limestone Laboratory Measurement Mean 5.2 þ 0.2 crock = 0.002 (constant)
Metasediment Laboratory Measurement Mean 1.4 þ 0.1 crock = 0.002 (constant) for depths 9 5 km
crock = 0.0003 (constant) for depths s 5 km
Granodiorite Glover and Vine [8] B 0.50 þ 0.1 crock = 0.128exp((30.2217/kB T)+
(30.819(P-2))
Amphibolite Glover and Vine [8] E1 0.68 þ 0.1 crock = 0.164exp((30.25177/kB T)+
(30.14(P-2))
Granulite Glover and Vine [8] F1 0.78 þ 0.1 crock = 0.0915exp(30.25489/kB T)
Olivine (Fo90 ) Xu et al. [10] H813 0.00 crock = 955 exp(31.73/kB T)
Melts
Basalt mantle Shankland and Wa¡ [11] ^ ^ cmelt = 18 200 exp((31.147^4.14U1033 P)/kB T)
Slab [14,15] ^ ^ cmelt = 30 constant values from 0.5 to 15 S/m
Note: P = lithostatic pressure (kbar), T = temperature (K), kB = Boltzmann's constant = 8.6171U1035 eV/³C.

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 63

tions were sparged under vacuum to remove dis- as a function of pressure, temperature and lithol-
solved gasses prior to use. The low frequency ogy.
(50 Hz) in-phase conductivity data derived from
these measurements have been used for each rock 2.4. Melt conductivities
type, and are given in Table 1. We have not
measured or included corrections for the temper- The conductivity of basaltic mantle melt as a
ature and pressure dependence of the surface function of temperature has been taken from the
lithologies since these lithologies are represented work of Shankland and Wa¡ [11], which enables
in the model by only a surface layer a few cells a pressure and temperature dependent Arrhenius
thick. equation for the melt conductivity to be written,
Rock conductivities for the remaining major and hence allow the melt conductivity at a given
crustal lithologies (granodiorite, amphibolite, pressure, temperature and depth in the model to
granulite) have been obtained from laboratory ex- be calculated, providing melt is present. The melt
periments that were carried out as a function of in the downgoing slab has been assumed to have a
temperature, lithostatic pressure and pore £uid granitic/rhyolitic composition as it is considered
pressure, all with 100% saturation with 0.5 M to be formed as a partial melt of mainly pelitic
NaCl (2.92 wt%) aqueous pore £uid [7,8]. The rocks (in amphibolite or granulite facies) [6] as
£uid with which these experiments were carried predicted by laboratory experiments on the melt-
out is in the lower range of salinities for the con- ing of pelites [12,13].
tinental crust [9] (2^10 wt%). We have corrected For the slab we have used a range of prescribed
the reported rock conductivity values to those for constant melt conductivities (0.5^15 S/m) that
rocks fully saturated with £uid salinities 1 M covers the likely range of melt conductivities
NaCl as a function of temperature, lithostatic based on experimental data available in the liter-
pressure and pore £uid pressure assuming that ature [14,15]. The model applies the basalt melt
the e¡ective conductivity of the rock at a given conductivity to all grid positions in the mantle,
depth, pressure and temperature is a linear func- and the granitic melt to grid positions in the crust.
tion of the £uid conductivity at the same depth A cut-o¡ temperature controls if a melt is possible
pressure and temperature. This salinity was chos- for each lithology. Thus, for the crustal slab the
en because it lies approximately half way between minimum cut-o¡ is 680³C, which is the temper-
the bounds of crustal salinities given by Nesbitt ature for the melting reaction involving the break-
[9]. Thus, we have used rock conductivities that down of muscovite at 0.5 MPa in the absence of
include the e¡ect of a 1 M NaCl £uid for the aqueous £uids [16]. For the mantle, the cut-o¡
conductivity and melt predictions presented in temperature is 1000³C, which is less than the sol-
this paper. This salinity provides a £uid conduc- idus of peridotites in order to check if melting in
tivity of approximately 50 S/m at lower crustal the mantle was required to ¢t the MT conductiv-
depths. ities.
The rock conductivities for the olivine (Fo90 )
under upper mantle conditions have been ob- 2.5. Temperature
tained from experiments carried out as a function
of temperature and pressure, and with controlled A 2D temperature model of the lithosphere in
oxygen fugacity by Xu et al. [10]. the Pyrenees has been calculated previously by
The data for each major rock type have been Zeyen and Fernändez [17] (Fig. 3b). We have cal-
subjected to a non-linear regression to obtain a culated a new temperature model (Fig. 3a) to be
characteristic Arrhenius equation for each rock consistent with the structure derived by the MT
type that depends both upon temperature and model by using the 2D code of Zeyen and Fer-
lithostatic pressure (Table 1). These equations nändez and their values of heat production and
are used in the model to calculate the rock matrix thermal conductivity. Zeyen and Fernändez as-
conductivity at a given grid position in the model sumed a thickness of the lithosphere beneath Ibe-

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

64 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

ria of more than 100 km and an isothermic each of the conducting phases, as the current
boundary condition at such a depth, our model may £ow from one phase to another if the two
considers the lithospheric thickness to be 80 km in phases have similar conductivities.
Iberia, based on the MT results. However, the The ¢rst two models represent the physical
variation in the calculated surface heat £ow be- bounds of conductivity resulting from an alternat-
tween both temperature models is within the var- ing parallel and a perpendicular arrangement of
iation of the measured data [17]. The temperature rock and melt. The randomly distributed model is
models almost coincide from the surface to a a weighted geometric mean of randomly distrib-
depth of about 40 km, the main di¡erence occur- uted volumes of rock and melt and falls between
ring at a greater depth due to the di¡erent litho- the previous two bounds. A slightly more discrim-
spheric thickness. inating set of bounds has been produced by Ha-
shin and Shtrikman [18]. Wa¡ [19] has proposed a
2.6. Lithostatic pressure mixing model based on a porous medium com-
posed of composite spheres of varying size, each
The lithostatic pressure at depth has been cal- of which has an outer shell of melt completely
culated for each cell by using the relevant rock covering an inner unmolten spherical core. Since
densities. These have been adjusted in order the high conductivity melt of all the spheres are in
to ¢t the gravity Bouguer anomaly data and contact, there is high e¡ective connectivity of the
coincide with those used by Zeyen and Fernändez melt phase. This model seems to make good geo-
[17], the only exception being the density of the metrical sense if we consider partial melting in
slab beneath 50 km, where a density of 3050 rocks to be a distributed process preferentially
kg/m3 was required to ¢t the Bouguer anomaly. occurring at the grain surfaces. The Wa¡ formula
Simple calculations using standard bulk mod- is functionally equivalent to the Hashin and
ulus values show that the density increase due to Shtrikman [18] upper bound. Finally we also con-
subduction results in a slab rock density of no sider a modi¢cation to Archie's formulation valid
more than 3080 kg/m3 . This is evidence that for two conducting phases [20].
no large melt fraction is expected, otherwise a The parallel layer model describes the highest
stronger negative Bouguer anomaly would be re- possible melt apparent connectivity and will there-
quired. fore give the lowest predicted melt fraction for
any given melt conductivity and observed MT-
derived conductivity.
3. Conductivity mixing models

A number of mixing models have been used to 4. Modelling results

calculate the electrical conductivity at depth.
These are (i) parallel alternating layers, (ii) per- The MT conductivity model of Pous et al. [1,2]
pendicular alternating layers, (iii) randomly dis- consists of rectangular blocks of constant conduc-
tributed volumes of melt and rock, (iv) Wa¡'s tivity with a marked contrast of conductivity be-
mixing model/Hashin^Shtrikman upper bound tween the blocks, while the true resistivity distri-
(HS+), (v) HS lower bound (HS-), and (vi) a bution (and the one predicted by the forward
modi¢cation of Archie's law that is valid for use modelling) consists of a gradual variation within
with a conducting rock matrix such as we ¢nd a lithology resulting from the continuous depen-
deep in the crust. All six mixing models require dence of conductivity on temperature and pres-
the rock matrix conductivity, melt conductivity, sure. This dependence can result in the presence
and a melt fraction. The connectivity that is im- of thin zones of high resistivity, which are not
portant for these mixing models is not necessarily resolvable by the MT technique. This lack of res-
the physical connectivity of the phases but the olution should be borne in mind when comparing
e¡ective electrical connectivity associated with the conductivity inferred from the MT data with

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 65

positions 100 and 120 km and depths 55 and

80 km. These values are shown in Fig. 4 for
each mixing model using a prescribed melt
fraction of 9.4% and for the observed MT data.
It is immediately clear that the models incorpo-
rating good melt connectivity agree reasonably
well with the MT data. This is because all the
curves in Fig. 4 have been calculated with a
melt fraction that has been chosen to give a
good ¢t of the Wa¡/HS+ conductivity prediction
to the MT data for a melt conductivity of 5 S/m.
The models describing partial or low connectiv-
ity do not approach the level of the MT data
Fig. 4. Predicted conductivity in the slab as a function of unless unrealistically large melt fractions are
melt conductivity using the six mixing models, with a melt used.
conductivity of 5 S/m and a prescribed melt fraction of
9.4%, and compared to the observed MT conductivity in the
Fig. 5 shows examples of the predicted conduc-
slab. tivities along the whole pro¢le calculated by each
of the mixing models. The data can be considered
in two parts; (i) the crust and mantle outside the
the conductivities predicted by our forward mod- slab, and (ii) the slab. Outside the slab, the melt
el. Accordingly, we have used a smoothed MT fraction has been set to zero so that the predicted
model (Fig. 2a), incorporating both gradual later- conductivities in the crust and mantle outside the
al and depth resistivity variations to compare with slab depend only upon the laboratory derived
the model's conductivity predictions and also as measurements of £uid saturated rock conductiv-
input to the inverse model. We have kept the MT ity. It is clear that the lithologies, £uid salinities
smoothing to a minimum, and have checked that and £uid conductivities that we have used model
its implementation provides the same MT re- the conductivity of the crust and mantle outside
sponse as the initial MT model. the slab very well. This is a robust indication that
our choices of laboratory derived rock conductiv-
4.1. Conductivity prediction using forward ity data and our use of 1 M NaCl salinity £uids
modelling is valid. Within the slab a melt conductivity of
5 S/m and a melt fraction of 9.4% have been
The forward modelling uses laboratory data of prescribed. These parameters have been chosen
£uid saturated rock and melt conductivities as a to ensure that the results of the HS+/Wa¡ model
function of pressure and temperature for each (Fig. 5b) show a close correspondence to the ob-
lithology together with a prescribed melt fraction served MT conductivities (Fig. 2a). The parallel
to produce a predicted conductivity pro¢le. The model slightly overestimates the observed MT
predicted conductivities can then be compared conductivities as a result of the better e¡ective
with the observed MT conductivities. We have connectivities implicit in the parallel model (Fig.
used 30 constant values for the melt conductivity 5a). A better correspondence between the electri-
of the slab (from 0.5 to 15 S/m), which are con- cal conductivities predicted by the parallel layer
sidered to include the likely range of melt conduc- model and the observed MT conductivities can be
tivities at the compositions, depths, temperatures achieved at slightly lower partial melt fractions
and pressures encountered in the Pyrenees. Since (8.7%) for 5 S/m melt conductivity. The remaining
we are primarily interested in the conductivities in models grossly underestimate the electrical con-
the down-going slab, we have taken the mean ductivity in the slab (Fig. 5c^f). Unrealistic values
value of predicted conductivity over all the cells of both partial melt fraction (approaching 100%)
in the model representing the slab between lateral and melt conductivity ( s 30 S/m) are required

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

66 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

Fig. 5. Conductivity predictions for each of the six mixing models with a melt conductivity of 5 S/m and a prescribed melt frac-
tion of 9.4%. (a) Parallel model. (b) HS+/Wa¡ model. (c) Modi¢ed Archie's law. (d) Random melt distribution. (e) HS model.
(f) Perpendicular model.

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 67

Fig. 6. Partial melt fraction (%) predictions for each of the six mixing models with a melt conductivity of 5 S/m. (a) Parallel
model. (b) HS+/Wa¡ model. (c) Modi¢ed Archie's law with m = 2. (d) Random melt distribution. (e) HS model. (f) Perpendicular
model. Colour bar (g) refers to parts (a) and (b), and colour bar (h) refers to parts (d)^(f).

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

68 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

Fig. 7. Comparison of the Wa¡ mixing model predictions of electrical conductivity and partial melt fraction for the new tempera-
ture model (a and b), and for the Zeyen and Fernändez [17] temperature model (c and d). Melt conductivity = 5 S/m, melt frac-
tion = 9.4%.

to attain the values of electrical conductivity in 4.2. Melt prediction using inverse modelling
the slab observed by the MT measurements. In-
deed, there is so little contribution to the overall The mixing models have also been used with
conductivity by the melt for the perpendicular the conductivities of the £uid saturated rock ma-
model (Fig. 5f), that this ¢gure can be used to trix, melt conductivities and the observed MT
indicate the conductivities that would be pre- conductivities to predict the melt fraction occur-
sent in the slab if no melt was present. Compar- ring at depth. Zones that produce zero predicted
ison of Fig. 5f with the MT conductivities (Fig. melt fraction are those where 1 M NaCl saline
2a) indicates clearly that for the chosen litholo- £uids in the rock matrix are su¤cient to explain
gies saturated with a 1 M NaCl aqueous £uid, a the observed conductivity at a given depth. Those
conduction mechanism other than £uid is re- zones where the predicted melt fraction is greater
quired. than zero require an additional conduction mech-

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 69

anism, which is assumed to be due to partial melt- tle conductivity as proposed recently by Mibe et
ing for reasons discussed later in this paper. al. [21].
Fig. 6 shows the predicted partial melt fractions
from such an inversion using a range of mixing
models and for a slab melt conductivity set to 5. Discussion
5 S/m. The mixing models that describe a well
connected melt (the parallel layer and HS+/Wa¡'s The results obtained from this 2D modelling
models) show that melt fractions in the slab are show a good agreement between the observed
always less than 9.4% with no melt occurring else- MT data and conductivity predictions using labo-
where (Fig. 6a,b). The well connected melt mixing ratory data saturated with saline £uids on similar
models have been tested for melt conductivities lithologies to those expected in the Pyrenees at
that are lower (1 and 2 S/m) and higher (7 and prescribed temperature and pressure conditions.
10 S/m). For the 1 S/m and 2 S/m melt con- The lithostatic pressure has been calculated by
ductivities, the slab melt fractions are less than using densities which ¢t the gravity Bouguer
41 and 22%, respectively, and for 7 S/m and anomaly, and the temperature has been obtained
10 S/m melt conductivity the melt fractions are in order to ¢t the surface heat £ow data and the
6.6 and 4.7%, respectively. Mixing models de- geometry of the lithosphere derived by the MT
scribing partial e¡ective connectivity (Fig. 6c,d) data. Saline £uids are already taken into account
or low melt e¡ective connectivity (Fig. 6e,f) pre- in the rock matrix conductivity as they are con-
dict unrealistically high melt fractions approach- sidered to be the main mechanism for conductiv-
ing 100% in and directly behind the slab as well as ity. However, in the deeper thickened crust
in the upper mantle. If the slab melt conductivity (depths greater than 30 km), a conduction mech-
is set to a lower (1 or 2 S/m) or higher (7 or anism other than either (i) aqueous £uids at the
10 S/m) value, the mixing models for partial and modelled proportions, or (ii) conduction through
low e¡ective connectivity melt provide predicted the main mineralogies must be present to explain
melt fractions that are marginally higher or lower, the high conductivities measured by MT. Higher
respectively, but still remain unrealistically high porosities than were used in this modelling would
(of the order of 100%). These results are a robust explain the conductivity if the rocks were satu-
indication that if melt is responsible for the high rated. However, such porosities are unrealistic
conductivities at depth, then it must exist in a for lower crustal rocks [16,23] even if high pore
substantially connected form. The lowest bound £uid pressures are present. If the porosity remains
of the melt fraction occurs when well connected the same as that used in the modelling, a higher
melt is present (parallel layer and HS+/Wa¡'s conductivity £uid phase saturating the rocks
models) with the highest melt conductivity of would also explain the raised conductivity if the
the melt (10 S/m). £uid is considered to be well connected. However,
The temperature model of Zeyen and Fernän- mineralogical considerations require that the
dez [17] was also used in our inverse modelling. granulites of the lower crust contain no aqueous
The result, shown in Fig. 7, is that even with such £uid phase, suggesting that the amounts of melt
a colder model the melt prediction is similar in required in the slab may be even higher than
most of the slab (i.e., 9.4% for the high connec- those indicated by the modelling. In any case, it
tivity HS+/Wa¡ model for a melt conductivity of should be noted that, for the crustal lithologies
5 S/m). By contrast, the Zeyen and Fernändez (pelitic rocks) at the PT conditions prevailing in
model also requires a small melt fraction in parts the Pyrenean thickened crust, a partial melt would
of the upper mantle in order to ¢t the MT form whatever the saturation and conductivity of
conductivity, which is unrealistic bearing in aqueous £uids within the original rock. This melt
mind typical values for the solidus of mantle ma- would quickly absorb any aqueous phase in the
terial. A decrease of the dihedral angle of aqueous rock. The presence of graphite and sulphides can
phases in the mantle would also enhance the man- also be invoked to enhance conductivity. Only

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

70 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

small amounts are necessary, but they would have is inferred from our modelling and corresponds to
to be present in a highly connected manner over the high connectivity (HS+/Wa¡ mixing models)
huge volumes to give the required MT conductiv- and a highest melt conductivity (10 S/m). We use
ities [24]. In the case of graphite, it would have to the HS+/Wa¡ model values rather than the par-
be formed and remain stable over the depth and allel layer model values as it is considered that the
temperature range over which high melt conduc- Wa¡ model represents a much more likely distri-
tivities are observed. Taking these points into con- bution of melt within the rock at depth. An upper
sideration, it is our view that the raised conduc- bound of melt fraction is not possible to be in-
tivities studied in this modelling are most likely to ferred from this modelling alone given that it de-
derive from the presence of partial melting at pends on the degree of melt e¡ective connectivity.
depth beneath the Pyrenees. However, a relatively small amount of melt
Considering the lithological section in the Pyr- should be expected from the following considera-
enees and the temperature model, melting is pos- tions. First, the absence of collisional related vol-
sible in the deeper crust, at depths greater than canism in the Pyrenees may, in part, be due to a
30 km. According to a number of laboratory de- limited melt fraction being present. Melt remains
hydration melting experiments [12,16,22,25] the in its source area until a `critical melt fraction' is
pressure and temperature conditions described in reached, whereupon the melt will segregate and
this paper at the locus of the maximum melting ascend. Most authors consider the critical melt
zone, are within the bounds of the experimental fraction to be about 20% (e.g., [16,26]). It has
phase relations for pelite melting. In particular, also been proposed that deformation enhances
the `melt in' boundaries for the synthetic musco- melt segregation by the increase of the permeabil-
vite^biotite and muscovite^quartz experiments ity of the rock mass through vein and dyke devel-
[25] ¢rst intersect the Pyrenean geotherm at a dis- opment (e.g., [27,28]). The absence of signi¢cant
tance of approximately 110 km along the transect. post-collisional deformation in the Pyrenees indi-
Noting that the maximum pressure attained in the cates that surface volcanism will occur only if the
melting experiments was 15 kbar, and that the melt fraction is above the critical value (20%).
`melt in' boundaries are linear, then extrapolating Secondly, the evolution and geometry of the colli-
these boundaries to 20 or 25 kbar (V60^80 km) sional orogen and the lithological characteristics
would result in a signi¢cant melting interval. In of the thickened crust also favour a limited melt
addition, the main phase initially controlling de- fraction. Most of the magmatism at convergent
hydration melting in both synthetic and natural orogens results from the subduction of hydrous
pelites is muscovite. However, biotite will also rocks from either oceanic crust or upper continen-
contribute to dehydration melting after muscovite tal crust. In the Pyrenees, however, oceanic crust
has been exhausted, such that melting can contin- did not exist before the collision, and crustal
ue to considerable depths in the subducted lower thickening has been accommodated by thrust-
crust. For locations less than V100 km along the stacking of upper crustal units above a subducted
transect, melting could not occur even with ex- lower crust [3]. Moreover, the Pyrenean orogen
trapolation of the experimental data because the reworks an older crust formed during the Hercy-
geotherm is too steep. In addition, at locations nian orogeny, which is di¡erentiated (formed by
greater than V160 km along the transect, crustal granodiorites, amphibolite and granulite facies
melting cannot occur because the depth of inter- rocks) and has low water content at deep levels
section of the geotherm with the pelite melting [29]. As a consequence, the combination of this
boundary would be in the upper mantle. There- distinct crustal geometry and the particular litho-
fore on independent petrological criteria, it is logical characteristics may result in both the ab-
clear that the most likely cause of the high con- sence of free water and a limited amount of rock
ductivities is the presence of melt beneath the Pyr- rich in hydrous minerals at depth. From the
enees. above considerations a melt fraction not exceed-
A lower practical bound of melt fraction (4.7%) ing 20% would explain the available geological

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72 71

and geophysical data. In addition, if melt was hara, and two anonymous referees for providing
present in amounts well above 10% the reduced data or advice. This work was made possible by
density of the rock would produce a more nega- the sponsorship of Acciones Integradas (HB1997-
tive gravity anomaly than that observed. 14)/The British Council (1837) and DGICYT
(PB95-269)[FA].

6. Conclusions

2D conductivity modelling has been carried out References

with geologically and geophysically de¢ned inputs
[1] J. Pous, J.J. Ledo, A. Marcuello, M. Daignie©res, Electrical
including laboratory data of the conductivities of resistivity model of the crust and upper mantle from a
£uid saturated lithologies corresponding to the magnetotelluric survey through the central Pyrenees, Geo-
Pyrenees, melt conductivity, and a range of sce- phys. J. Int. 121 (1995) 750^762.
narios where the melt fractions of various struc- [2] J. Pous, J.A. Mun¬oz, J.J. Ledo, M. Liesa, Partial melting
of subducted continental lower crust in the Pyrenees,
tural elements of the model have been varied. The
J. Geol. Soc. 152 (1995) 217^220.
modelling results indicate that a good match to [3] J.A. Mun¬oz, Evolution of a continental collision belt:
the MT data can be obtained. The presence of ECORS-Pyrenees crustal balanced cross-section, in: K.
aqueous £uids alone is su¤cient to explain the McClay (Ed.), Thrust Tectonics, Chapman and Hall,
majority of the upper and lower crust throughout London (1992) 235^246.
[4] P. Choukroune, ECORS Team. The ECORS Pyrenean
the pro¢le. The conductivity of the uppermost
deep seismic pro¢le re£ection data and the overall struc-
mantle can be explained fully by the conductivity ture of an orogenic belt, Tectonics 9 (1989) 23^39.
of olivine (Fo90 ) obtained from laboratory mea- [5] A. Souriau, A. Granet, A tomographic study of the litho-
surements, and no mantle melting is required. The sphere beneath the Pyrenees from local and teleseismic
subducted Iberian lower crust has an anomalously data, J. Geophys. Res. 100 (1995) 18117^18134.
[6] D. Vielzeuf, Relations de phases dans le facie©s granulite et
high conductivity which cannot be explained by
implications gëodynamiques, Ann. Sci. Univ. Clermont II
the presence of aqueous £uids or rock matrix con- 79 (1984) 1^288.
duction, and is unlikely to be explained by the [7] P.W.J. Glover, F.J. Vine, Electrical conductivity of car-
presence of interconnected graphite, sulphides or bon-bearing granulite at raised temperatures and pres-
other conducting phases. An additional conduc- sures, Nature 360 (1992) 723^726.
[8] P.W.J. Glover, F.J. Vine, Beyond KTB-The electrical con-
tion mechanism is therefore required, which given
ductivity of the deep continental crust, Surv. Geophys. 16
the pressure and temperature conditions and crus- (1995) 5^36.
tal lithologies is interpreted as being melt. The [9] B.E. Nesbitt, Electrical resistivities of crustal £uids,
modelling indicates that the lower bound of the J. Geophys. Res. 98 (1993) 4301^4310.
melt fraction is 4.7% assuming a melt conductivity [10] Y.S. Xu, B.T. Poe, T.J. Shankland, D.C. Rubie, Electrical
conductivity of olivine, wadsleyite and ringwoodite
of 10 S/m and providing that the melt is well
under upper-mantle conditions, Science 280 (1998) 1415^
connected. The lack of surface magmatic activity 1418.
in the Pyrenees is consistent with the melt fraction [11] T.J. Shankland, H.S. Wa¡, Partial melting and electrical
falling below the critical melt fraction required for conductivity anomalies in the upper mantle, J. Geophys.
melt segregation and transport to the surface, Res. 82 (1977) 5409^5417.
[12] A.E. Patin¬o Douce, N. Harris, Experimental Constraints
partly by the history and lithology of the Pyre-
on Himalayan anatexis, J. Petrol. 39 (1998) 689^910.
nees, and partly by the lack of extensional fea- [13] J.M. Pickering, A.D. Johnston, Fluid absent melting be-
tures in the area. haviour of a two-mica metapelite: experimental con-
straints on the origin of the Black Hills granite, J. Petrol.
39 (1998) 1787^1804.
[14] H.S. Wa¡, D.F. Weill, Electrical conductivity of mag-
Acknowledgements
matic liquids: e¡ects of temperature, oxygen fugacity,
and composition, Earth Planet. Sci. Lett. 28 (1975) 254^
We wish to thank H. Zeyen, M. Fernändez, Je¡ 260.
Roberts, Tom Shankland, K. Hayashi, H. Tsuka- [15] J.J. Roberts, J.A. Tyburczy, Partial-melt electrical con-

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart

72 P.W.J. Glover et al. / Earth and Planetary Science Letters 178 (2000) 59^72

ductivity: In£uence of melt composition, J. Geophys. Res. [24] P.W.J. Glover, Graphite and electrical conductivity in the
104 (1999) 7055^7065. lower continental crust: a review, Phys. Chem. Earth 21
[16] J.D. Clemens, D. Vielzeuf, Constraints on melting and (1996) 279^287.
magma production in the crust, Earth Planet. Sci. Lett. [25] A.E. Patin¬o Douce, E.D. Humphreys, D. Johnston, Ana-
86 (1987) 287^306. texis and metamorphism in tectonically thickened conti-
[17] H. Zeyen, M. Ferna©ndez, Integrated lithospheric model- nental crust exempli¢ed by the Sevier hinterland, Western
ling combining thermal, gravity, and local isostacy analy- North America, Earth Planet. Sci. Lett. 97 (1990) 290^
sis: Application to the NE Spanish Geotransect, J. Geo- 315.
phys. Res. 99 (1994) 18089^18102. [26] A.B. Thompson, J.A.D. Connolly, Melting of the conti-
[18] Z. Hashin, S. Shtrikman, A variational approach to the nental crust: Some thermal and petrological constraints
theory of the elastic behaviour of multiphase materials, on anatexis in continental collision zones and other tec-
J. Mech. Phys. Solids 11 (1963) 12^140. tonic settings, J. Geophys. Res. 100 (1995) 15565^
[19] H.S. Wa¡, Theoretical consideration of electrical conduc- 15579.
tivity in a partially molten mantle and implications for [27] J.D. Clemens, C.K. Mawer, Granitic magma transport by
geothermometry, J. Geophys. Res. 79 (1974) 4003^4010. fracture propagation, Tectonophysics 204 (1992) 339^
[20] J.F. Hermance, The electrical conductivity of materials 360.
containing partial melts: a simple model from Archie's [28] E.H. Rutter, The in£uence of deformation on the extrac-
Law, Geophys. Res. Lett. 6 (1979) 613^616. tion of crustal melts: a consideration of the role of melt-
[21] K. Mibe, T. Fujil, A. Yasuda, Control of the location of assisted granular-£ow, in: M.B. Holness (Ed.), Deforma-
the volcanic front in islands arcs by aqueous £uid con- tion-enhanced Fluid Transport in the Earth's Crust and
nectivity in the mantle wedge, Nature 401 (1999) 259^262. Mantle (1997) 82^110.
[22] A.E. Patin¬o Douce, J.S. Beard, Dehydration-melting of [29] D. Vielzeuf, J.D. Clemens, C. Pin, E. Moinet, Granites,
biotite gneiss and quartz^amphibolite from 3 to 15 Granulites and Crustal Di¡erentiation, in: D. Vielzeuf
kbar, J. Petrol. 36 (1995) 707^738. and Ph. Vidal (Eds.), Granulites and Crustal Evolution,
[23] J.D. Clemens, G.T.R. Droop, Fluids, P^T paths and the Kluwer Academic Publishers, Dordrecht (1990) 59^85.
fates of anatectic melts in the Earth's crust, Lithos 44
(1998) 21^36.

EPSL 5425 26-4-00 Cyaan Magenta Geel Zwart