The Scientific World Journal

The Scientific World Journal / 2014 / Article

Research Article | Open Access

Volume 2014 |Article ID 793236 | 32 pages | https://doi.org/10.1155/2014/793236

Multivariate Analysis, Mass Balance Techniques, and Statistical Tests as Tools in Igneous Petrology: Application to the Sierra de las Cruces Volcanic Range (Mexican Volcanic Belt)

Academic Editor: N. Hirao
Received26 Aug 2013
Accepted25 Nov 2013
Published05 Mar 2014

Abstract

Magmatic processes have usually been identified and evaluated using qualitative or semiquantitative geochemical or isotopic tools based on a restricted number of variables. However, a more complete and quantitative view could be reached applying multivariate analysis, mass balance techniques, and statistical tests. As an example, in this work a statistical and quantitative scheme is applied to analyze the geochemical features for the Sierra de las Cruces (SC) volcanic range (Mexican Volcanic Belt). In this locality, the volcanic activity (3.7 to 0.5 Ma) was dominantly dacitic, but the presence of spheroidal andesitic enclaves and/or diverse disequilibrium features in majority of lavas confirms the operation of magma mixing/mingling. New discriminant-function-based multidimensional diagrams were used to discriminate tectonic setting. Statistical tests of discordancy and significance were applied to evaluate the influence of the subducting Cocos plate, which seems to be rather negligible for the SC magmas in relation to several major and trace elements. A cluster analysis following Ward’s linkage rule was carried out to classify the SC volcanic rocks geochemical groups. Finally, two mass-balance schemes were applied for the quantitative evaluation of the proportion of the end-member components (dacitic and andesitic magmas) in the comingled lavas (binary mixtures).

1. Introduction

Several conventional mineralogical, geochemical, and isotopic tools, using a limited number of variables (e.g., bivariate, trilinear, multielement, and semilogarithmic diagrams), have usually been applied to establish a qualitative or semiquantitative view of igneous petrological mechanisms [1, 2]. Particularly, the interaction between, at least, two magmas is one of the most important mechanisms of compositional diversification of igneous rocks [3]. According to genetic relations between the original or resident magma and the later invasive magma, two scenarios could be expected [4, 5]: (a) successive pulses of magma derived from a common source intersect in time and space or (b) unrelated chemical distinct magmas, derived from different sources are involved in the interaction episode. Additionally, different styles of the interaction phenomena are related to the variation of physicochemical parameters (e.g., [3, 6, 7]): (a) the initial contrast in chemical composition, temperature, and viscosity, (b) the relative mass fractions and the physical state of interacting magmas, and (c) the static versus dynamic environment of interaction. These processes have been broadly divided into (a) magma mingling, a route characterized by a physical juxtaposition and intermingling of contrasting compositions, with little or no chemical homogenization, and (b) magma mixing, where the physical and chemical conditions promote the homogenization of contrasting geochemical and isotopic features, resulting in a single magma of intermediate composition. If a magma mixing/mingling model is proposed, it must include statements specifying (a) the initial compositions of the resident and invasive magmas, (b) the modal mineralogy of the magmas prior to mixing, and (c) the proportions of resident and invasive magmas [4]. A quantitative assessment could be obtained from multivariate statistical techniques [8]. Although these methods have been used with classification purposes in igneous rocks [9], their use to understand magma mixing/mingling processes is still limited [7, 1013].

On the other hand, magma mixing/mingling processes have been observed in diverse tectonic settings. Consequently, a complete vision of these magmatic localities, commonly dominated by rocks with > 52% (the subscript refers to the adjusted silica from the SINCLAS computer program [14, 15]), would be facilitated from the tectonic regime. However, a restricted number of conventional diagrams are available for tectonic discrimination of intermediate ( = 52–63%; [16, 17]) and acid ( > 63%; [1, 18]) magmas. Additionally, these schemes have been critiqued as a result of a statistically wrong treatment of compositional data, eye-drawn subjective boundaries for different tectonic fields, and lack of representation of the entire statistical population [19, 20]. S. P. Verma and S. K. Verma [21] and Verma et al. [22], to solve the limitations of the tectonic discrimination conventional schemes, have proposed a set of new discriminant-function-based multidimensional diagrams for intermediate and acid magmas from four tectonic settings (island arc, continental arc, continental rift + ocean island, and collision).

In this context, Velasco-Tapia et al. [23] recently reported, based on mineralogical, geochemical, and Sr-Nd isotopic conventional tools, that the formation of the Sierra de las Cruces (SC) volcanic range (3.7 to 0.5 Ma; central part of the Mexican Volcanic Belt (MVB); Figure 1) was mainly controlled by a magma mixing/mingling process. In this work, as an example, multivariate techniques (linear discriminant, cluster, and principal component analysis), discordancy and significance statistical tests, and mass-balance approaches were applied to establish the tectonic setting and to obtain a quantitative picture of the magmatic evolution of this volcanic range.

2. Geological Synthesis

The SC volcanic range is an elongated volcanic range, extending in a NNW-SSE direction for ~65 km, with a width varying between 47 km to the north and 27 km to the south (Figure 2; [2325]). According to K-Ar geochronological data [26], the main mass of SC volcanic range was erupted between 3.7 and 1.8 Ma. After that, in the middle Pleistocene (~0.5 Ma), another volcanic event produced andesitic domes, being labeled as Ajusco period. It has been considered as the transition to the Sierra de Chichinautzin monogenetic eruptive period (<40 ka; [2729]).

On the basis of morphostructural and radiometric age criteria, the SC volcanic range has been divided into four sectors bounded by E-W faults [23, 24]: (a) northern sector (SCN; 2.9–3.7 Ma), (b) central sector (SCC; 1.9–2.9 Ma), (c) southern sector (SCS; 0.7–1.9 Ma), and (d) las Cruces-Chichinautzin transition sector (SCT; ~0.5 Ma). The northern and central sectors are characterized by morphostructures controlled by N-S and NE-SW fault systems. In contrast, E-W faults have ruled the morpholineaments and drainage patterns observed in the southern sector and the transition region.

The SC stratovolcanoes underwent alternated episodes, associated with faulting, of effusive and explosive activity. Porphyritic andesite to dacite lava flows (Lava Dacítica Apilulco; thickness < 4 m) with planar fracturing subparallel to the surface constitute the main effusive products. They generally show a mineralogical assemblage of plagioclase + amphibole + orthopyroxene ± clinopyroxene ± quartz + Fe-Ti oxides. Spherical to ellipsoidal magmatic enclaves occasionally occur in these lava flows. They are randomly distributed along the volcanic range, although the number and size apparently increase towards the north. Majority of the magmatic enclaves display a few millimeters to 4 centimeters in diameter, although in some northern outcrops they reach ~20 cm in diameter. The explosive products consist in pyroclastic deposits (Brecha Piroclástica Cantimplora; thickness = 1–4 m), conformed by dacitic blocks (20–30 cm), pumice clasts (<15 cm), and ash, that occurred intercalated with the lava flows.

Velasco-Tapia et al. [23] developed an extensive study in the SC volcanic range that includes detailed petrography, mineral chemistry, whole-rock geochemistry, and Sr-Nd isotopic data. These authors reported that several disequilibrium features confirm the significant role of the magma mingling/mixing processes between andesitic and dacitic magmas with concomitant fractional crystallization. The SC magmas were probably generated at different levels of the continental crust by partial melting. The magma mixing/mingling evidence includes (a) normal and sieved plagioclases in the same sample, rounded and embayed crystals, and armored rims over the dissolved crystal surfaces; (b) subrounded, vesicular magmatic enclaves, ranging from a few millimeters to ~20 centimeters in size (mineralogical assemblage: plagioclase + orthopyroxene + amphibole + quartz ± olivine ± Fe-Ti-oxides); (c) crystals with reaction rims or heterogeneous plagioclase compositions (inverse and oscillatory zoning or normally and inversely zoned crystals) in the same sample; and (d) elemental geochemical variations and trace element ratio more akin to magma mixing and to some extent diffusion process. Andesitic enclaves have been interpreted as portions of the intermediate magma that did not mix completely (mingling) with the felsic host lavas.

3. Methods

In the present work ten samples, collected along the SC volcanic range (Figure 2; SCN: SC46, SC52, and SC52a; SCS: SC51, SC53, and SC58; SCT: SC03, SC16, SC22, and SC60), were studied to obtain new petrographic and geochemical data. Modal compositions were determined by point counting on thin sections using a Prior Scientific petrographic microscope. Approximately 500 points per sample were counted in order to obtain a representative mode (Table 1).


SampleLocalityLat. (N)°Long. (W)°TexturePhenocrystsGroundmass textureRockDisequilibrium evidence
OlOpxCpxPlgQtzAmpTypeOl + QtzQtz-RPlg-N + SE

SC03Cantimplora99°14′34′′19°11′35′′VP946MI
SC16Rancho Agustín99°19′40′′19°11′30′′P7022620MIDE
SC22Volcán Negro99°23′06′′19°10′00′′VT8515TI
SC46Los Puercos99°28′04′′19°31′36′′P4582018MFDE
SC51Cerro Prieto99°16′55′′19°18′42′′P33591025VFDE
SC52S Miguel Tecpan99°24′33′′19°31′12′′VP2741428MFDE
SC52aS Miguel Tecpan99°24′33′′19°31′12′′VP2377MME
SC53Santiago99°16′53′′19°15′48′′VP69724VFDE
SC58Garambullos99°15′56′′19°15′03′′P3258631VFDE
SC60Quellamecal99°14′25′′19°10′33′′P536230VF

Modal data are percentages of phenocrysts + microphenocrysts calculated on a vesicle and groundmass free basis. Texture: P: porphyritic, VP: vesicular-porphyritic, and VT: vesicular trachytic. Groundmass represents 60–90% of the rocks. Groundmass texture: M: microlithic, T: trachytic, and V: vitreous. Phenocrysts: Ol: olivine, Opx: orthopyroxene, Cpx: clinopyroxene, Plg: plagioclase, Qtz: quartz, and Amp: amphibole. Rock types: I: intermediate magmas without disequilibrium evidence, F: felsic magmas without disequilibrium evidence, IDE: intermediate comingled lava, FDE: felsic comingled lava, and ME: magmatic enclave. Disequilibrium evidences: Ol + Qtz: olivine + quartz, Qtz-R: quartz with a reaction rim, Plg-N + S: plagioclases with normal and sieved texture, and E: ellipsoidal chilled andesitic enclave.

Major and trace element composition of these SC volcanic rocks (Tables 2 and 3) were determined in ActLabs laboratories (Ancaster, Canada), using the “4LithoRes” methodology (for details consult webpage http://www.actlabsint.com/). Major elements were analyzed by inductively coupled plasma-optical emission spectrometry (ICP-OES) with an analytical precision <2% and accuracy typically better than 5% at 95% confidence level, based on analysis of diverse geochemical reference materials (GRM). Trace element concentrations were determined by inductively coupled plasma-mass spectrometry (ICP-MS) with an analytical precision 3–6% (occasionally reaching 9-10%) and an accuracy typically better than 7–12% for most elements at the 95% confidence level, based on analysis of diverse GRM.


SampleSC03SC16SC22SC46SC51SC52SC52aSC53SC58SC60

SectorSCTSCTSCTSCNSCNSCNSCNSCSSCSSCS

TASDA BTADDDBAADD

Major-element measured composition (% m/m)
SiO260.8854.6453.4464.2965.9160.7348.3454.8163.2166.71
TiO20.9810.9121.5420.6610.5830.7851.6361.1280.6950.582
Al2O316.7219.8015.5115.9815.2416.1016.9118.9916.8915.85
4.575.438.664.284.005.439.356.675.003.90
MnO0.0990.0920.1360.0650.0610.0620.1050.1030.0730.062
MgO6.122.236.711.522.022.436.592.341.371.68
CaO5.572.667.433.693.874.445.564.594.303.90
Na2O2.973.473.984.074.014.012.363.684.404.56
K2O0.800.801.532.502.521.891.300.841.682.23
P2O50.140.220.630.180.180.210.190.260.130.17
LOI0.608.950.112.702.303.017.597.013.060.24
Total99.43099.20499.67899.906100.70499.09799.931100.421100.80899.884

CIPW norm
Q18.73023.35820.60021.53016.4765.22316.30218.98820.458
Or4.6215.2609.14215.12315.18211.6728.3865.34210.19413.261
Ab25.52032.67834.05035.53934.59135.46321.79733.51738.23038.839
An27.13213.08920.15517.67716.45121.26528.76222.68421.03916.298
C1.2309.9250.3302.1454.6330.333
Di10.4311.5680.2641.685
Hy18.93311.09318.5437.3057.56010.61726.66812.0757.5796.444
Ol0.202
Mt1.6122.0993.0401.6951.5642.1783.1462.4941.9701.506
Il1.8921.9282.9611.2951.1281.5573.3922.3061.3561.113
Ap0.3290.5681.4760.4310.4240.5070.4800.6490.3080.396
Mg-v77.71951.68466.87148.90557.63454.6663.94047.75242.47353.718
FeOt/MgO0.6722.1911.1612.5331.7822.011.2772.5653.2832.089

TAS: rock classification following the Le Bas et al [36] scheme. A: andesite, BA: basaltic andesite, BTA: basaltic trachyandesite, and D: dacite.
Adjusted composition (% m/m) and CIPW norm calculated applying SINCLAS program [14, 15]. Mg-v = /( ), atomic; and calculated from adjusted FeO and Fe2O3 following Middlemost [34].

SampleSC03SC16SC22SC46SC51SC52SC52aSC53SC58SC60

SectorSCTSCTSCTSCNSCNSCNSCNSCSSCSSCS

TASDABTADDDBAADD

La14.120.034.823.425.224.116.218.811.117.1
Ce31.247.977.841.240.638.240.941.221.834.1
Pr4.037.2810.106.546.126.715.195.142.994.35
Nd17.129.742.926.524.728.424.321.612.617.4
Sm3.96.19.05.44.85.65.54.62.93.6
Eu1.241.772.671.521.371.661.651.571.081.10
Gd3.96.28.04.44.35.05.34.62.93.3
Tb0.60.91.10.70.70.70.80.70.50.5
Dy3.35.25.73.83.74.04.63.92.62.7
Ho0.71.01.00.70.70.80.90.80.50.5
Er1.92.72.92.02.12.22.52.21.51.6
Tm0.280.370.420.300.300.300.360.320.220.22
Yb1.71.82.52.01.91.92.21.91.41.4
Lu0.240.320.400.300.300.290.330.290.230.21
Sc1519119153718138
V1023915088751092161195951
Cr24628260606015036015017040
Co181029109154021159
Ni88110303050110805020
Cu219430102020602020
Ga13222018202122242121
Rb133285961402263858
Sr380303763521502582445569454368
Y20322820211922211518
Zr136156237156160158162194143149
Nb6.05.417.04.04.03.03.08.013.04.0
Cs2.12.82.71.11.22.1
Ba276412648542571481344660388481
Hf3.44.45.44.24.24.44.74.83.84.2
Ta0.400.331.100.50.50.300.200.70.30.6
Pb72111114112310911
Th1.83.04.16.76.74.03.05.03.48.2
U0.61.31.22.62.51.61.11.11.43.1

4. Sierra de las Cruces Database and Evaluation Scheme

4.1. Mineralogical and Geochemical Database

A more complete SC database of the mineralogical modes and the whole-rock geochemical composition was established from the new as well as the published information reported by Velasco-Tapia et al. [23]. CIPW norms for samples were calculated on a 100% anhydrous adjusted basis of major element composition, with ratios adjusted depending on the rock type [34]. Rock classification was based on the total alkali-silica (TAS) scheme [35, 36]. All computations (anhydrous and iron-oxidation ratio adjustments, norm compositions, and rock classifications) were automatically done using the SINCLAS software [14, 15].

4.2. Linear Discrimination Analysis

The tectonic affinity of the SC volcanic rocks was established applying new discriminant-function-based multidimensional diagrams for intermediate ( = 52–63%) and acid ( > 63%) rocks using the linear discriminant analysis (LDA) of natural logarithm ratios of major elements, immobile major and trace elements and immobile trace elements. These diagrams [21, 22] were proposed to discriminate island arc (IA), continental arc (CA), within-plate (continental rift, CR, and ocean island, OI, together), and collisional (Col) settings. Based on the earlier work of Verma and Agrawal [39] and the modifications outlined by Verma [40], these diagrams also provide probability estimates for individual samples, which were used in the present work.

Firstly, the nature of intermediate or acid magma for each sample was confirmed from the SINCLAS software [14, 15], under the Middlemost [34] option for Fe-oxidation adjustment. After that, a series of natural logarithms of element ratios were estimated for all samples. This transformation provided a Gaussian character to the distribution data, a basic condition of the LDA. After that, the ln-ratio data were used to estimate two discriminant functions (DF1 and DF2), obtained from the LDA (canonical analysis), and the individual probability for each sample to a tectonic regime. This statistical exercise was first performed to discriminate between IA + CA, CR + IO, and Col settings and four times for all possible combinations of three groups at a time out of four groups (IA, CA, CR + OI, and Col). Details of the statistical methodology and LDA equations have been reported in [21, 22]. It is important to note that the discrimination analysis was carried out considering the four SC sectors. All LDA equations were incorporated in a STATISTICA for Windows (Statsoft, Inc., Tulsa, OK, USA) spreadsheet and discrimination diagrams were constructed from these results.

4.3. Discordancy and Significance Tests

In order to better understand the contribution of the subducted Cocos plate to the SC magmas, the methodology put forth and practiced by Verma [38] was applied. This approach basically consists of comparing the magmas closer to the Middle America Trench (MAT) to those farther from it; that is, the SC sectors were statistically compared as two groups. The null hypothesis (H0: the two groups did not differ significantly at strict 99% confidence level) and the alternate hypothesis (HA: the two groups differ significantly at 99% confidence level) were tested by Fisher and Student’s -tests (UDASYS software, [37]). Because the significance tests require that the data be normally distributed, single-outlier type discordancy tests were applied at strict 99% confidence level, for which DODESSYS software of Verma and Díaz-González [41] was used.

4.4. Cluster Analysis

The principal aim of this statistical tool is to partition observations into a number of groups. It is expected that the observations within a cluster are as similar as possible, whereas the differences between the clusters are as large as possible. In magma mingling scenario, this technique would be helpful for the SC sample distribution in resident, invasive, and comingled magmas.

In this work, cluster analysis was performed using the natural logarithm of major elements () and representative trace (transition: Co, V; rare earth: La, Eu, Yb; lithophile: Ba, Sr, U; high-field strength: Hf, Y, Zr) elements to ratios by using a hierarchical cluster method (HCM; [42]). Geochemical ratios were previously standardized (z-scores) by means of where is the standardized value of , the th variable for the th sample, is the mean value of the th variable, and is its standard deviation. Additionally, the normality of each standardized variable was confirmed by the Shapiro-Wilks test. Cluster analysis applied a Ward’s linkage rule, which linked iteratively nearby points through a similarity matrix and performed an ANOVA test to evaluate the distance between clusters [43]. The adopted procedure gives equal weight to each geochemical ratio. The measure of similarity was simply the distance as defined in Euclidean space. The distance between two samples is given by where denotes the th variable measured on object in sample and is the th variable measured on object in sample . The results of the cluster analysis were graphically displayed in three dendrograms with units in Euclidean values, corresponding to northern, central, and southern-transition SC sectors.

The weight of geochemical log-ratios in the cluster approach was determined from the results obtained in a principal component analysis (PCA). It has been defined as an orthogonal linear transformation for reducing the dimensionality of a dataset by expressing it as the combination of a small number of linearly independent factors or “principal components.” Each factor will be a function of the individual contributions of the original variables [44]. The greatest variance for the transformed data was linked to the first principal component, whereas the second variance magnitude was related to the second principal component, and so on. The PCA considers a data matrix, ( rows × columns; rows represent different samples, and columns give a particular chemical component; each component which has been standardized yielded a zero empirical mean). The transformation is stated by a set of -dimensional vectors that map each row vector of to a new vector of principal component factors given by

Individual variables of considered over the data set successively inherit the maximum possible variance from , with each loading constrained to be a unit vector. The first principal component satisfied where the quantity to be maximized is known as Rayleigh quotient. The th component was determined by subtracting the principal components from :

The vector associated with this component and showing the maximum variance from this new matrix would be defined as

All calculations related to cluster analysis were carried out using the STATISTICA for Windows software.

4.5. Mass-Balance Evaluations

Nixon [31] applied a simple mass-balance scheme for the quantitative characterization of binary mixtures and end-member compositions in the Iztaccíhuatl volcano (central MVB). The author suggested that, despite the compositional heterogeneity, if a chemical component can be found whose concentration is invariant in time and known in the mix and in each of the end-members, it is possible to treat quantitatively the magma mixing process.

Mixing proportions may be calculated considering the lever principle and the composition of the comingled magma subsequently described for all chemical components. The amount of a component in the mixed lava could be represented by where , and and represent the weight fraction and concentration, respectively, of element in subscripted end-members and and mixture . The composition of an end-member could be estimated by where constituent . In this work, this mass-balance approach (model A) was applied to SC lavas, being restricted to those sectors where the end-member compositions were available and to those components that exhibit a statistically significant linear coherence in -Harker diagrams. This test involved the evaluation, at 99% confidence level, of Pearson product-moment correlation coefficient and the sample size (). Details and required caution in the use of have been reported in Bevington and Robinson [45].

On the other hand, Zou [33] reported a mass-balance approach to explain the and geochemical ratios (where , , , and represent major or trace elements) in SC comingled lavas as a product of a mixture of two components 1 and 2. The variation in the and geochemical ratios could be modeled by the hyperbolic equation (condition ):

In this model, the to coefficients have been defined as

where the geochemical ratios in the components 1 and 2 are

The proportion of the first component could be estimated by

In this work, the scheme described by Zou ([33], model B) was applied to evaluate the mixing/mingling process in the SC northern sector. All calculations of mixing models were carried out using the STATISTICA for Windows software.

5. Results

Ten samples of SC database proved to be intermediate magmas. The set of major element based diagrams (; Table 4 and Figure 3) showed a collisional setting with total percent probability value (% prob) of about 45.8%. However, immobile major and trace element based diagrams (; Table 4 and Figure 4) indicated a within-plate regime, although with a relatively low % prob of only about 38.1. Unlike other sets of diagrams, a continental arc setting can be inferred from those based on immobile trace elements (; % prob = 39.7; Table 4 and Figure 5). It is important to note that intermediate samples from southern and transition sectors (1.9 to 0.5 Ma) represent the main contribution to the collisional and within-plate settings.


Figure nameaFigure typeaTotal number of samplesNumber of discriminated samples
ArcWithin-plateCollision
IA + CA
( )b
IA
( )b
CA
( )b
CR + OI
( )b
Col
( )b

Major elements - -51003  
(0.009–0.767)
7  [
(0.152–0.848)
1-2- 1008  
(0.108–0.944)
2  
(0.015–0.858)
1-2-51005  
(0.068–0.946)
5  
(0.030–0.861)
1- -51003  
(0.011–0.802)
7  
(0.147–0.959)
2- -5104  
(0.159–0.778)
1  
(0.010–0.650)
5  
(0.032–0.756)
All major element based diagrams   
% prob
{50}{0}  {0.0} 
{0}  {0.0} 
0%
{17}  {16.552} 
37.9%
{9}  {7.112} 
16.3%
{24}  {20.061} 
45.8%

Major and trace elements - -593  
(0.005–0.846)
6  
(0.016–0.886)
0
1-2- 91  
(0.004–0.584)
2  
(0.012–0.545)
6  
(0.012–0.984)
1-2-591  
(0.016–0.518)
6  [ ] 
(0.042–0.567)
2  
(0.045–0.942)
1- -593  
(0.005–0.816)
6  
(0.019–0.855)
0
2- -595  [ ] 
(0.002–0.921)
4  
(0.012–0.855)
0
All major and trace element based diagrams   
% prob
{45}{3}  {3.037} 
{5}  {8.023} 
17.8%
{13}  {12.649} 
28.1%
{22}  {17.163} 
38.1%]
{2}  {7.193} 
16.0%

Trace elements - -5106  
(0.019–0.822)
2  
(0.001–0.961)
2  
(0.016–0.933)
1-2- 1007  
(0.026–0.938)
3  
(0.001–0.958)
1-2-5101  
(0.009–0.467)
7  
(0.050–0.676)
2  
(0.090–0.942)
1- -5105  
(0.012–0.785)
3  
(0.002–0.940)
2  
(0.026–0.941)
2- -5107  
(0.026–0.925)
2  
(0.001–0.958)
1  
(0.016–0.838)
All trace element based diagrams   
% prob
{50}{6}  {4.715} 
{6}  {9.459} 
18.9%
{21}  {19.856} 
39.7%
{10}  {8.503} 
17.0%
{7}  {12.178} 
24.4%

“Figure name” corresponds to one of the three sets of diagrams based on major elements, immobile major and trace elements, and immobile trace elements, respectively, whereas “Figure type” gives the tectonic fields being discriminated where the tectonic group numbers are as follows: 1—IA (island arc), 2—CA (continental arc), 3—CR (continental rift) and 4—OI (ocean island) together as within-plate, and 5—Col (collision); —mean one standard deviation of the probability estimates for all samples discriminated in a given tectonic setting; these are reported in .
bProbability estimates for different tectonic groups are summarized after the number of discriminated samples as follows: —range of probability values estimated for IA + CA combined setting, —for IA, —for CA, —for CR + OI, and —for Col. Boldface font shows the expected or more probable tectonic setting; the final row gives a synthesis of results as       %prob where one has the following: —number of samples plotting in all five diagrams which are reported in the column of total number of samples whereas the sum of samples plotting in a given tectonic field is reported in the respective tectonic field column, —sum of probability values for all samples plotting in a given tectonic field which are reported in the respective tectonic field column, and %prob —total probability of a given tectonic setting expressed in percent after assigning the probability of IA + CA to IA and CA using weighing factors. For details of principles, equations, and application rules of the multidimensional diagrams see [21].

A relatively large number of samples () from SC database proved to be of acid magma. In contrast to intermediate magmas, all diagrams indicated a subduction-related setting for the SC acid magmas, with total percent probability values for this tectonic regime of about 74.1%, 63.0%, and 68.7%, respectively, for the major, major and trace, and trace element based diagrams (Table 5 and Figures 6, 7, and 8). The results of the tectonic setting are further evaluated from discordancy and significance tests in the Discussion section below.


Figure nameFigure typeTotal number of samplesNumber of discriminated samples
ArcWithin-plateCollision
IA + CA
( )
IA
( )
CA
( )
CR + OI
( )
Col
( )

Major elements - -54646  
(0.700–0.987)
00
1-2- 46046  
(0.815–0.969)
0
1-2-546046  
(0.666–0.965)
0
1- -54627  
(0.056–0.933)
019  
(0.031–0.804)
2- -54646  
(0.767–0.995)
00
All major element based diagrams   
% prob
{230}{46}  {39.234} 
{27}  {35.511} 
17.2%
{138}  {152.57} 
74.1%
{0}  {0.0} 
0.0%
{19}  {17.938} 
8.7%

Major and trace elements - -54646  
(0.556–0.946)
00
1-2- 463
(0.010–0.651)
43  
(0.324–0.934)
0
1-2-5463  
(0.007–0.671)
43  
(0.296–0.919)
0
1- -54633  
(0.025–0.907)
013  
(0.045–0.699)
2- -54646  
(0.660–0.953)
00
All major and trace element based diagrams   
% prob
{230}{46}  {36.007} 
{39}  {47.423} 
23.4%
{132}  {127.66} 
63.0%
{0}  {0.0} 
0.0%
{13}  {27.658} 
13.6%

Trace elements - -54646  
(0.656–0.973)
00
1-2- 461
(0.003–0.560)
45  [ ] 
(0.440–0.990)
0
1-2-546046  
(0.599–0.913)
0
1- -54616
(0.001–0.953)
030  
(0.047–0.975)
2- -54646  
(0.744–0.996)
00
All trace element based diagrams   
% prob
{230}{46}  {37.178} 
{17}  {31.627} 
14.9%
{137}  {145.56} 
68.7%
{0}  {0.0} 
0.0%
{30}  {34.795} 
16.4%

For explanation see the footnote of Figure 6. For details of principles, equations, and application rules of the multidimensional diagrams see [22].

On the other hand, the hierarchical agglomeration process was carried out for each SC sector (SCN: 22 samples; SCC: 12 samples; and SCS and SCT: 22 samples) and their results were summarized in three dendrograms with units in Euclidean values (Figures 9(a)9(c)). The statistical parameters (mean, minimum, maximum, and standard deviation) associated with the centroid of each cluster are reported in Table 6.

(a)

ElementNorthern SC sector ( )
N1 ( )N2 ( )N3 ( )
Min.Max.s Min.Max.s Min.Max.s

57.652.7760.384.265.163.2567.191.569.067.9469.980.8
1.10.711.790.60.690.590.820.070.5220.4860.5520.031
16.515.3818.461.716.215.5416.830.515.915.1116.490.5
1.681.422.170.421.231.081.500.130.890.830.960.05
5.14.057.231.83.082.703.760.332.222.082.390.11
0.1030.0950.1150.0110.0700.0590.0820.0080.05740.0530.0600.0026
6.76.237.190.52.51.154.170.81.110.6981.5770.30
6.346.076.590.264.43.415.180.53.313.083.790.30
3.32.583.660.64.284.094.530.134.374.204.480.11
1.4421.4191.4620.0222.231.842.610.302.512.292.800.19
0.1640.1430.2070.0370.1930.1510.2920.0360.13440.1300.1410.0043
La14.012.316.22.02013.634.8619.815.123.63.2
Eu1.261.031.650.341.321.042.150.321.070.911.250.12
Yb1.901.602.200.301.761.262.700.401.521.161.700.22
Ba344329358155004145716053047257850
Co2822401012.39.016.02.26.46.07.00.5
Cr290230360709060160343220407
Hf3.52.94.71.03.963.204.600.404.133.84.70.39
Sr459445474155204316016041035156670
Th2.872.803.000.125.33.66.71.26.03.48.21.4
U1.030.901.100.122.101.402.600.442.41.103.000.6
V160123216509275109106052737
Y20.718222.3191234616.112.022.03.2
Zr122100162351461091621515113318418

(b)

ElementCentral SC sector ( )
C1 ( )C2 ( )C3 ( )C4 ( )
Min.Max.s Min.Max.s

58.32965.163.7667.301.961.33864.263.6165.160.5
1.1090.710.630.770.080.7910.690.650.780.05
17.04316.315.8916.840.517.40016.4615.9917.100.40
1.7781.241.081.350.151.4241.291.161.570.13
5.0793.112.703.360.364.0683.232.903.910.33
0.1180.07830.07600.08100.00250.0710.0780.0700.0880.007
3.6632.21.513.060.83.9183.02.313.610.5
6.7644.53.795.230.74.9624.783.975.370.42
4.2904.354.274.490.124.2704.364.164.620.17
1.4972.1632.1362.1800.0231.5921.771.661.950.11
0.3290.210.160.240.050.1650.1570.1410.1650.009
La26.826.023.627.72.211.114.312.119.12.3
Eu2.091.681.452.120.381.101.070.991.340.12
Yb2.301.871.602.100.251.501.481.301.700.16
Ba5075504856026030940036444729
Co2613.010.015.02.61713.912.017.02.0
Hf3.903.733.603.900.153.203.593.403.700.11
Sr81360043468314045346945150721
U1.701.941.802.020.121.201.381.001.700.22
V1508971103161008881997
Y24.019.717.021.02.314.015.613.020.02.2
Zr138137.71371391.21141311231386

(c)

ElementSouthern and transition SC sectors ( )
ST1 ( )ST2 ( )ST3 ( )
Min.Max. Min.Max. Min.Max.

59.354.0361.832.564.663.3065.970.967.464.9269.411.8
1.090.841.560.230.720.630.920.090.610.540.660.05
18.115.6822.042.116.715.7617.750.616.2615.9517.040.44
1.591.112.100.281.281.111.440.121.030.861.170.13
4.53.185.990.83.212.783.610.302.582.142.920.33
0.1050.0790.1380.0180.0710.0520.0880.0140.0540.0220.0790.021
4.02.486.781.72.31.413.510.71.60.492.811.0
5.62.967.511.44.534.045.110.393.93.234.870.7
4.003.024.390.444.514.284.770.144.344.054.590.26
1.330.782.010.431.841.632.090.152.191.942.350.15
0.270.140.640.160.170.110.250.050.1550.1350.1770.018
La1811.534.8714.511.118.22.219.016.925.23.5
Eu1.61.132.670.51.131.041.340.091.241.101.490.16
Yb2.01.32.70.51.521.012.300.341.741.402.100.30
Ba4202766601504163694713746943449925
Co191029512.910.020.03.38.85.012.02.9
Hf4.23.45.40.73.683.304.100.243.903.604.200.28
Sr5003037631304744165334041036449550
U1.040.601.320.281.500.802.000.342.01.53.10.6
V11039150338359961269518412
Y2313.032.2614.911.017.01.82216368
Zr16212923738136125149814812616414

The studied rocks from northern SC sector (Table 6; Figure 9(a)) were distributed in three general clusters (N1 [13.6%], N2 [54.5%], and N3 [31.9%]). The PCA calculation indicated that the ~94.2% of geochemical variability of samples from northern SC sector could be explained by three factors. The factor F1 contributed with 57.4%, being associated with major (excepting Na and P) and transition elements; rare earth elements and yttrium ruled a contribution of 18.6% by means of the factor F2 (Figure 10(a)). The principal component F3 (a function of Na, P, and Sr) explained the 8.2% of the chemical variability.

The samples from central SC conformed four groups (C1 [8.3%], C2 [25.0%], C3 [8.3%], and C4 [58.3%]; Table 6 and Figure 9(b)). A ~94.1% of the chemical variability can be explained by means of five factors. The factor F1 (45.0%) is controlled by Si and alkali composition. A 32.0% of the compositional heterogeneity has been associated with the incompatible elements using the principal component F2 (Figure 10(b)). The factor F3 (ruled by Mg, Ca, and HFSE) contributed with a 10.6%.

The samples of SCS and SCT were agglomerated in three geochemical groups (ST1 [36.4%], ST2 [40.9%], and ST3 [22.7%]; Table 6 and Figure 9(c)). PCA calculations have revealed that a ~90% of the geochemical composition could be explained as a function of five principal components. The factor F1, associated with major elements (excepting Na and K), Co, and Eu, contributed with 42.8%. F2 factor, which represents a 24.7%, is controlled by Ba, K, and U (Figure 10(c)). An 11.9% of the chemical heterogeneity is explained by the factor F3, a variable ruled by Na, K, and V composition.

The mass-balance approach for magma mixing (model A) used by Nixon [31] was applied to the geochemical data from SC northern sector (i.e., intermediate N1 cluster interacting with felsic N3 group resulting in N2 comingled lavas). The mixing analysis was essentially limited to , , , , , , , Co, Cr, Ni, and V, since all these constituents exhibit a statistically significant linear coherence in Harker diagrams (–0.98; ; statistically significant at 99% confidence level; Figures 11 and 12) and have relatively small concentration ranges in felsic N3 end-member (Table 6).

The proportion of the intermediate N1 end-member in each N2 mixed lava was calculated using (7) and the average composition of the intermediate () and felsic () end-members. Calculated proportions exhibit internal consistency for majority of the chemical components (Figure 13). For each sample, the estimated proportions display a Gaussian distribution (their normality behavior was proved by a Schapiro-Wilks test), covering between ~15 and 47% in average proportion of the andesitic N1 end-member (Figure 14).

On the other hand, the mixing model B [33] was applied to lavas of the northern SC sector. The coefficients to ((10a)–(10d)) of the hyperbolic mixing equation (9) were established for twelve geochemical ratio-ratio systems (Table 7):   : /, /, V/Ba, V/U, Cr/Th, and : /;    /Eu, /Hf, /Ta, /Zr, Ga/Ni, and : /V). Figures 15 and 16 show some examples of the ratio-ratio diagrams for the SCN lavas, including the average composition of the intermediate () and felsic () end-members (black filled square and circle) and their hyperbolic mixing models (black solid line). The application of model B revealed that the percentages (100*) of the component N1 in each of the comingled lavas N2 range from 11 to 58% (Figure 14). Each mean and its uncertainty were estimated from a statistic sample of twelve ratio-ratio systems displaying a Gaussian behavior (normality proved by a Schapiro-Wilks test).


Ratio-ratio system -axis -axisHyperbolic mixing equation coefficients

1[Fe2O3]adj/[K2O]adj[SiO2]adj/[FeO]adj0.81−9.6045.164.7
2[Fe2O3]adj/[Al2O3]adj[SiO2]adj/[FeO]adj0.81−44.5−22364.7
3V/Ba[SiO2]adj/[FeO]adj−49.2−193967927584
4V/U[SiO2]adj/[FeO]adj−49.2−9.9567.27584
5Cr/Th[SiO2]adj/[FeO]adj−481−24.214818168
6Cr/Yb[SiO2]adj/[FeO]adj−481−3.53−43.518167
7[MgO]adj/Eu[SiO2]adj/V−224−96−25.3398
8[MgO]adj/Hf[SiO2]adj/V−225−451−3.61398
9[CaO]adj/Ta[SiO2]adj/V149−6612.7247
10[CaO]adj/Zr[SiO2]adj/V149−16840280247
11Ga/Ni[SiO2]adj/V20382600−5518167
12Ga/Rb[SiO2]adj/V2037−82801685167

6. Discussion

6.1. Tectonic Setting

The MVB (Figure 1) has been considered as a very tectonically complex zone. In the framework of the theory of plate tectonics, the origin of this volcanic province has been explained by means of the subduction of Cocos and Rivera plates under the North American plate. However, several geological, geophysical, and geochemical characteristics observed in central MVB and the entire province do not support this simple model. Particularly, a strong controversy regarding the tectonic regime has been widely documented in the literature (e.g., [29, 30, 38, 4653]).

How to interpret the seemingly contradictory results obtained in the tectonic discrimination analysis for the SC magmas (Tables 4 and 5)? A transitional continental arc to within-plate setting can be tentatively considered as a consistent model for the central MVB. Felsic magmas display geochemical features consistent with an origin from the upper continental crust. The genesis of the majority of the Mexican crustal source rocks has been associated with continental arc regime. Afterwards, a change in the tectonic setting could be related to a relatively fast variation in the Cocos plate subduction angle.

However, the Cocos plate tectonic evolution is an issue that has not been solved. Pérez-Campos et al. [54] pointed out that the history of volcanism has been used to infer the evolving geometry of subduction. According to this model, during earlier Eocene the volcanic arc in central Mexico was nearer to the coast and parallel to the trench consistent with steep subduction. In late Eocene (30 Ma) there was a hiatus, thought to be associated with a flattening process. At 20 Ma, after a 10 Ma lull, volcanic activity resumed. At ~10 Ma, the western part of the Cocos plate separated to form Rivera plate. At about this time, the development and propagation of a tear in the subduction plate have been suggested, culminating with the lower portion of the Cocos plate breaking off. The west-east propagating volcanism along the MVB reached the longitude of Mexico City at about 7 Ma. Additionally, Peláez Gaviria et al. [55] have reported changes during the last 3.5 Ma in the plate configuration at the north of the Middle America Trench (MAT) as a result of (a) the propagation of the Pacific-Cocos Segment of the East Pacific Rise (EPR-PCS), (b) the collision of the EPR-PCS with the MAT at 1.7 Ma, and (c) the formation of the Rivera Transform.

Actually, subhorizontal subduction of Cocos plate has been inferred by Pérez-Campos et al. [54], Husker and Davis [56], and Pacheco and Singh [57] from seismic data obtained from a dense network. Particularly, the dip angle of Cocos slab decreases gradually from ~50° to 0° along the labeled Michoacan segment of the Mexican subduction zone [57]. However, this quasihorizontal subduction and a very shallow subducted slab (at most at about 40 km in depth) are not thermodynamically favorable conditions for arc-related magma generation [58].

The diminution or even cessation of arc-related volcanism observed in the south-central Andes has been related to subhorizontal subduction of the Nazca plate [59]. The SC intermediate rocks could be a volcanism generated under this complex condition of the tectonic transition to an extensional regime. Additionally, Velasco-Tapia and Verma [29] have inferred, from inverse and direct immobile trace element modeling, combined 87Sr/86Sr and 143Nd/144Nd isotopic ratios, and the use of multidimensional log-ratio discriminant-function-based diagrams, that mafic magmas from the Sierra de Chichinautzin (the post-SC volcanic event of <40 ka) were undoubtedly generated by partial melting of continental lithospheric mantle in a within-plate setting.

Although the previous studies and this work represent significant contributions to the understanding of the origin of the volcanism in the central MVB, more geological-geophysical-geochemical collaborative research is needed to clearly understand the evolution of the tectonic regime in this area and the entire MVB.

6.2. Application of Discordancy and Significance Tests

The acid rock data of SC were placed in two groups: Gr1 close to the MAT (consisting of the data from the southern and transition sectors) and Gr2 farther away from the MAT (data from the northern and central sectors). A statistical comparison of these groups was carried out using Fisher test and Student’s -test. The results are summarized in Table 8. No statistically significant difference was observed between the two groups for any of the elements listed in Table 8 (see true for all elements in both one-sided and two-sided columns of Table 8). The same is true for the Nb-anomaly as well as for ratios of large-ion lithophile elements (LILE) to light rare earth elements (LREE) and LILE to high-field strength elements (HFSE) (see [38] for the importance of these ratios for subduction processes). Therefore, the negligible contribution from the subducted slab to the SC magmas can be safely inferred. The intermediate rock data were not so numerous and, therefore, are not reported here, although they confirmed the results for acid rocks.


ElementGroup AGroup B DfSign _calc _criteria
One-sided
H0
One-sided
CL_
One-sided
_criteria
Two-sided
H0
Two-sided
CL_
Two-sided

(a) Major elements
Gr 2Gr 1321141.00.4712.421True 502.701True 50
Gr 2Gr 1321141.00.5092.421True 502.701True 50
Gr 2Gr 1321141.00.6872.421True72.52.701True44.9
Gr 2Gr 1321141.0+0.5712.421True 502.701True 50