The ∼AD 1250 effusive eruption of El Metate shield volcano (Michoacán, Mexico): magma source, crustal storage, eruptive dynamics, and lava rheology

Abstract

Medium-sized volcanoes, also known as Mexican shields due to their andesitic composition and slightly higher slope angles in comparison to Icelandic shields, occur across the Trans-Mexican Volcanic Belt and represent nearly one third of all volcanic edifices in the Michoacán-Guanajuato Volcanic Field (MGVF). Many questions about their origin and eruptive dynamics remain unanswered. Here, we focus on El Metate, the youngest (∼AD 1250) monogenetic shield volcano of the MGVF and the most voluminous (∼9.2 km³ dense rock equivalent) Holocene eruption in Mexico. Its eruptive history was reconstructed through detailed mapping, geochemical analysis (major and trace elements, Sr-Nd-Pb isotopic data), and rheological study of its thick andesitic flows. Early and late flow units have distinct morphologies, chemical and mineralogical compositions, and isotopic signatures which show that these lavas were fed by two separate magma batches that originated from a heterogeneous mantle source and followed distinct differentiation paths during their ascent. Thermobarometry calculations constraining the conditions of crystallization indicate a temporary storage of the last erupted magma batch at a depth of ∼7–10 km. Lava rheology was estimated using petrographic characteristics, geochemical data, and flow dimensions. The magma viscosity increased from 10²–10³ Pa s prior to eruption through 10⁶–10⁸ Pa s during ascent, to 10⁹–10¹¹ Pa s during lava emplacement. Though magma viscosity was quite high, the eruption was purely effusive. The explosive eruption of such a large magma volume was probably avoided due to efficient open system degassing (outgassing) of the magma as it ascended through the uppermost crust and erupted at the surface.

Introduction

Small- to medium-sized shield volcanoes are an important component of many volcanic fields on Earth (e.g., Idaho, USA, Greeley 1982; Hawaii, USA, Decker et al. 1987; Michoacán, Mexico, Hasenaka 1994; Iceland, Rossi 1996; the Cascades, USA, Hildreth 2007) and other planets (e.g., Moon, Guest and Murray 1976; Venus, Guest et al. 1992; Mars, Baptista et al. 2008). The Trans-Mexican Volcanic Belt (TMVB), one of the most complex and active continental arcs worldwide, displays a large number of such medium-sized volcanoes. In particular, they represent nearly one third of the ∼1500 edifices of the ∼40,000-km² Michoacán-Guanajuato Volcanic Field (MGVF, Fig. 1), which is not only the largest monogenetic field in the TMVB (Hasenaka and Carmichael 1985a) but also of the entire Earth (Valentine and Connor 2015). Although the exact causes responsible for the great number of monogenetic volcanoes in the MGVF (which also happens to be the region where the TMVB reaches its greatest width) are still poorly understood, their high frequency of occurrence may be related to the unique geometric configuration of the subduction zone (Pardo and Suárez 1995; Gómez-Tuena et al. 2007; Johnson et al. 2009; Blatter and Hammersley 2010; Kim et al. 2012) that controls the location and size of magma generation areas and the magnitude of crustal extension. We speculate that the large number of monogenetic volcanoes over such a wide area might be related to the near-horizontal position at a depth of 90–120 km of this segment of the subducting oceanic Cocos plate underneath the continental North America Plate (Kim et al. 2012). Such a low subduction angle at depth might be inducing partial hydrous melting of the mantle wedge over a wide area underneath a ∼40-km-thick continental crust, instead of favoring the more common case in which magmas are generated and repeatedly rise along a much narrower zone leading to the formation of a chain of stratovolcanoes.

Fig. 1

Digital elevation model of the Michoacán-Guanajuato Volcanic Field (MGVF, outlined in red) showing the location of El Metate shield volcano. Yellow rectangle indicates the study area shown in detail in Fig. 2. Small-sized and medium-sized volcano database modified after Hasenaka (2009, personal communication). Inset map at the lower right corner shows the location of the MGVF within the Trans-Mexican Volcanic Belt (TMVB)

The origin of the magmas, the eruption dynamics, and the distribution of volcanic edifices (especially of scoria cones) in the MGVF have been the focus of many previous studies (e.g., Williams 1950; Fries 1953; Hasenaka and Carmichael 1985a, b, 1987; Connor 1987, 1990; Hasenaka et al. 1994; Ownby et al. 2007, 2011; Gómez-Tuena et al. 2007; Johnson et al. 2009; Guilbaud et al. 2009, 2011, 2012; Pola et al. 2014; Siebe et al. 2014), yet there are relatively few studies focusing on the medium-sized volcanoes (Hasenaka and Carmichael 1986; Roggensack 1992; Ban et al. 1992; Hasenaka et al. 1994; Hasenaka 1994). The volumes of the MGVF shield volcanoes vary between 0.5 and 10 km³ and are considerably larger than those of typical scoria cones (average of 0.021 km³) but smaller than those of the two stratovolcanoes (Tancítaro and Patamban) in this field (∼50 km³, Hasenaka 1994). These medium-sized volcanoes have generally slightly steeper slopes (5–15°) than Icelandic shields (Hasenaka 1994; Roggensack 1992; Whitford-Stark 1975) and distinctly more evolved compositions (basaltic-andesitic to andesitic), which led to their designation as Mexican shields (Hasenaka 1994). Although outnumbered by smaller volcanoes (mostly scoria cones), they nevertheless represent nearly 70 % of the total volume erupted since 1 Ma (Hasenaka 1994) and hence played a considerable role in the formation of the MGVF. However, the source, storage, and transport as well as the physical properties (density, viscosity, volatile content, etc.) of the magma involved in these eruptions remain poorly constrained.

In this paper, we provide new insights on magmatic processes during the ∼AD 1250 eruption of El Metate, which is both the most recent monogenetic shield in Mexico and the largest andesitic effusive eruption during the Holocene worldwide (Chevrel et al. 2016). The only previous geochemical study of El Metate lavas (Losantos et al. 2014) did not specify sampling locations and reported only major element data. Here, we first summarize the main results of a recent study focused on the age, volume, chronology, and impact of the eruption (Chevrel et al. 2016) and then present new data on lava morphology. Next, we describe in detail the geochemistry (major, trace elements, and Sr-Nd-Pb isotopic data) and petrology of the distinct lava units and use this information to constrain the magma source, storage conditions, and lava rheology and to propose an eruptive model.

Age, volume, and chronology of lava flow emplacement

El Metate volcano (N 19° 32′ 19″; W 101° 59′ 34″, 2910 m asl, Fig. 1) erupted a volume of 9.2 km³ (dense rock equivalent) of andesitic lava (Chevrel et al. 2016). Detailed fieldwork shows that the eruption was purely effusive since neither ash nor spatter deposits from El Metate were found in its vicinity (Chevrel et al. 2016). If El Metate had initiated with a Strombolian phase (with ash deposits now mostly buried below the thick lava flows), the old edifice that crops out ∼1.5 km south of El Metate’s summit (Fig. 2) should be covered by several meters of fresh coarse ash. However, we only found a 70-cm-thick strongly altered scoriaceous ash fallout (dated at 13,480 ± 50 years BP that predates significantly El Metate’s eruption) under a well-developed clayey soil.

Fig. 2

Digital elevation model of El Metate shield volcano showing emplacement sequence (1–13) of the different lava flows and surrounding volcanoes. Sampling locations of analyzed rocks are indicated

In addition, paleosols are absent between the different flow units. The latter also display similarly well-preserved surface morphologies, indicating that the entire edifice was formed during one single eruptive event. The eruption took place ∼AD 1250, and since it is absent from colonial chronicles, it must have ended well before the invasion of the Spaniards that occurred only 275 years later in the 1520s. Because of the large area covered by the lava flows (103 km²) and the concomitant disruption of the regional hydrological network, local human populations might have been forced to migrate to the neighboring lacustrine basins (Zacapu and Pátzcuaro) in the east contributing to the social processes that led to the rise of the Tarascan empire that initiated around that time (Chevrel et al. 2016).

The morphology of El Metate is remarkably well preserved and exposed (Fig. 3). The volcano is composed of a central dome (1 km wide, 300 m high) from which several voluminous lava flows that form distinct units radiate (Fig. 2). Lava flows directed to the south reached longer distances (<15 km) than flows emitted to the N, due to different topographical gradients and the presence of the older prominent Paracho shield to the N. For this reason, the ∼10-km wide and ∼600-m high shield is slightly asymmetric. Thirteen flow units were identified from their stratigraphic relations (Fig. 2). Lava flow units F1 to F4 are clearly below lava units F5 to F13, and thus, these form the early and late lava groups, respectively. The lateral contacts of early lava units are however not exposed; thus, their relative time of emplacement is not known. The proposed numeration from F1 to F4 is hence tentative, except for F4b, which issued from unit F4 and thus must be younger than F4. By contrast, the relative time of emplacement of late lavas can be much better constrained due to clear contacts between the flows (Fig. 2) and sharp compositional differences between stratigraphically lower F5 to F8 units and stratigraphically higher F9 to F13 units. Nevertheless, the relative order between flow units F6 and F7 is only partly known and their composition is similar (see below); hence, their numeration is also tentative (these uncertainties do not affect our final eruptive model, see below).

Fig. 3

Aerial view of El Metate (2910 m asl) from the east with the stratovolcano Tancítaro in the background and the village of Turícuaro in the foreground. Different El Metate lava flow units (F4, F4b, F7, F8, F11, and F12) are indicated. Photo taken February 7, 2010

Lava flow morphology

The early and late lavas are morphologically distinct. The fronts of the early flows average <40 m thick and their surface texture ranges from rubbly-‘a’ā to blocky in contrast to late flows that have thick (<150 m) fronts and blocky surfaces (terms used as defined in Kilburn 2000). In comparison with early lavas whose proximal surfaces are not exposed, most of the late flows can be traced from source to front where some flows split into multiple lobes. These flows have well-defined open channels separated from wide flat-topped margins (levées) by a clear shear zone that is continuous from source to front (Fig. 4). The channels are typically narrow near the source and broaden downslope, as the topographic gradient decreases, forming fan-shaped or straight flows depending, respectively, on whether they flowed on a near flat or more inclined topography (Fig. 4). The lava surfaces inside the channels display thick and widely spaced cross-flow ridges (ogives).

Fig. 4

Schematic representation of major lava flow features. a Oblique view from the NW of the fan-shaped flows F8 and F11. b Vertical view of the straight channel flow F6

The morphological features observable at El Metate are typical for highly viscous silicic coulées, e.g., the Chao dacitic flow (de Silva et al. 1994), the coulées at San Pietro Island (Cioni and Funedda 2005), or the rhyolitic obsidian flow of Cordón Caulle (Tuffen et al. 2013). Nevertheless, it is worth mentioning that the size of the El Metate lava flows is exceptionally large. The longest (F6) is on average 15 km long, <4 km wide, and ∼70 m thick, corresponding to a volume of 2.2 km³, and the thickest (F11) is on average 5 km long, ∼2 km wide, and 150 m thick, corresponding to a volume of 1.8 km³. These flows are among the most voluminous single lava lobes so far reported (Wadge and Lopes 1991). Flows of similarly large sizes are usually more silicic [1–3 km³ San Pedro andesitic flow, Chile (Francis et al. 1974); 4.4 km³ Payun Matru trachytic flow, Argentina (González Díaz 1972); 2.1 km³ Copales dacitic lava flow from Ceburoco, Mexico (Sieron and Siebe 2008); 1–2 km³ dacitic to rhyolitic Big Glass Mountain, USA (Eichelberger 1975); 24 km³ Chao dacitic flow, Chile (Guest and Sanchez 1969)]. Thus, El Metate produced some of the most voluminous andesitic flows that have been reported, and their young age has made it possible to study them in detail.

Sampling and analytical methods

Rock samples were collected from the interiors of surface blocks, to avoid alteration. We first sampled the flow fronts that were of easy access and then collected rock samples during long hikes along the flow margins or across their interiors where possible (Fig. 2). Twenty-seven whole rock samples from El Metate were analyzed for major and trace elements at Activation Laboratories Inc., Ancaster, Canada (analytical methods, including detection limits and analytical uncertainties, are reported in Table 1). Thin sections from each lava unit were studied under a polarizing microscope. Modal mineralogical analyses were carried out with the aid of an automated stage. Mineral compositions were determined using a JEOL JXA-8900R electron microprobe at the Laboratorio Universitario de Petrología (LUP), Instituto de Geofísica, UNAM, Mexico City. The measuring conditions were set at an accelerating voltage of 20 kV and a beam current of 20 nA (with a diameter of 1 μm), and counting times were 40 s for all elements. Elements were calibrated from an assortment of mineral standards taken from SPI supplies® reference standards for X-ray analysis (SPI# 2757-AB). Sr-Nd-Pb-isotope ratios were determined for six samples representative of the distinct lava groups at the Laboratorio Universitario de Geoquímica Isotópica (LUGIS), Instituto de Geofísica, UNAM. Sr-Nd isotope ratios were determined with a Thermo Scientific Triton Plus thermal-ionization mass spectrometer (TIMS) equipped with nine Faraday cups, and Pb isotopes were determined with a Finnigan MAT 262 TIMS equipped with eight Faraday cups. All the measurements were done in static mode. Further details of the analytical procedures can be found in Table 2 and in Schaaf et al. (2005).

Table 1 Whole rock major (in wt%) and trace element (in ppm) analyses of rock samples from El Metate lavas

Table 2 Sr-Nd-Pb isotope ratios of representative El Metate samples

Results

Mineralogy and petrography

Volume percent of the different phases present in the samples (crystals, groundmass, vesicles) and average mineral and glass compositions are reported in Tables 3 and 4, respectively. All lavas contain plagioclase as phenocrysts (0.7–16.9 vol%) and microphenocrysts (12.5–30.8 vol%) and augite and hypersthene often as phenocrysts (<3 vol%) and as microphenocrysts (0.2–13.5 vol%) embedded in a light-brown groundmass consisting of microlites of the same mineral phases plus oxides and minor interstitial glass (Table 3). In addition, the early lavas F1 and F4b contain 2.4–6.6 vol% olivine phenocrysts, in contrast to all late lava units (F5 to F13 and dome) that contain 1.2–6.3 vol% hornblende phenocrysts and lack olivine. The other early lavas (F2, F3, and F4) do not contain olivine or hornblende. Vesicles represent <9 vol% of the lavas, except in rocks from the summit dome area and from late-stage proximal lavas (F11 to F13) that contain 11 to 29 vol% vesicles.

Table 3 Modal mineralogical analyses (vol%) of El Metate rock samples

Table 4 Average mineral and glass compositions of El Metate lavas

Olivine phenocrysts are up to 1 mm long, euhedral to subhedral, and show resorption textures with reaction rims of pyroxene (Fig. 5a, b). Pyroxene phenocrysts are <0.5 mm, they occasionally show disequilibrium textures, and the microphenocrysts are often found as clusters. Hypersthene microphenocrysts are elongated and have oxide inclusions. Plagioclase phenocrysts are tabular, <0.5 mm long, and frequently oscillatory-zoned and commonly contain inclusions of glass, apatite, pyroxene, and rarely opaques, which were not observed inside the acicular microphenocrysts and microlites. Hornblendes are <1 cm long and show variable morphology. Some are pristine and others present disequilibrium textures. Some of the crystals have a rim made of a microgranular assemblage of pyroxene + plagioclase + oxide + glass and/or opacite (Fig. 5c, d), whereas others are fully replaced by these materials, forming mineral ghosts (Fig. 5d). Enclaves consisting of coarser tabular plagioclase + destabilized pyroxene + melt are rarely found.

Fig. 5

Photomicrographs under crossed polars of El Metate lavas: a olivine (Ol) phenocryst with reaction rim of microcrystalline pyroxene (sample no. 14286, F1). b Cluster of olivine and augite (Cpx) phenocrysts (sample no. 1205, F4b). c Brown hornblende with thick microcrystalline reaction rim (gabbroic-type; sample no. 14268, F5). d Hornblende (Hbl) with opacite-rim (black-type) and hornblende ghost embedded in a seriate groundmass consisting mostly of plagioclase (Plag) and some orthopyroxene microlites and glass (sample no. 14287, F6)

Major and trace element contents

Based on the classification of LeBas et al. (1986), El Metate’s lavas are calc-alkaline basaltic andesites to andesites (57 to 62 wt% SiO₂, Table 1). Our results are in general agreement with previously published data (Fig. 6, Hasenaka et al. 1994; Losantos et al. 2014).

Fig. 6

Total alkalies (Na₂O + K₂O) vs. SiO₂ diagram after LeBas et al. (1986) for all analyzed volcanic rocks. Data from Losantos et al. (2014; the two different symbols represent two distinct groups representing different crystallization trends) and Hasenaka et al. (1994) for El Metate lavas are also plotted

In spite of their similar SiO₂ and Na₂O + K₂O contents (Fig. 6), early and late lavas display distinct trends in major element plots (Fig. 7). Late lavas have distinctly higher Al₂O₃, CaO, Na₂O, and Sr and lower Fe₂O_3T, MgO, TiO₂, and Ni at a given SiO₂ weight percent content than early lavas, although the most evolved units of both groups have similar compositions. In early lavas (F1 to F4), Fe₂O_3T, MgO, TiO₂, and Ni sharply decrease while CaO slightly decreases, Al₂O₃ displays a scattered pattern, and K₂O increases constantly with increasing SiO₂. In contrast, late lavas (F5 to F13) show a tendency to decrease in Al₂O₃, a sharp decrease in CaO, weak decreases in Fe₂O_3T and MgO, constant Ni, and variable K₂O with increasing SiO₂.

Fig. 7

Harker diagrams showing selected major (in wt%) and trace (in ppm) elements plotted against silica. Data from Losantos et al. (2014; the two different symbols represent their two crystallization trends) and Hasenaka et al. (1994) for El Metate lavas are also plotted. Arrows represent modeling of fractional crystallization for both lava groups (proportions in wt%): model 1: 3.3 olivine, 5.5 plagioclase, 0.3 clinopyroxene, 1.6 Ti-Fe oxide, 0.18 apatite; model 2: 2.7 olivine, 8.8 plagioclase, 0.9 clinopyroxene, 0.7 Ti-Fe oxide; model 3: 8 hornblende, 8 plagioclase, 4 cpx, 0.2 apatite; model 4: 6 hornblende, 10 plagioclase, 1.1 Ti-Fe oxide, 0.35 apatite

The trends observed in late lavas indicate a general progressive increase in the degree of differentiation of the flows with the emplacement sequence from F5 to the formation of the final summit dome. Also, sampling along the longest lava units (F5, F6) reveals some systematic variations in major elements with distance from the vent (<2 wt% SiO₂). As noted above, the emplacement sequence could not be constrained for early lavas, except that unit F4b was emplaced by a late breakout from unit F4. Unit F4b is less evolved than F4, the opposite of what would be expected if differentiation increased over time.

Interestingly, Sr contents in the most mafic (<59 wt% SiO₂) late lavas (F5, F6, F7, F8) are abnormally high (>1300 ppm), decrease in more evolved subsequently erupted units (F9, F10, F11), and are lowest in the most silicic and last erupted lavas (F12 and F13 with ∼800 ppm) which are similar in this respect to early lavas (F2–F4; Fig. 7).

Early and late lavas have also distinct trace and rare earth element (REE) patterns (Fig. 8). Although both groups span similar ranges in light REE and are depleted in heavy REE, the late lavas show a stronger depletion (Fig. 8a). Also, late lavas show a marked decrease in REE with increasing SiO₂, whereas early lavas show constant to slightly increasing REE with SiO₂ (La shown as an example in Fig. 7). The elevated Sr and lower Y contents of late lavas yield higher Sr/Y ratios (60–140) than early lavas (<50). Both lava groups show trace element patterns typical for arc lavas but early lavas are less depleted in Ta and Nb and slightly more enriched in large ion lithophile elements (LIL) than late lavas (Fig. 8b). None of the samples shows a Eu anomaly.

Fig. 8

a Chondrite-normalized (Sun and McDonough 1989) REE compositions of analyzed rocks from El Metate shield. b Trace element abundances of analyzed rocks from El Metate shield normalized to primordial mantle (McDonough et al. 1992)

Isotopic composition

The first Sr-Nd-Pb isotope dataset of six samples representative of the chemical diversity of El Metate lavas (three samples from each lava group) is reported in Table 2. The variations measured are significant and nearly cover the entire range of MGVF products (Fig. 9). Early lavas are isotopically quite homogeneous (⁸⁷Sr/⁸⁶Sr 0.704007–0.703767, ¹⁴³Nd/¹⁴⁴Nd 0.512789–0.512819, ²⁰⁶Pb/²⁰⁴Pb 18.63836–18.68909, ²⁰⁷Pb/²⁰⁶Pb 15.5831–15.5912), while the late lavas cover a wider range (⁸⁷Sr/⁸⁶Sr 0.703127–0.703727, ¹⁴³Nd/¹⁴⁴Nd 0.512797–0.512942, ²⁰⁶Pb/²⁰⁴Pb 18.54352–18.62874, ²⁰⁷Pb/²⁰⁶Pb 15.5547–15.5766). The small isotopic variations of the early lavas do not correlate with SiO₂. In contrast, variations in late lavas correlate with the inferred eruption sequence and SiO₂ contents. Late lavas become more radiogenic and more SiO₂ rich over time, and the last, most silicic eruptive unit (F12) is isotopically similar to the early lavas (Fig. 9a). All the data follow a linear trend in plots of ⁸⁷Sr/⁸⁶Sr ratios vs. εNd (Fig. 9c) and ²⁰⁶Pb/²⁰⁴Pb vs. ²⁰⁷Pb/²⁰⁴Pb (Fig. 9d). The earliest unit of the late lavas (F5) is only slightly less radiogenic than the subducting oceanic crust (MORB-COCOS), whereas the last erupted lava (F12) and the early lavas (F1, F4, F4b) both plot toward values reported for the granodioritic upper crustal basement.

Fig. 9

Isotopic ratios (Sr-Nd-Pb) for six selected samples from El Metate (see also Table 2). a ⁸⁷Sr/⁸⁶Sr ratio vs. SiO₂. b ⁸⁷Sr/⁸⁶Sr ratio vs. ¹⁴³Nd/¹⁴⁴Nd. c εNd vs. ⁸⁷Sr/⁸⁶Sr ratio. d ²⁰⁶Pb/²⁰⁴Pb vs. ²⁰⁷Pb/²⁰⁴Pb. Field for the MORB-COCOS rocks after Class and Lehnert (2012), MGVF field after Blatter and Hammersley (2010), oceanic sediment data from Verma (2000), and granodiorite analysis from sample of the local basement in the Jorullo area. DM depleted mantle, EM enriched mantle. Red arrows indicate the temporal evolution of the late lavas

Thermobarometry and hygrometry

Crystallization pressure (P), temperature (T), and water content (H₂O) were estimated for the different lava units. We applied the olivine-liquid thermometer (based on Beattie 1993) and the two-pyroxene thermobarometer (based on Brey and Köhler 1990) using the equations of Putirka (2008), the hornblende thermobarometer and hygrometer from Ridolfi et al. (2010), and the plagioclase-liquid hygrometer from Waters and Lange (2015). Details of the methods and tests for equilibrium are given in Online resource 1, and the results are presented in Table 5 and Fig. 10. Assuming a maximum pressure of crystallization of 8 kbar (∼24 km; Hasenaka and Carmichael 1987), dry conditions (olivine composition is weakly affected by H₂O), and equilibrium between mineral core and bulk rock, olivine phenocrysts in the most mafic early lava (F1) are constrained to have formed at 1176 ± 35 °C, whereas olivines in more evolved early lava (F4b) crystallized at 1135 ± 35 °C. The lowest temperature of olivine growth was estimated at 1065 °C for the most mafic early lava (F1) and 1081 °C for the more evolved early lava (F4b), assuming dry conditions, a pressure of 5 kbar (corresponding to the highest pressure estimated for pyroxene crystallization, Fig. 10), and using the compositions of olivine rims and a residual melt calculated by subtracting the mineral core composition from the bulk rock. Two-pyroxene microphenocrysts in the most mafic early lava (F1) formed at 1076 ± 35 °C and between 2.5 and 4.4 ± 2.8 kbar, whereas pyroxenes forming rims around olivine yield slightly higher T (∼1100 °C) and low P (1 atm). Pyroxenes in more evolved early lavas (F4 and F4b) formed at lower T (in average 988 ± 35 °C) and over a larger P range (2.3–5.5 ± 2.8 kbar). In comparison, two-pyroxene phenocrysts and microphenocrysts in the late lavas formed over an even wider P-T range that extends to lower P and T (0–4 ± 2.8 kbar; 938–987 ± 35 °C). [Note that the Putirka (2008) model is also valid for hydrous melts (K. Putirka, personal communication)]. The crystallization conditions of hornblende in late lavas are constrained at 2.7–3.1 ± 0.4 kbar and 922–972 ± 22 °C. Results from the hornblende hygrometer yield a water content of 5.5–6.1 ± 1 wt% H₂O. Two-pyroxene microlites at hornblende rims formed at a P-T range of 0–3 ± 2.8 kbar and 910–1040 ± 38 °C. Water contents were also estimated with the plagioclase-liquid hygrometer using the plagioclase phenocryst compositions. For early lavas, we used on the one hand the previously calculated crystallization T of olivine cores at 8 kbar and bulk rock composition and, on the other hand, the previously calculated crystallization T of olivine rims at 5 kbars and a recalculated residual melt composition (see above and Table 5). Both condition sets yield 1.9 to 3.5 ± 0.35 wt% H₂O. In the case of the late lavas, this method yields 4.5 to 5.5 ± 0.35 wt% H₂O at the P and T conditions obtained from the hornblende thermobarometer and considering equilibrium of the minerals with the bulk rock.

Table 5 Results of thermobarometry and hygrometry

Fig. 10

Pressure-temperature diagram showing results of different mineral-equilibrium thermobarometry methods applied in this study (see also Table 5). Data obtained by the olivine-liquid thermometer (assuming dry condition and pressure at 8 and 5 kbars), the two-pyroxene thermobarometer, and hornblende thermobarometer are plotted as circles, diamonds, and squares, respectively. Representative uncertainties for each model are indicated. The errors given here are inherited from the models applied. Arrows indicate P-T path of early (blue) and late (red) lavas

In summary, mineral-melt equilibria indicate that the magmas feeding the two lava groups followed a distinct crystallization history during ascent. The magma feeding the early lavas crystallized olivine at high T and plagioclase at moderate water contents (<3.5 wt% H₂O), followed by minor pyroxene crystallization. In contrast, the late lavas were fed by a magma that crystallized hornblende, pyroxene, and plagioclase at lower T (900–1000 °C) and higher H₂O (<6 wt%). In both cases (early and late lavas), mineral crystallization depths are poorly constrained (0–15 km) except for hornblendes in the late lavas, which formed at depths of ∼7–10 km.

Magma and lava rheology

The viscosity of the magma and lava can be estimated on the basis of petrological and geochemical parameters (petrological approach, see details in Online resource 2). This approach considers that the interstitial melt has a Newtonian rheology that depends on T, melt composition, and H₂O content, while the content and shape of crystals and bubbles account for the deviation of the mixture from this behavior. Here, the melt viscosity is calculated using the model of Giordano et al. (2008). The effect of crystals is taken into account considering two populations of crystals and applying the equation of Krieger and Dougherty (1959); Eq. 2, Online resource 2) with the fitting parameters of Mueller et al. (2010). The effect of vesicles is complex (Manga et al. 1998) and briefly discussed below.

Using this approach, the pre-eruptive viscosity of the magma that fed the lava flows was estimated from a recalculated melt composition (calculated by subtracting the composition of the phenocrysts from the whole rock, Online resource 3), the H₂O content obtained from hygrometry, the crystallization T of phenocrysts obtained from geothermometry, and the amount of phenocrysts in the lavas. The calculated viscosities (Table 6; Online resource 4) vary from 3 × 10² Pa s for the most mafic early lavas (F1, 1176 °C, 1.9 wt% H₂O, 19 vol% crystals) to 7 × 10³ Pa s for the most evolved early lavas (F4, 989 °C, 1.9 wt% H₂O, 9 vol% crystals) and from 7 × 10² to 10³ Pa s for the late lavas (922–972 °C, 4.5–5.5 wt% H₂O, 6–19 vol% crystals).

Table 6 Magma and lava viscosities estimated from petrological data and groundmass composition

The viscosity of the lavas upon eruption was calculated using the same approach, assuming a low dissolved water content (0.1 wt% H₂O), the crystallization T of microphenocrysts, and the amount of microphenocrysts + phenocrysts in the lava. The results vary between 3 × 10⁶ and 2 × 10⁷ Pa s for the early lavas (987–1076 °C, 40–50 vol% crystals) and between 4 × 10⁷ and 10⁸ Pa s for the late lavas (922–972 °C, 31–48 vol% crystals; Table 6, Online resource 4). These are minimum values since the amount of microlites has not been taken into account. The lower viscosity value of the early lavas compared to the late lavas is in agreement with the observed thinner flow front of the early lavas in contrast to the thicker blocky front of the late lavas. Assuming dry interstitial melt increases the values by 0.2 to 0.4 log units depending on the SiO₂ content. Proximal lavas contain up to 30 vol% bubbles, which can increase or decrease the viscosity by 0.6 to 0.7 log units depending on the capillary number of the bubbles (Fig. 5 in Rust and Manga 2002).

The bulk apparent viscosity of the flows is however likely to be several orders of magnitude larger than the viscosity of the molten lava given by petrological estimates in particular because of the formation of a resistant crust (e.g., Griffiths and Fink 1993; Kerr et al. 2006; Castruccio et al. 2013). Several rheological models have been developed to extract the apparent viscosity of the flows from their dimensions (e.g., Nichols 1939; Hulme 1974; Griffiths and Fink 1993; Pinkerton and Wilson 1994). We have applied some of these models (see Online resource 2) to the latest flows (F6, F7, F8, and F11) because these are fully exposed, which allows measuring their dimensions entirely. The average flow width and thickness required in the equations were derived from 7 to 9 profiles drawn across each of these flow units (Fig. 11) and the total flow length represents an average of 10 vent-front measurements. The underlying slope was approximated from the slope of the flow surface over the entire length of the flow because the pre-eruption topography is buried under adjacent flows. This represents a minimum value, as lavas tend to fill depressions and lower slope gradients. The thermal diffusivity used for the calculations is 4.21 × 10⁻⁷ m² s⁻¹, as in Kilburn and Lopes (1991). Note that error propagation for these models (calculated following Chevrel et al. 2013, Appendix A.2, and Lefler 2011) yields up to 40 % error due to the large standard deviation obtained from flow heights (up to 10 %). Physical constants, dimensions, and rheological parameters calculated by this approach are given in Table 7 and presented below.

Fig. 11

Topographic profiles (perpendicular to flow direction) across well-exposed El Metate lava flows (F6, F7, F8, and F11) from which width, thickness, and channel width in the case of F6 were estimated

Table 7 Geometric flow parameters, viscosities, yield strengths, and emplacement durations of El Metate lava flows calculated from the morphological approach

The simplest rheological model used is the one originally defined by Nichols (1939) who adapted the equation of Jeffreys (1925) to describe channelized lava flowing down an inclined slope. The derived equation assumes Newtonian rheology and requires knowledge of the flow mean velocity (Eq. 6 in Online resource 2). This was derived from the flow dimensions, following the G _z approach developed by Pinkerton and Sparks (1976), Hulme and Fielder (1977), and Pinkerton and Wilson (1994) and assuming cooling-dependent flow advance (results from these calculations are presented below). Applying the Jeffreys’ equation yielded apparent viscosities of 6 × 10⁹ Pa s for F7, ∼10¹⁰ Pa s for F6 and F8, and slightly higher for F11 (10¹¹ Pa s).

Hulme (1974) developed equations describing the unconfined flow of ideal Bingham liquids on an inclined plane. These predict the formation of lateral levées and a simple relation between yield strength and flow dimensions. Based on this theory, Wilson and Head (1983) derived an equation that relates the apparent viscosity of the flow within the channel to the effusion rate and yield strength (Eq. 9 in Online resource 2). For unit F6, the occurrence of a well-developed channel allowed us to apply this equation, giving an apparent viscosity of 6 × 10⁸ Pa s.

Considering a Bingham model, bulk yield strength of the flows can be estimated from their thickness (Eq. 7 in Online resource 2). This yielded values of ∼10⁵ Pa for all lava units, except for F8, which had lower yield strength (8.4 × 10⁴ Pa). For unit F6, Hulme’s equation gave a similar value (1.9 × 10⁵ Pa).

The effusion rates estimated from the G _z approach mentioned above are 20 and 60 m³/s for units F7 and F6, respectively, corresponding to a lava front advance of ∼30 m/day. For F8 and F11, the effusion rate was estimated at 9–12 m³/s, equivalent to a lava front advance of a few meters per day. These values were then used to derive emplacement times, considering the individual lava flow volumes that range between 0.5 and 2.2 km³ (Table 7). This yields emplacement times of between ∼500 days for F6 and F8, ∼330 days for F7, and ∼2300 days for F11 (t _Gz on Table 7). Note that this model assumes a constant effusion rate and therefore underestimates the emplacement duration of the lava flow.

Finally, we applied the model of Kilburn and Lopes (1991) who consider the flow as a Newtonian fluid whose expansion is controlled by the resistance of a cooling crust, hence approximating a Bingham rheology. The equation they derive allows calculation of the emplacement duration of a flow-field based only on its final dimensions and the underlying slope (Eq. 11 in Online resource 2). Applying this equation yields slightly higher values for F6 (600 days) and F11 (∼2600 days) but a lower value for F7 (265 days) and a much higher value for F8 (∼1300 days). An average of the emplacement times gives 1.5, 0.8, 2.5, and 6.7 years for F6, F7, F8, and F11, respectively (Table 7).

Discussion

Magma source and evolution

Our data allows us to distinguish at least two different magma batches at El Metate that produced the early and late lava groups, respectively. SiO₂ contents in both groups vary over a similarly wide range; nevertheless, they have distinct morphologies, mineralogy, major element trends as a function of SiO₂, trace element concentrations, and isotope compositions. Their mineral assemblages also formed under different conditions, revealing separate evolutionary histories. We consider below some of the processes that may have caused the chemical variations.

Fractional crystallization modeling

Fractional crystallization may have caused the compositional trends observed within each group (Fig. 7). To test this, we first modeled major element trends using a mass balance approach, employing a least-square fit model (Herrmann and Berry 2002) and a specially designed spreadsheet. Parameters used as input were the least evolved whole rock compositions of each trend observed, average compositions for phenocryst phases measured in these same samples, Fe-Ti oxide compositions measured in late lavas, and a representative composition for apatite from the literature (details in Online resource 5). Results are plotted along with the data in Fig. 7. The observed variations in the two groups of lava can be modeled for most of the elements by fractionation of a combination of the mineral phases present in the samples, namely olivine, plagioclase, clinopyroxene, apatite, and Fe-Ti oxides for early lavas and plagioclase, hornblende, clinopyroxene, apatite, and Fe-Ti for late lavas (Fig. 7). The models however fail to reproduce the variations of Na₂O (Fig. 7). Some discrepancies in TiO₂ for early lavas may be due to higher Ti/Fe ratios in these lavas compared to late lavas.

To further test these results, modeling of selected trace element abundances was undertaken using the mineral assemblages determined by major element modeling, the Rayleigh fractionation equation, and a range of partition coefficients taken from recent literature (Tiepolo et al. 2007; Sisson 1994; Laubier et al. 2014; Prowatke and Klemme 2006; Tepley et al. 2010). For early lavas, the calculations reproduce fairly well the slight increase of REE elements with increasing SiO₂ that is observed. However, the sharp decrease in Ni with increasing SiO₂ could not be reproduced because of the low fractionating amounts of olivine (∼3 wt%). This trend would require 15 wt% olivine fractionation, which is inconsistent with major element trends, and may suggest mixing with a source having lower Ni contents. For late lavas, the calculations fail to reproduce the sharp decrease of most REE (e.g., La, Gd) with respect to SiO₂ because these elements are incompatible in all the mineral phases fractionating, except for apatite, which fractionates in too small proportions (<0.4 wt%) to cause a significant effect. Furthermore, the sharp decrease in Sr with increasing SiO₂ would require fractionation of 70 wt% of plagioclase instead of the 6 wt% modeled using major element trends. In conclusion, although fractional crystallization can explain most of the observed variations in the compositions of early lavas, it fails to reproduce trace element variations in late lavas.

The evolution of late lavas with time may instead reflect progressive mixing with the most evolved member of the early lavas (F4). This could explain the convergence of both groups in isotopic, most major element, and some trace element (Sr, Ni) graphs (Figs. 7 and 9), but cannot account for the progressive depletion of late lavas in REE with increasing SiO₂ given that the early lavas have high REE contents. Alternatively, late lavas may have mixed with another, more radiogenic source depleted in trace elements that did not erupt from El Metate. The chemical data of Ownby et al. (2011) for volcanic rocks in the nearby Tancítaro area does not, however, include any composition with such low HREE values.

Source heterogeneities and crustal assimilation

The different isotopic compositions of early lavas (F1-F2-F4b) and late lavas (F5-F11-F12) indicate that they originated from a source affected by continental crust assimilation or from a heterogeneous mantle. More isotopic systems should be used to distinguish between these two hypotheses.

If crustal assimilation would be an important factor, its degree would need to be significant in order to explain the isotopic differences (Fig. 9c, d). Mineralogic evidence of strong contamination is essentially absent since only few Qz-xenocrysts were observed. These must have been picked up shortly before eruption.

Instead, the differences in isotopic and trace element compositions observed between the two lava groups and within the late lavas might have been caused by mantle heterogeneities. These could be derived from variable degrees of enrichment in fluid-mobile elements (LIL: light REE, Rb, Cs, etc.) of the fluids expelled from slab-derived sediments and differences in the extent of depletion in immobile elements (heavy REE, high-field strength elements like Nb, Ta, Ti) by previous melting events as already postulated (e.g., Johnson et al. 2009 and references therein) for other volcanic rocks in the TMVB. The lower concentration of both mobile and immobile elements and the less radiogenic signature of the late lavas in comparison with the early lavas suggest that the former issued from a more depleted mantle source that was affected by lesser degrees of fluid enrichment. The cause of the abnormal Sr-rich composition of the most mafic late lavas (F5–F9) is not clear, but the near-MORB isotopic signature of these lavas excludes contamination by an old radiogenic Sr-rich crust. The high Sr/Y ratios of the late lavas may thus reflect a mantle source contaminated by Sr-rich fluid (or melt). Finally, the poorly fractionated HREE (of both lava groups) suggest that hornblende was a residue in the source, which has been interpreted by Ownby et al. (2011) to suggest an origin by partial melting of hornblende-rich gabbro-norite in the lower crust. However, Ortega-Gutiérrez et al. (2008) suggest that the lower crust in this area might consist of high-grade hydrous metamorphic rocks. Both cases seem discordant with the MORB-like isotopic composition of the late lavas, but more detailed geochemical studies are needed to better define the nature of the source.

Eruptive model

The more straightforward and simple model would consist of a first magma batch ascending continuously through the crust and erupting the early lava group followed by a second magma batch that then forms the late lavas. However, we prefer a model in which the magma that formed the late lavas stagnated in the crust before ascent and eruption of the early lava.

The magma batch that formed the late lavas (hornblende-bearing) required some residence time in the upper crust (∼7–10 km depth) to allow for phenocryst growth. This would imply cooling and a concomitant viscosity increase, hindering further ascent and eruption. This could be overcome if a second, less viscous olivine-bearing batch passes by, suffering only minor cooling (limited stagnation upon ascent, see Fig. 10), and opens the path, erupting first. Accordingly, we propose a model that includes three main stages (Fig. 12):

Fig. 12

Schematic model depicting magma ascent from the mantle, temporary storage in the upper crust, and eruptive sequence of early (blue) and late (red) lavas

• Stage A. Some time prior to the eruption, a first voluminous batch with a heterogeneous, near-MORB isotopic signature, MgO-poor but Sr and H₂O-rich low-viscosity magma (10²–10³ Pa s) was emplaced in the upper crust at depths of ∼7–10 km where it underwent mainly plagioclase and hornblende crystallization and fractionation.

• Stage B. This was followed by the ascent of a second smaller magma batch that was isotopically more radiogenic, less differentiated (MgO richer) and hotter (>1100 °C). This batch likely bypassed the first batch before erupting at the surface and forming the early lava group (F1–F4b). The lack of chemical and textural evidence for mixing between the two magmas suggests little interaction between them. Although the exact sequence of flow emplacement is not known for early lavas, it seems more likely that the most mafic, olivine-bearing, hottest (∼1076 °C), and less viscous lava (F1) was the first to erupt. The subsequent more evolved, olivine-free magma crystallized pyroxene at shallow levels (∼10 km) and produced the more silicic early lavas (F2–F4) at lower temperatures (∼990 °C). A small lava volume of intermediate composition, likely trapped in the complex plumbing network or resulting from another magma input, erupted subsequently out of F4 forming the small lava lobe F4b. All these early lavas have relatively low emplacement viscosities (10⁶–10⁷ Pa s) and form thin (<40 m) rubbly-‘a’ā to blocky flows.

• Stage C. Once the initial source exhausted and the vent opened, the first magma batch (from stage A, as described above) that stalled at depth followed on the heels in the same dike system to reach the surface producing the hornblende-bearing late lava group (F5–F13). This larger batch shows progressive enrichment in SiO₂, decreasing Sr contents, and increasingly radiogenic isotope compositions with time, possibly due to mixing with an unknown third component. This magma erupted entirely effusively at ∼922–972 °C and lavas rapidly developed high viscosities (from 10⁷ to 10¹¹ Pa s) during emplacement to form thick (60–150 m) blocky lava flows. The flows that were emitted to the south (F6–F7) nevertheless reached long distances from the vent (<15 km) probably due to the combination of low heat loss during emplacement (efficient thermal insulation by the formation of a carapace) and a great volume emitted on a pronounced gradient.

Eruptive dynamics

If erupted explosively, the large magma volume involved in the El Metate eruption would have produced a major Plinian eruption (volcanic explosivity index >5) with a tremendous impact on surrounding populations (Chevrel et al. 2016). It is worth noting that all eruptions of large volumes of andesitic magma in the Holocene have been explosive (Table 3 in Chevrel et al. 2016), and hence, in this respect, El Metate’s eruption is intriguing. For hazard assessment purposes, it is of fundamental importance to understand why magmas with similar compositions and physical properties can display contrasting eruption styles: Why did El Metate erupt effusively and not explosively? Magma viscosity is not the only parameter determining how volatiles degas from the magma (explosive vs. effusive style). Degassing processes also depend on whether closed vs. open system conditions become established in upper crustal levels prior to eruption (e.g., Cashman 2004). The effusive style of El Metate implies efficient open system degassing, through an open conduit and permeable walls, of a magma that ascended at rates that were sufficient to sustain the effusive activity throughout the eruption allowing continuous passive gas escape. In this context, the mass eruption rate was estimated for El Metate (assuming continuous and sustained activity) by multiplying the effusion rate (obtained by the morphological approach) for late lavas (5–50 m³/s) by the density of typical andesitic lava with 0–30 % of vesicles (1820–2600 kg/m³). This yielded a mass eruption rate in the range of 10⁴–10⁵ kg/s, a value that falls at the limit between the explosive and effusive eruption styles for a wide range of observed historic cases (Cashman 2004). This value is similar to the mass eruption rate reported for the initial violent Strombolian activity of Paricutin volcano (Pioli et al. 2008) but only slightly higher than the long-term effusion rates observed at lava lakes and domes (Pyle 2000).

The texture of hornblende crystals is strongly dependent on decompression rates because this affects the time available for the crystals to react to decreasing melt water contents (Rutherford and Hill 1993; Rutherford and Gardner 2000; Rutherford and Devine 2003). The rates calculated above correspond to the limit at which the minerals start to break down in the experiments of Rutherford and Hill (1993). This may explain why, in El Metate lavas, some crystals are still intact while others have partly to fully suffered breakdown and recrystallization. Alternatively, there may have been heterogeneities in the decompression rate experienced by the ascending magma due, for example, to velocity gradients across the conduit. Opacite rims around the crystals likely grew by oxidation and dehydrogenation during lava emplacement (Garcia and Jacobson 1979).

It is worth noting that El Metate’s early lava type (MgO-rich andesite with abundant olivine) is frequent in the MGVF (e.g., Jorullo, Paricutin, etc., Hasenaka and Carmichael 1987), while the El Metate late lava type (hornblende-bearing MgO-poor andesite) occurs more rarely and commonly forms smaller edifices. Specific conditions must have existed at El Metate to allow for such a large volume of an initially hydrous, low-T, crystal-rich magma to reach the surface and erupt in a solely effusive manner. Intense passive degassing of the magma must have occurred prior to or during ascent to prevent explosive eruption. We envision a scenario in which the second magma batch that produced the early lavas (which were the hotter, more fluid, and also more frequent in the MGVF) opened the conduit allowing gas escape from the hornblende-bearing first magma batch (at that time still residing at depth) and therefore facilitating its final ascent by reducing the pressure gradient. Brenna et al. (2010) proposed a similar scenario, according to which a low volume magma acted as a trigger or path-opener for the eruption of a shallower derived, larger volume magma and that the two magmas used the same single-dike plumbing system.

The above scenario leads to the question of why the hornblende-bearing first magma batch did not reach the surface first but instead became temporarily stalled at depth. On the basis of petrological experiments, Carmichael (2002) demonstrated that a decompressing andesitic hydrous (but relatively cold) magma, such as El Metate’s first batch, that reaches hornblende stability at low pressures will crystallize readily, yielding to viscosity increase, and thus more likely become stalled during ascent, than a hotter and less hydrous magma (such as El Metate’s second magma batch). The eruption of El Metate’s late lavas thus exemplifies the specific conditions under which hornblende-bearing crystal-rich magmas can erupt. The above implies that, given the evidence for a hydrous mantle below the MGVF (Johnson et al. 2009), large volumes of this magma type may be stalled in the crust underneath the MGVF.

Lava emplacement and duration of the eruption

Active andesitic flows of the size of those emitted by El Metate have never been witnessed, which makes it difficult to make interpretations about their emplacement style. The viscosities derived from petrological parameters are first-order approximations because they do not take into account heterogeneities in crystal and bubble distribution, crystal-crystal interactions, strain-rate effects (Caricchi et al. 2007; Costa et al. 2009; Lavallée et al. 2007), or viscous heating (Cordonnier et al. 2012). Morphologically derived parameters are also subject to much uncertainty given that most of them assume simple, nonrealistic rheology (Newtonian). More complex formulations (e.g., Castruccio et al. 2010, 2014) require consideration of a large set of additional petrographic and morphometric parameters (fraction of the crystals formed after flow emplacement, parameterization of the yield strength and shear rate dependency as a function of crystal content, flow thickness during the eruption, etc.) that cannot be inferred for flows whose emplacement was not directly observed, and would require more complex modeling which is beyond the scope of this paper.

Nevertheless, the viscosities and effusion rates calculated here are quite comparable to those reported for smaller active blocky lava flows (e.g., Lonquimay, Chile; Naranjo et al. 1992; Santiaguito, Guatemala; Harris et al. 2004) but slightly higher than expected for andesitic flows (e.g., Kilburn 2000; Harris and Rowland 2015). The inferred yield strength values (∼10⁵ Pa) are also similar to values obtained for intermediate to silicic lava flows and domes (Harris and Rowland 2015).

The effusion rates estimated here might be too high because their calculation considers a constant magma output rate, although the latter was observed to decrease over time in the few active flows that have been observed directly. For example, the effusion rates during the eruption of the andesitic lava flow from Lonquimay, Chile, started at 80 m³/s and decreased rapidly to <10 m³/s within the first 50 days, for a total duration of 330 days (Naranjo et al. 1992). Furthermore, well-isolated flows could continue advancing despite a decrease or cessation of the lava supply. An example for this is the rhyolitic lava flow from Cordón Caulle, Chile, which continued to advance as far as 3.6 km from the vent, although the lava supply had ceased 6–8 months earlier (Tuffen et al. 2013). Still today, it is not possible to anticipate with certainty how long the Cordón Caulle flow will remain active. In addition, the time that elapsed between the extrusion of each of El Metate’s lava flows is unknown. However, given the young age (∼AD 1250) of this volcano and in the absence of any reference to its activity in the early colonial chronicles, it is safe to assume that El Metate’s lavas must have been essentially cold by the time of the arrival (AD 1520s) of the Spaniards in Michoacán. Accordingly, the maximum emplacement duration of the entire volcano must have been less than 275 years (Chevrel et al. 2016). On the other hand, assuming a continuous and successive (one after the other) eruption of all lava flows (representing ∼10.8 km³, non-DRE; Chevrel et al. 2016) at a rate of ∼10 m³/s, the formation of El Metate should have taken at least ∼35 years.

Conclusions

Stratigraphic relations and lava flow morphological parameters together with viscosity estimates all indicate that the entire eruption of El Metate was purely effusive and continuous. At the end, 103 km² of terrain was covered by lava flows forming a monogenetic shield with a total volume of 9.2 km³. Hence, El Metate was formed by the largest andesitic effusive eruption worldwide and the most voluminous eruption in Mexico during Holocene times, and it certainly had a significant impact on the environment and human populations (Chevrel et al. 2016).

Our detailed study of El Metate volcano, the youngest andesite shield of the MGVF, shows that its lava flows originated from two separate magma batches that had different mineralogical and chemical compositions, as well as distinct Sr-Nd-Pb isotope signatures. The source for both batches was a subduction-modified heterogeneous mantle. Mineral thermobarometry reveals that after initial ascent through the crust, the first batch became temporarily stalled at a depth of ∼7–10 km, allowing for crystallization and fractionation. Then, the second batch ascended and evolved independently, bypassed the first batch without major mingling or mixing, and erupted effusively to form the early lava flows (F1–F4). The first batch followed immediately after, along the same conduit and produced the late lava flows (F5–F13).

The continuous ascent rates must have been slow enough to allow extensive open system degassing (outgassing) of magma as it was rising to the surface, thereby preventing explosive eruption. Instead, effusive outpouring of a large volume of andesitic magma formed more than a dozen lava flows. Fed by a relatively high and sustained mass eruption rate (10⁴–10⁵ kg/s), some of the most voluminous (∼2 km³) lava flows reached distances of up to ∼15 km from the vent, aided by the pre-existing slope and low heat loss during emplacement. The morphologies and compositions of early Mg-rich lavas indicate lower viscosities (10⁶–10⁷ Pa s), while the late hornblende-rich lavas (10⁷ to 10¹¹ Pa s) formed thicker viscous flows. Estimates of flow emplacement duration yielded ∼2 years for the longest and ∼7 years for the thickest flow, respectively. Successive emplacement of all flows probably took ∼35 years.

Finally, El Metate is one of ∼400 medium-sized volcanoes in the MGVF. Whether many of them followed a similar eruption pattern is unclear, however no other Mexican shield displays such a well-exposed morphology. Other previously studied Mexican shields such as Cerro Paracho or El Estribo are clearly composite and polygenetic (Pola et al. 2014; Siebe et al. 2014; Chevrel et al. 2016). Although El Metate represents an exceptional volume of lava, many other examples of similarly thick blocky andesite lava flows exist in the MGVF. Many of them are isolated and have a much smaller volume but may represent other examples for the effusive eruption of a similar magma type. Further studies of such voluminous blocky flows are needed to both better understand their mechanism of emplacement, as well as their role in the tectonic evolution and eruption dynamics of the MGVF.