Introduction

Pyroclastic density currents (PDCs) are mixtures of hot volcanic particles and gas that flow downslope under the influence of gravity. They are highly energetic and account for about a third of the total fatalities (278,800 deaths) due to volcanic activity worldwide since 1600 AD (Auker et al. 2013). Pyroclastic fallouts consist of volcanic particles that have traveled through the atmosphere before falling to the ground. They are not as threatening as PDCs but their widespread effects can produce severe economic loss (Oxford-Economics 2010) and even climatic changes (Pitari and Mancini 2002). In the last two decades, analogue modeling and numerical simulations allowed major improvements concerning the knowledge of these phenomena. Nevertheless, the study of modern and ancient pyroclastic deposits remains fundamental to calibrate numerical models, elaborate future eruptive scenarios, and understand emplacement processes (Cas and Wright 1987).

Chachimbiro compound volcano (0°28' N, 78°17' W, 4034 m asl) is located 80 km north of Quito, in the northern part of the Ecuadorian Volcanic Front (EVF), the westernmost volcanic range related to the subduction of the Nazca Plate below South America (Fig. 1). In this region, Holocene activity is characterized by powerful but low-frequency explosive eruptions of acid andesite to dacite, with Volcanic Explosivity Index (VEI) ranging from 3 to 6, and by lava dome extrusions (Von Hillebrandt 1989; Andrade et al. 2014; Hall and Mothes 2008; Hidalgo et al. 2008; Robin et al. 2008). The long-lasting evolution of the Chachimbiro volcano comprises three successive phases named Huanguillaro, Tumbatú, and Hugá (Bernard et al. 2011). Only limited information is available on the Holocene activity of this volcano (Beate 2001), but recent investigation led the identification of at least four eruptions associated to the Hugá phase in the last 6,000 years (Bernard et al. 2011). In this paper, we date and study the deposits of the most important eruption from this period. This eruption left a complex well-preserved PDC deposit that crops out abundantly on the eastern side of the volcanic complex, in the Urcuquí district (Fig. 2). It also produced a pyroclastic fall deposit presently buried within the andisol of the Chachimbiro highlands.

Fig. 1
figure 1

a Geodynamic context of Ecuador, and b location of Chachimbiro volcano (modified from Monzier et al. 1997). CAR Carnegie Ridge; NAB North Andean Block; NAZ Nazca Plate; SA South American Plate

Full size image
Fig. 2
figure 2

Outcrops of the 3640–3510 years BC eruption deposits. a Quarry in the lowlands; b outcrop in Chachimbiro highlands showing the fallout layer and a younger fall deposit from Hugá edifice; c superposition of PDC and fall deposits

Full size image

The aim of this paper is to describe the 3640–3510 BC deposits in order to identify the source and the eruptive dynamics of this major eruption. Results of stratigraphical analysis as well as geochronological, geochemical, and physical volcanology studies are presented with a special focus on the most violent episode of the eruption: the emission of a high-energy PDC.

Methods

The PDC and fallout deposits have been sampled and studied in 95 locations in the field. Radiocarbon ages were obtained on six samples. Four age determinations were carried out by the LMC14 (Laboratoire de Mesure du Carbone 14), Bondy (France), using accelerator mass spectrometry (AMS). Two conventional determinations were obtained at the Center of Isotope Research, Groningen University (Netherlands), and using pre-treatment and analytical procedures described in Mook and Streurman (1983). Calibration of the 14C ages was realized using OxCal 3.1 (Bronk Ramsey 2005) and atmospheric data from Reimer et al. (2004).

Major and trace element whole-rock analyses were carried out at the Université de Bretagne Occidentale. Chemical analyses were obtained on agate-ground powders by inductively coupled plasma–atomic emission spectroscopy (ICP-AES), except for Rb measured by flame atomic emission. Relative standard deviations are 1 % for SiO2 and 2 % for the other major elements except for low concentrations (<0.50 %) for which absolute standard deviation is 0.01 %. Analytical procedure is described in detail in Cotten et al. (1995). Major element compositions for mineral and glasses were obtained at the Laboratoire Magmas et Volcans (Université Blaise Pascal, Clermont-Ferrand, France) using a CAMECA Camebax SX-100 electron microprobe, operating at 15 kV accelerating voltage, with a beam current of 15 nA for mineral analyses, and 8 nA and a 5-μm defocused beam for glass analyses. The estimated error for single analysis is of about 1 % when the element concentration is higher than 10 % and of 2–5 % when it is lower than 10 %.

Some physical characteristics of the deposit were observed and/or measured directly in the field (thickness, structure, and texture of the deposits). For the pyroclastic fallout deposit, the maximum clast size (both lithics and pumice) was obtained by measuring the three axes of the five largest clasts on an exposure surface of ∼0.5 m2 (Biass and Bonadonna 2011). The PDC deposits were mechanically sieved at 1 − Ф intervals from −5 to 4Ф (i.e., from 32 mm to 64 μm, with Φ = −log2 d, d being the particle diameter in millimeters) and weighed at 10−2 g for grain-size analysis. This study was completed with density measurements performed on the −4 to −5 Ф fraction (16–32 mm) using the method described in Houghton and Wilson (1989) and based on Archimedes’ principle.

Age of the eruption

Three of the six radiocarbon ages were obtained on charcoal included in the PDC deposit. The other three come from two peat samples and a palaeosoil in the Chachimbiro highlands. The ages are consistent with the stratigraphy observed in the field (Table 1). The palaeosoil (5,420 ± 40 years BP) and the lower peat (5,190 ± 40 years BP) give a maximum age for the eruption that is slightly younger than the age obtained on a palaeosoil (5,690 ± 50 years BP) in the Chachimbiro highlands by Beate (2001). The upper peat gives a minimum age for the pyroclastic fall deposit (4,615 ± 40 years BP). The consistency of the ages obtained on charcoal samples from the PDC deposit (Table 1) rules out any contamination process and allows calculation of a precise age for the event. Their combination leads to a single calibrated age of 3640–3510 years BC within a probability of 92.9 %.

Table 1 Radiocarbon age determinations and calendar age conversions. Calibration uses the OxCal 3.1 program (Bronk Ramsey 2005) and atmospheric data of the Intcal04 calibration curve (Reimer et al. 2004). Converted dates have a 1-year resolution but are rounded off to the nearest 5 years AD
Full size table

In a sector called Cruz Tola, we found archaeological remains under the PDC deposit in the form of an irrigation channel. This indicates that there were human settlements in the Urcuquí district in the Early Formative Period (4400–1450 cal BC, Zeidler 2008). In this area there are also “Tolas”, small mounds constructed by humans to celebrate religious rites. Archaeological investigations show that the “Tolas” were built by the Caranquis during the Late Period, between 1250 and 1550 AD (Moreira Pino 2012), and therefore not directly related to the irrigation channel. Abundant obsidian and ceramic fragments in this area indicate that the Urcuquí district was an important place during pre-Columbian times.

Petrology of the juvenile samples

Whole-rock major and trace elements

SiO2 contents of juvenile fragments from the PDC (five samples) and fallout (three samples) deposits range from 67.3 to 68.9 wt% (Table 2). Nevertheless, the PDC samples show a very narrow silica range (68.4 ± 0.1 wt%). Since these blocks were collected in different localities, we consider their compositions representative of a homogeneous source. According to the classification of Peccerillo and Taylor (1976), the analyzed rocks are medium K-rich calc-alkaline rhyodacites (Fig. 3a). Compared to the whole Chachimbiro magmatic suite, which range from basaltic andesites to dacites, the 3640–3510 BC juvenile fragments are enriched in SiO2 and depleted in MgO (Fig. 3b), as well as in FeO and CaO (not shown). Highly fractionated spectra for HREE are emphasized in the La/Yb vs. Yb diagram (Fig. 3c) where the high La/Yb ratios result from the Yb depletion in these rocks. In spider diagrams, they plot in the medium to upper part of the compositional range defined by the Chachimbiro rocks for the LILE (Cs, Rb, Ba, Th, Sr) and LREE (La and Ce). On the contrary, the analysis show a progressive depletion in MREE (Sm and Eu), HREE (Dy and Yb), and Y (Fig. 3d). Spider diagrams also show Nb negative anomaly typical of arc magmas. Globally, even though the 3640–3510 BC products appear more differentiated than the other samples from Chachimbiro compound volcano, they follow the same geochemical trend. Thus, we consider that they most likely come from the Chachimbiro magmatic system.

Table 2 Whole-rock and trace element geochemical analysis for the 3640–3510 BC eruptive products. Major elements expressed as oxides in wt% and trace element concentrations in ppm
Full size table
Fig. 3
figure 3

a Classification diagram for volcanic arc rocks (Peccerillo and Taylor 1976). b MgO vs. SiO2 (wt%) diagram showing strong differentiation of 3640–3510 BC samples with respect to other rocks of Chachimbiro compound volcano. c La/Yb vs. Yb diagram; high La/Yb ratios are mainly due to Yb depletion. d Spider diagrams normalized to N-MORB composition by Sun and McDonough (1989)

Full size image

Petrography and mineral chemistry

The juvenile samples from the PDC deposit have a porphyritic-microlitic texture with about 30 vol% of phenocrysts. Textural banding due to variable degrees of vesicularity is also observed. The phenocryst mineral assemblage is composed of plagioclase (15 vol%) + amphibole (9 vol%) + Fe-Ti oxides (3 vol%) + biotite (2 vol%) + quartz (1 vol%) + traces of apatite (Fig. 4). The groundmass consists of 80 vol% of plagioclase, amphibole, Fe-Ti oxides and quartz microlites, and 20 vol% rhyolitic glass. Microprobe analyses are presented in the Online Resource 1.

Fig. 4
figure 4

Phenocryst assemblage in the 3640–3510 BC lava observed (microscope field, 5 mm). Plg plagioclase, Amp amphibole, Ox oxide; Bio biotite

Full size image

Plagioclase is the most abundant phase (Fig. 5a), present as zoned phenocrysts, xenocrysts, inclusions in amphibole, and microlites. Euhedral phenocrysts (between 1 and 2.5 mm in size) show mostly reverse zoning with Na-rich cores (An 33.6 ± 3.0 mol%) evolving to more Ca-rich rims (An 41.4 ± 1.7 mol%). Xenocrysts are anhedral andesine with An contents of 43.5 ± 2.6 mol%. Inclusions in amphibole display similar andesine compositions (An 47.4 ± 3.9 mol%). Microlites consist of acicular plagioclase showing a homogeneous composition at the limit oligoclase-andesine (An 29.5 ± 0.7 mol%) indicating equilibrium conditions with the surrounding analyzed volcanic glass. Amphibole is present as green and brown euhedral phenocrysts (<1 mm) and microlites. According to the classification of Leake et al. (1997) they are magnesio-hornblende and tschermakite (Fig. 5b). No reaction rims are optically observed and no notable difference in composition occurs between the narrow margins of crystals (<20 μm) and the inner adjacent zones. This indicates that, prior to the eruption, the crystals were mostly in equilibrium with the magma and that the ascent to the surface was fast enough to avoid any destabilization (Rutherford and Hill 1993). Fe-Ti oxides are present as titano-magnetite and ilmenite, both as euhedral microphenocrysts (about 0.3 mm in size) and microlites (<100 μm), as well as inclusions in amphibole. Both oxides are present as individual crystals without forming solid exsolution. Biotite is present as euhedral pleochroic phenocrysts of about 1–3 mm long. Two slightly different compositions are observed (Mg#, 0.54–0.56; K2O, 9.3 wt%; TiO2, 3.6 wt%; and Mg#, 0.64–0.65; K2O, 9.3 wt%; TiO2, 2.8 wt%). Quartz is present as euhedral microphenocrysts and microlites. Apatite appears as an accessory phase (<1 vol%), present as euhedral microphenocrysts (<0.3 mm). The volcanic glass from the groundmass shows a very homogeneous rhyolitic composition (78.8 ± 0.7 SiO2 wt%).

Fig. 5
figure 5

Classification of minerals a Plagioclase classification of 3640–3510 BC eruption PDC juvenile products. Plagioclase phenocrysts vary from andesine to oligoclase. Two K-feldspars were also distinguished. int intermediate zones in phenocrysts; inc inclusions. b Amphibole compositions (cations in structural formula) plotted on the classification diagram of Leake et al. (1997)

Full size image

Pre-eruptive conditions

Geo-barometry

Our mineralogical study has shown the crystallization of hornblende mostly at equilibrium with the surrounding melt. The Al-in-hornblende barometer is based on the increase of Al in hornblende as a function of pressure (Johnson and Rutherford 1989). We used the new empirical thermobarometric formulations from Ridolfi et al. (2010), which are based on a rigorous analysis of the physical-chemical, compositional and textural relationships of amphibole stability. These authors considered experimental results, H2O-CO2 solubility models, as well as natural amphibole-bearing calc-alkaline products. We applied their formulations to the amphibole and found two groups of amphibole cores associated to different crystallization pressures. The first group displays an equilibrium pressure of 408 ± 30 Mpa (1 sigma SD) and the second of 228 ± 33 Mpa (Table 3) the presence of two magmatic reservoirs. The corresponding depths are respectively 14.4 ± 1 km and 8.0 ± 1.1 km.

Table 3 Pre-eruptive pressure, temperature and water content of the rhyodacite feeding the 3640–3510 BC eruption
Full size table

Geo-thermometry and oxygen fugacity

Ridolfi et al. (2010) formulations allow also the estimation of equilibrium temperature and oxygen fugacity (Table 3). The higher-pressure amphibole was formed at temperatures of 936 ± 22 °C with a log fO2 of −10.6 ± 0.4 (ΔNNO 0.58 ± 0.29). The temperatures calculated for the amphibole that crystallized at shallower depth are lower (858 ± 22 °C) with a log fO2 of −11.5 ± 0.3 (ΔNNO 1.10 ± 0.3). High water contents are estimated with these formulations, with H2O of 7.8 ± 0.4 wt% for the deeper amphibole and 7.1 ± 0.5 wt% for shallower amphibole. These values are supported by the presence of euhedral biotite in the 3640–3510 BC samples.

Given the pcresence of titano-magnetite and ilmenite, we applied the thermometric formulations based in on the equilibrium of these two oxides. The 19 oxide pairs that fulfilled the equilibrium requirement of Bacon and Hirschmann (1988) were used to apply the thermometric formulations of Andersen and Lindsley (1988) and Ghiorso and Evans (2008). Considering that the oxides’ equilibrium should represent the nearest surface conditions we used the QUILF program (Andersen et al. 1998), first varying T and oxygen fugacity, in order to obtain an oxygen fugacity value to fix in a second run. The obtained log fO2 was −11.7, a very close value to that obtained for the shallower amphibole. According to the QUILF calibration and fixing the oxygen fugacity and the pressure to 230 MPa for shallow-depth conditions, the average equilibrium temperature for the 19 pairs was 768 ± 44 °C. The same oxide pairs were used in the FeTiOxideGeotherm (version 1.1) application of Ghiorso and Evans (2008). This application releases two temperatures, one based in the Fe-Ti exchange (679 ± 12 °C) and one in the Fe-Mg exchange (775 ± 52 °C) with a ΔNNO of 1.65 ± 0.02.

Finally, we used Eq. 24a from Putirka (2008) for a plagioclase-liquid thermometer. This equation allows estimating equilibrium temperature between liquid and plagioclase within an error of ± 36 °C. From 49 liquid–plagioclase pairs, only 14 satisfied the equilibrium condition where stated by the author (KDAb-An for temperatures below 1,050 °C is of 0.1 ± 0.05). We tested Eq. 24a with a fixed water content of 7 wt% and a pressure of 230 MPa, both estimated with the shallow-depth amphibole. Calculated temperature is 722 °C (Table 3), consistent with those calculated with oxides pairs.

Physical characteristics of the deposits

PDC deposit

Mapping

The PDC deposit is easily recognized all over the Urcuquí district along the roads and in small quarries (Fig. 2a). It covers a fan-shaped area of about 62 km2 including the main towns of Urcuquí district (Urcuquí, Tumbabiro, and San Blas) (Fig. 6). The thickness of the PDC deposit ranges from a maximum of 15 m at the foot of the peak La Viuda to few decimeters near Urcuquí. High thickness values (>2 m) are found in the five principal valleys of the area (Cachiyacu, Chuspihuaycu, la Banda, Añaburo, and Pigunchuela). Overall, the deposit thickness increases toward the center of the deposit area and the northwest.

Fig. 6
figure 6

Areal distribution and thickness of the 3640–3510 BC PDC deposit. The crosses correspond to thickness measurements carried out in 69 outcrops

Full size image

Stratigraphy

The PDC deposit overlies directly either an indurated volcanic ash deposit (Fig. 2a) known in Ecuador as Cangahua (Hall and Mothes 1997) or a brown ash layer. In few outcrops, the brown layer is transformed into a soil. The basal contact is generally flat with few erosive features and burned vegetation. The top of the PDC deposit is also relatively flat and forms terraces in the valleys. The deposit shows a complex stratification with lateral thickness and facies variations. Thirty-seven sections of this deposit are presented in Fig. 7.

Fig. 7
figure 7

Stratigraphic columns of the 3640–3510 BC PDC deposit with main units. Stratigraphic sections located in the valleys are in bold italic and normal font when located in the interfluves. The color of the stratigraphic columns corresponds to the sector where it was described (see the inset map). The letters in brackets correspond to the grain-size samples. Cal 14C ages are in red

Full size image

In order to present a comprehensive description, we simplified the deposit stratigraphy using the principal field characteristics. The deposit is principally composed of two layers (Fig. 8). The lower layer (L1) is a massive juvenile-rich pyroclastic deposit with some large rhyodacite blocks (>50 cm) and charcoal fragments. L1 is mostly concentrated in the valleys but can also be found in the interfluve areas. In some localities, this layer is roughly divided in two parts mostly due to difference of coloration (columns 009, 021, 083, and 108). The top of L1 displays gas escape structures. This layer contains banded and prismatic jointed blocks, and accidental clasts that are mostly oxidized andesite and lutite from the cretaceous basement (Natividad and Rio Cala units; Vallejo et al. 2006). The amount of accidental material is higher near the Cachiyacu river. In some outcrops, a fine-grained basal layer (L1a) is found below or instead of L1. This layer is slightly laminated and is typically found in the interfluve area.

Fig. 8
figure 8

Stratigraphy of the 3640–3510 BC PDC deposit

Full size image

The upper layer (L2) is the most widespread and is found both in valleys and in interfluve areas. This layer consists of stratified coarse ash and lapilli laminae. L2 has a higher content in accidental material than L1 with scarce large blocks (>20 cm) although it remains mostly composed of juvenile material (>90 vol%).

An ash layer (LA) with lenses of lapilli is found on top of the upper layer (Fig. 8). This layer, that can be up to 3 m thick, appears mostly in the valleys and is interpreted as reworking of the fine material from the PDC deposit.

Volume

As an isopach map would not correctly represent the spatial distribution of a topographically controlled deposit, the volume of the deposit has been estimated using the Crandell (1989) method. The area has been divided into 10 segments, and an average thickness is attributed to each one (Fig. 6). It is important to note that in each segment the standard deviation on the thickness values can reach 20 to 62 % (Table 4) suggesting a possible large uncertainty on the volume calculation. This method gives a total volume of about 55 × 106 m3.

Table 4 Areas, average thicknesses, and volumes of the 3640–3510 eruption PDC deposit. Segment numbers correspond to those in Fig. 6
Full size table

Density and vesicularity of clasts

The juvenile material is a porphyritic-microlitic rhyodacite with variable vesicularity. Dense rock equivalent density has been measured by water pycnometry on five rock powders (<2 μm). The results give a dense rock density of 2.60 ± 0.05 g cm−3. Bulk density measurements using the immersion method were carried out on 1,491 juvenile clasts from the lower and upper PDC layers (Table 5). The results show a wide range of densities varying from 0.67 to 2.59 g cm−3. The maximum density measured confirms the dense rock density obtained by water pycnometry. Therefore, the density measurements were converted to vesicularity. The global vesicularity histogram, showing the volumetric abundance of clasts by range of vesicularity, indicates that the distribution of vesicularity in the PDC deposit is roughly bimodal (Fig. 9a). A principal mode appears for both layers at 5–10 % of vesicularity and a secondary mode at 50–55 % characterizes better the upper layer. L1 has about 80.4 vol% of dense clasts (vesicularity < 25 %) contrary to L2 which presents a higher content in vesicular clasts (28.1 vol%). These results indicate a vertical variation of the clasts vesicularity in the PDC deposit. Nevertheless, the global shape of the density histograms (Fig. 9a) is the same suggesting that the material came from the same explosion and the vertical evolution may be due to transport and sedimentation processes. We present two density histograms made with samples from the same outcrop that support this general trend (Fig. 9b, c).

Table 5 Density of juvenile clasts in the 16–32 mm fraction. Average density, vesicularity, and accidental content are balanced volumetrically
Full size table
Fig. 9
figure 9

a Vesicularity histograms of the main layers (L1 and L2) of the PDC deposit; the histogram for each unit has been created using the average of individual histograms (18 histograms for the lower layer L1 and 14 for the upper layer L2). b Vesicularity histogram of the sample 098-B from the lower layer L1 and c the sample 098-C from the upper layer L2. The abundance of clasts for each vesicularity range has been volumetrically balanced. Note the difference between the lower and upper layer in general and at outcrop scale, showing a vertical evolution

Full size image

Grain size distribution

In general, the deposits consist mostly of very fine ash to coarse lapilli particles but also contain a low fraction of extremely fine ash (<63 μm or greater than 4φ) and blocks (>64 mm or less than −6 φ) (White and Houghton 2006). A detailed grain-size analysis was realized on 70 samples for the −5 to > 4 φ fractions (32 mm to <63 μm) obtained by mechanical dry sieving. The extremely fine ash fraction represents about 2 % of the whole PDC deposit but variations from the base to the top from are notable: 1.3 wt% for the lower layer, 2.4 wt% for the upper layer. The layers L1a and LA present a higher amount of fines with 3.4 and 6.7 wt% respectively. The grain-size histograms (Fig. 10) and the Inman (1952) graphical statistics (Fig. 11, Online Resource 2) show that the PDC deposit displays broad grain-size distribution variability. The layer L1a is typically well sorted with a medium to very coarse ash mode and a normal distribution (Fig. 10a). L1 is relatively homogeneous and poorly sorted with a fine-medium lapilli mode and a positive skewness (Fig. 10b, c). L2 has the largest range of grain-size distribution from well sorted to poorly sorted with a grain-size mode ranging from medium ash to medium lapilli (Fig. 10d, e). The ash layer (LA) presents a medium ash mode at the limit between well sorted and poorly sorted (Fig. 10f). There is no evidence of longitudinal variation in L1. On the contrary, L2 changes from a fine-medium lapilli mode in the central area (Fig. 10d) to a coarse ash mode in the outer area (Fig. 10e). It is important to note that L2 usually displays laminae of coarse ash and fine-medium lapilli even in the outer area. As a whole, a crude normal grading characterizes the PDC deposit, except where L1a is present.

Fig. 10
figure 10

Typical grain-size histograms of the main units of the PDC deposit. Grain-size ranges are in Phi units (Φ = −log2 d)

Full size image
Fig. 11
figure 11

Inman (1952) graphical parameters for the grain-size distribution of the PDC samples. The pyroclastic flows (solid line) and surge fields (dashed line) are from Walker (1971). The directed blast field (red line) is modified from Komorowski et al. (2013)

Full size image

Taking into account this information, most of the PDC deposit shows the grain-size characteristics of pyroclastic surges and in particular directed blasts (Fig. 11). In general, the grain-size distribution in the PDC deposit is correlated to the facies. The coarser material is found in the valley facies and in Cachiyacu area while the finer material is represented in the outer zone and in the upper units in Pigunchuela area.

Pyroclastic fallout deposit

Distribution and stratigraphy

The fallout deposit covers at least 70 km2 with an elliptical shape elongated toward the west due to the dominant wind direction during the eruption. In the study area, the thickness of the fall deposit varies from 120 cm near Chachimbiro hot springs to 7 cm in the Western Cordillera highlands where it has been conserved in peat bogs. In the highlands, the fallout layer is found within the andisol at ∼1.5 m depth, and about 0.5 m below a more recent lapilli fallout deposit from the Hugá phase (Fig. 2b). Near the peak La Viuda, the fallout lies directly on top of the upper unit of the PDC deposit (Fig. 2c). The eruption products are compositionally similar to those from the PDC deposit and consist of 80 vol% of rhyodacite pumices and 20 vol% of accidental material (principally lutite). These values correspond to an accidental-rich fallout. As the fallout is poorly exposed in the field, no grain-size or density analysis has been performed on this deposit.

Volume and plume height

Thickness and maximum pumice size obtained in 25 outcrops allowed the interpretation of isopach and isopleth maps (Fig. 12; Online Resource 3). The volume of deposits (Table 6) has been estimated using different methods (Pyle 1989; Fierstein and Nathenson 1992; Legros 2000; Bonadonna and Houghton 2005; Bonadonna and Costa 2012). The result from the Power Law model has been discarded due to the low m coefficient (1.174). Bonadonna and Houghton (2005) show that a low m coefficient would produce a large uncertainty on the volume estimate. The exponential-1 segment (Pyle 1989) and 1 isopach (Legros 2000) models lead to low, unrealistic estimated volumes. The results from the exponential-2 segment (Fierstein and Nathenson 1992) and Weibull (Bonadonna and Costa 2012) models are similar and fit the data series better. Therefore, we use those to estimate a range of volumes (18–27 × 106 m3). The plume height has been calculated at 13.5 km using the Weibull function (Bonadonna and Costa 2013).

Fig. 12
figure 12

Isopach and isopleth map of the fallout deposit. Main potential source locations of the eruption are indicated as well as the segment of El Angel fault

Full size image
Table 6 Volume calculations for the 3640–3510 BC fallout deposit
Full size table

Discussion

Eruption size and eruptive parameters

The parameters of the 3640–3510 BC eruption at Chachimbiro volcano are summarized in Table 7. According to the total volume of tephra (73–82 × 106 m3) the eruption must be considered moderate to large with a Volcanic Explosivity Index of 3 (Newhall and Self 1982). This eruption is the largest Holocene eruption of Chachimbiro volcano. The Dense Rock Equivalent volume of the juvenile material is estimated at 49-53 × 106 m3, which corresponds to 4.1 in magnitude (Pyle 2000). The intensity (9.8) and the height of the eruptive column (13.5 km) that produced the tephra fall ranks as sub-Plinian, similar to the 1992 Mount Spurr and 2006 Tungurahua eruptions (McGimsey et al. 2001; Eychenne et al. 2013). As thickness variations in this topographically controlled deposit are important (Table 4), it is difficult to precisely determine the uncertainty for these results. As regards the PDC deposit, our field work (69 thickness measurements, >1 measurement/km2) allows us to consider reliable our estimate of its volume. Conversely, due to the poor exposure of the pumice layer in the Chachimbiro highlands, our estimates for deposits of the sub-Plinian phase are tentative.

Table 7 Eruptive size and parameters
Full size table

Source of the 3640–3510 BC eruption

Beate (2001) suggested that the source of the fallout deposit was a hill called Rodeopamba in eastern Chachimbiro foothills (Fig. 6 and 12). This hill, however, is covered by a thick andisol which includes the fallout deposit, as exposed in a recent road cut. Therefore, even if this hill represents the remains of a dome, the dome is older than the eruption. The PDC and fallout deposit maps (Fig. 6 and 12) indicate that the eruption did not come from the central vent (Loma Albují) or from the Cochapata hills. This interpretation is confirmed by the chemical composition of the PDC and fallout samples, which is clearly different from that of the rocks sampled in the highlands (Fig. 3). Isopach and isopleth maps clearly indicate a source from or close to the peak La Viuda (3314 m a.s.l.), about 6.3 km east of Chachimbiro summit, even though La Viuda peak is formed by old andesite lava flows from a former edifice belonging to the Huanguillaro construction phase (Bernard et al. 2011). At the base of the peak La Viuda, we observed an outcrop which consists of deformed shale belonging to the Cretaceous Natividad Unit (Vallejo et al. 2006). Clasts of this material are abundant in the eruption deposit, particularly in the fallout layer. Below La Viuda, on both sides of the Cachiyacu valley, large rhyodacite blocks with the same composition as in the eruption deposits are included in a landslide deposit. Therefore, we suggest that the source of the eruption was a moderate-sized dome emplaced in this valley at the foot of La Viuda peak at an altitude of 2,800–2,900 m asl. According to the calculated DRE volume (Table 7), this dome was probably about 600–700 m wide, and 200–250 m high. It is likely that most, if not all, of the dome was destroyed by the eruption. No further volcanic activity has been observed in this area. The source location coincides with a segment of the El Angel fault system that passes through Chachimbiro volcano (Fig. 12). This fault system marks the northern limit of the Ibarra Basin (Andrade 2009).

Eruptive dynamics

PDCs are hot mixtures of gas and volcanic particles driven by gravity (Druitt 1998; Branney and Kokelaar 2002). Such mostly turbulent currents are produced by various mechanisms including, but not limited to: (1) fallback or continuous fountaining of an eruptive mixture from a vertical eruption column; (2) gravitational collapse of a growing lava dome or a steep lava front; (3) rapid decompression of a shallow magmatic body (Valentine and Fisher 2000). PDC deposits are characterized by a large variety of facies with two end-members, pyroclastic flows and surges (Fisher and Schmincke 1984; Cas and Wright 1987; Branney and Kokelaar 2002). Pyroclastic surges are dilute currents forming deposits that mantle topography and tend to be thicker in depressions (Valentine 1987). Their deposits are typically laminated and enriched in dense lithics and crystals. There are several subtypes of pyroclastic surges: base surges, ash-cloud surges, ground surges, and lateral blasts (Druitt 1998). Pyroclastic flows are highly concentrated currents forming deposits controlled by topography and tending to pond in valleys. They are generally massive, and poorly sorted deposits. Sometimes, the deposits present a subtle grading of large clasts. This grading is inverse for vesicular pyroclasts and normal for dense pyroclasts. There are also subtypes of pyroclastic flows mostly based on the components of the deposit or depositional processes: i.e., ignimbrites, ash flows, scoria flows, block-and-ash flows (Cas and Wright 1987; Fisher 1990; Freundt et al. 2000). The distinction between a pyroclastic flow and a pyroclastic surge is not always obvious and it is probable that a continuum of facies and processes exists between these two end-members (Wilson and Houghton 2000; Burgisser and Bergantz 2002).

In order to identify the source mechanism of this PDC, we have chosen two complementary approaches: (1) sedimentology of the PDC deposit; (2) mobility of the PDC.

The PDC deposit has the following characteristics:

  • A fan-shaped deposit with an aperture angle of 100°. The deposit mantles the landscape with an irregular thickness showing pounding in the valleys;

  • A crude stratification with at least one massive layer at the base and a more laminated layer to the top;

  • A juvenile-rich deposit with a wide range of density and an apparent vertical evolution from poorly vesicular clasts at the base to highly vesicular clasts to the top;

  • A generally fines-depleted, lapilli rich, well- to poorly sorted deposit.

Based on these characteristics, the 3640–3510 BC PDC deposit fits mostly the description of a high-energy turbulent stratified PDC deposit. In detail, these characteristics are very similar to those of directed blast deposits from decompression of highly pressurized cryptodomes such as at the eruption of Mount St. Helens in 1980 (Hoblitt et al. 1981) or lava domes (Pelean activity) such as at the eruption of Merapi volcano in 2010 (Komorowski et al. 2013). The graphical parameters of the grain-size distribution (Fig. 11) show also a great correspondence with surge and directed blast deposits. Also, the global vesicularity distribution (Fig. 9) is similar to that of the Bezymianny deposit (Belousov et al. 2007). Compared to other eruptions with directed blasts, the 3640–3510 BC event most resembles the Soufrière Hills and Merapi examples as we did not find the abundant soil erosion features described for Bezymianny and Mount St. Helens deposits (layer A in Belousov et al. 2007). Compared to the typical stratigraphy for blast deposits proposed by Belousov et al. (2007), the lower layer (L1) is similar to the layer B, while the upper layer (L2) has characteristics similar to those of the layer C. Komorowski et al. (2013) also describe bipartite units in the Merapi 2010 PDC deposit. They interpret the lower layer as the deposit of the erosive flow head while the upper layer corresponds to the turbulent upper part and wake of the PDC. This interpretation seems valid, and in particular explains the longitudinal and lateral variations of thickness and facies in the 3640–3510 BC PDC deposit. Below L1, the fine-grained basal layer (L1a) is interpreted as segregation of fines decoupling from the front of the PDC when it reaches a topographic barrier such as a transversal valley. This hypothesis agrees with the fact that this layer mostly occurs in the interfluve areas and also fits the experimental results obtained by Andrews and Manga (2011).

Some mobility factors have been calculated from the distribution of the PDC deposit. The travel distance (L) of the 3640–3510 BC eruption is about 10.5 km and the drop height (H) is estimated to be 800–900 m. The apparent friction coefficient (Heim 1932) of the 3640–3510 BC PDC (H/L = 0.076–0.086) is extremely low as well as the aspect ratio of the deposit (AR = 3–4 × 10−4; definition of Dade 2003). These parameters are clearly incompatible with typical dome collapses that have typical H/L > 0.2 and AR > 10−2 (Calder et al. 1999; Dade 2003; Saucedo et al. 2005). The low AR is common for surges and ignimbrites (Dade 2003), but the extremely low H/L (<0.1) is scarce. The mobility factor A/V 2/3 (Hayashi and Self 1992) calculated for the 3640–3510 BC PDC is around 515. This value is greater than those of dome-collapse flows (34 ± 4) and relatively close to surges (430–460) of Soufrière Hills (Calder et al. 1999). This value is low compared to that of large directed blast associated to cryptodomes decompression (Bezymianny ∼1,460; Mount St. Helens ∼2,610) and intermediary compared to medium size directed blast associated to dome decompression (Soufrière Hills ∼360; Merapi ∼750; Montagne Pelée ∼815) (Belousov et al. 2007; Komorowski et al. 2013). At Chachimbiro, the fact that the main dispersion axis of the PDC was oriented almost perpendicular to the local slope may explain the relatively low value of A/V 2/3 compared to Merapi and Montagne Pelée events.

The mobility of the PDCs can also be assessed using their ability to surmount obstacles. The 3640–3510 BC PDC crossed several valleys and was able to climb a hill at 8 km from the source. Using the height gained (H G = 253 m) between the bottom of the Pigunchuela valley and the top of the Mindaburlo hill (Fig. 13), the calculated minimum velocity of the blast is about 70 m s−1 at 5 km from the source. A more realistic approach developed by Francis and Baker (1977) estimates a frictional loss (F) that takes into account the height lost along the emplacement path (H L = 553–653 m):

Fig. 13
figure 13

Topographic profile between La Viuda peak and Mindaburio hill (see location in Fig. 6) with principal stratigraphic sections of the PDC deposit. The legend for the sections is the same as Fig. 7

Full size image
$$ F=1\hbox{--} {\left({H}_{\mathrm{G}}/{H}_{\mathrm{L}}\right)}^{1/2} $$
(1)
$$ v=10{\left(2\times g\times {H}_{\mathrm{G}}/100\times F\right)}^{1/2} $$
(2)

This model gives 120–135 m s−1 at 5 km from the source. These results are compatible with velocities measured or inferred from other directed blasts (Belousov et al. 2007; Komorowski et al. 2013). According to Andrews and Manga (2011), PDCs do not pass barriers >1.5 times the current thickness. Therefore, the 3640–3510 BC PDC was at least 170 m thick at 8 km from the source.

Reconstruction of the 3640–3510 BC eruption

Magma storage, evolution and migration

The chemical composition of the 3640–3510 BC eruption products suggests that the magma was issued from the Chachimbiro magmatic system but that it underwent a more extended differentiation process than the other rocks from the compound volcano. Amphibole crystallization shows that this process occurred at two depths (about 14.4 and 8.0 km) correlated with a trend of decreasing temperatures. We propose that the hot deep reservoir is the main Chachimbiro magma reservoir while the shallow reservoir had an independent evolution from the Chachimbiro feeding system and is probably controlled by El Angel fault system. Because of differentiation processes or tectonic movements, the magma started its migration from the shallow reservoir toward the surface. As there is no evidence of disequilibrium on the minerals (in particular the amphiboles), this migration must have been relatively rapid. When the cool viscous magma reached the surface it accumulated as a lava dome, the external part of which cooled and outgassed.

Source-transport-depositional system of the PDC

We use the concept of a source-transport-depositional system developed by Fisher (1990) and modified by Komorowski et al. (2013) to describe the emplacement of the 3640–3510 BC PDC. The reason why the lava dome exploded is not known. Komorowski et al. (2013) observed that directed blasts are generally associated to domes or cryptodomes with very high growth rates. They suggest that elevated magma flux reduced the time required to outgas the inner part of the dome and create an overpressure inside the dome. This hypothesis is supported by the density measurements that indicate more highly vesicular juvenile fragments in L2 than in L1. In the case of the 3640–3510 BC eruption, it is probable that a part of the volatile-rich magma that participate to the sub-Plinian phase was stored in the conduit rather than in the lava dome. This magmatic column may have also exerted pressure on the lava dome. Based on the historical data available on Merapi, Soufrière Hills, and the Montagne Pelée, we propose a pre-blast dome growth rate of the order of 10 to 38 m3 s−1 (Komorowski et al. 2013). According to the volume of magma erupted during the PDC phase, the dome growth phase must have lasted between 13 and 49 days.

Lateral or directed blasts related to the decompression of a gas-pressurized magma body may follow three scenarios: (1) explosion of a cryptodome as the result of a giant landslide (i.e., Bezymianny 1956, Mount St. Helens 1980; Hoblitt et al. 1981; Belousov 1996); (2) failure and explosion of a growing lava dome in association with a giant landslide (i.e., Tacana Late-pleistocene and Soufrière Hills 1997; Voight et al. 2002; Macías et al. 2010); (3) failure and explosion of a growing lava dome without giant landslide (i.e., Montage Pelée 1902 and Merapi 2010; Boudon et al. 2005; Komorowski et al. 2013). There is no evidence of a landslide scar or a debris avalanche deposit associated to the 3640–3510 BC eruption, so a simple dome failure and explosion seems likely. In the case of Merapi event, the directed blasts (at least two main episodes) are included in a complex series of volcanic processes such as retrogressive dome collapses (Charbonnier et al. 2013). The 3640–3510 BC PDC was different in that it was shortly followed by the sub-Plinian phase without previous or further activity. Therefore, we interpret the deposits of the PDC and fallout as the result of a monogenetic vent within the Chachimbiro volcano. The dome’s failure could have been triggered by a combination of factors. The high growth rate probably produced a highly unstable lava dome. In addition, the formation of a poorly permeable carapace is likely to cause internal gas overpressure. Finally, a tectonic event associated to El Angel fault system or a volcano-tectonic earthquake due to the magma migration toward the surface might have contributed to the dome’s destabilization.

The south-southeast direction of the explosion was clearly controlled by topographic elements that include the peak La Viuda and El Angel fault scarp (ENE in this sector). The decompression of the lava dome produced a lateral explosion that removed most of the dome and formed a high-energy turbulent stratified PDC. The front of the >170 m-thick PDC moved rapidly, allowing the PDC to cross several valleys and to overcome topographic obstacles such as Mindaburlo hill. Sedimentation was strongly controlled by the topography, and caused ponding of the most concentrated part of the PDC in valleys and segregation of more dilute PDCs able to spread further laterally and longitudinally. Taking into account the stratigraphy of the 3640–3510 BC PDC, the velocity estimated (120–135 m s−1 at 5 km from the source), and the historical record of directed blasts (Belousov et al. 2007; Komorowski et al. 2013) it is probable that the PDC was initiated in only a few minutes. The directed blast was followed by a vertical eruption that produced a sub-Plinian column that lasted about 1 h (Table 7). When the eruptive column reached the low stratosphere it rapidly drifted to the west producing a pyroclastic fall in the Chachimbiro highlands.

Hazard assessment

The identification of the source and the dynamics of an eruption is critical for hazard assessment and monitoring purposes. Lava domes are common features in the Ecuadorian Volcanic Front, but the 3640–3510 BC eruption differs from other comparable eruptions in some aspects. There are only two documented rhyodacitic domes in the Ecuadorian Volcanic Front (EVF): (1) Pilongo, in the Iliniza volcanic complex (Hidalgo et al. 2007); and (2) Fuya-Fuya, in the Mojanda Fuya-Fuya volcanic complex (Robin et al. 2009). Pilongo is a moderate-size (650 m diameter, 240 m high) satellite dome, located 6 km northeast of Iliniza summit, and lies directly on top of a fault system traversing Iliniza volcano. Fuya-Fuya is a large (3 km diameter, 600–800 m high) polygenetic dome. None of these domes produced a directed blast. In the EVF, Guagua Pichincha, Imbabura, and Chachimbiro volcanoes have experienced directed blasts associated with giant landslides (Robin et al. 2008; Bernard et al. 2011; Le Pennec et al. 2011). Directed blasts, some related to giant landslides, have been documented at numerous volcanoes worldwide (Belousov et al. 2002; Boudon et al. 2005; Macías et al. 2010; Komorowski et al. 2013) and are thought to be a common phenomenon that must be taken into account when assessing volcanic hazards. The 3640–3510 BC eruption highlights the fact that satellite domes formed in a single eruption can also produce directed blasts. Such satellite domes also enlarge significantly the area potentially affected by high-energy PDCs. Accordingly, hazard maps must better take into account local tectonic elements. For monitoring purposes, the dome growth rate should be used to assess the probability of a dome blast, based on data from modern eruptions (Komorowski et al. 2013).

Conclusions

Our study provides evidence of a large eruption dated between 3640 and 3510 years BC. Mapping and sedimentological analysis indicate that the eruption was produced by the sudden decompression of a medium-sized rhyodacite satellite dome that extruded 6.3 km east of Chachimbiro central vent. Rock and mineral chemistry point out that the magma of this eruption came from a deep reservoir (∼14.4 km deep) shared with the Chachimbiro feeding system. Then, it migrated along El Angel fault up to a shallower independent reservoir (∼8.0 km deep) before reaching the surface.

While the surface of the dome was slowly cooling and outgassing, the internal pressure increased because gas and magma were fed into it, and pressure eventually surpassed the lava resistance and produced a violent directed blast. A tectonic or volcano-tectonic earthquake might also have triggered destabilization of the lava dome. The high velocity (>120 m s−1) pyroclastic density current produced by dome decompression traveled more than 10 km and was able to cross several deep valleys and overpass a >200 m high hill 8 km from the source. It spread over 62 km2 and destroyed a large area of the Urcuquí district, burying remains of human settlements. The directed blast was followed by a sub-Plinian eruption that produced a 7–11 × 106 m3 (DRE) fallout deposit in Chachimbiro highlands.

This eruption was the largest from Chachimbiro volcanic complex in the last 6,000 years. Even though apparently rare, eruption of satellite domes must be considered for hazard assessment in the Ecuadorian Volcanic Front as they can produce highly explosive eruptions. In particular, the probability of a directed blast event must be considered as high in the case of a fast-growing dome.