Introduction

Plinian eruptions are amongst the most intense and potentially destructive natural events on Earth in terms of energy released and rates at which material volumes are mobilised. The reconstruction of eruptive events from the geological record of a particular volcano, including the possible extent, distribution and thermal structure of pyroclastic density currents (PDCs), is essential for understanding the potential hazards from future eruptions (e.g. at Somma-Vesuvius volcano; Zanella et al. 2014). A key approach for understanding the complex dynamics of explosive eruptions is the study of emplacement temperatures of PDCs. The temperature of emplacement of PDC deposits, as described by several authors (McClelland et al. 2004; Cioni et al. 2004; Zanella et al. 2007, 2014; Porreca et al. 2006, 2008; Paterson et al. 2010; Lesti et al. 2011), is a function of the original temperature of the magma at the vent, the collapse level from the eruption column (Woods and Bursik 1991; Shea et al. 2011), the transport regime in the PDC (McClelland et al. 2004; Lesti et al. 2011), the heat loss due to interaction with the environment (Gurioli et al. 2005; Di Vito et al. 2009), magma-water interaction (Thomas and Sparks 1992; Porreca et al. 2006, 2008) and the nature and size of components constituting the PDC (Martí et al. 1991; Woods and Bursik 1991; Thomas and Sparks 1992).

Paleomagnetic methods provide reliable tools for determining emplacement temperatures of ancient ignimbrites (McClelland et al. 2004; Paterson et al. 2010; Lesti et al. 2011). In this work, we carried out a detailed study of partial thermal remanent magnetization (pTRM) of lithic clasts embedded in three distinct ignimbrites emplaced during the 4.6 ka Plinian Fogo A eruption on São Miguel Island, in the Azores Archipelago. In a methodology paper, Pensa et al. (2015a) verified the reliability of thermal results from pTRM of lithic clasts by comparing them with equivalent data from reflectance (Ro%) of charcoal fragments, at sites where both lithics and charcoal were present close together in the same outcrop within the ignimbrites associated with the 4.6 ka Fogo A eruption.

In this companion paper to Pensa et al. (2015a), we reconstruct the emplacement temperature distribution of the Fogo A ignimbrites using pTRM data obtained from both areal and up-stratigraphy sampling. We consider the sedimentological characteristics of the three ignimbrites as the basis for analysing the factors that produced differences in emplacement temperatures. This study represents therefore the first attempt to evaluate and compare emplacement temperatures of multiple ignimbrites originating during the same eruptive sequence. By comparing the eruptive temperature of the magma, determined using the magnetite-ilmenite geothermometer (Powell and Powell 1977; Spencer and Lindsley 1981; Andersen and Lindsley 1985), and the emplacement temperature of the ignimbrites, we can assess the heat loss experienced by each PDC and ignimbrite as a consequence of eruption dynamics, flow transport and emplacement processes.

Geological setting of Fogo volcano

Fogo volcano is one of the three active quaternary strato-volcanoes that dominate São Miguel Island, the largest island of the Azores Archipelago (Portugal) in the mid-Atlantic Ocean (Fig. 1). The volcano lies at the centre of São Miguel Island and has a diameter of 15 km and an elevation of ~950 m a.s.l. It is characterised by a small central summit caldera (3 km × 2.5 km) as a consequence of repeated collapses and explosive eruptions (Moore 1990).

Fig. 1
figure 1

Location of Azores Archipelago and Sao Miguel Island; dispersal area of fallout deposits of Fogo A eruption and the 1653 eruption (from Walker and Croasdale 1970); isopach map of basal fallout deposit of the Fogo A eruption (from Walker and Croasdale 1970)

Full size image

The sub-aerial depositional history of Fogo Volcano started during the late Pleistocene (Gandino et al. 1985; Wallenstein 1999) and is marked by several voluminous trachytic Plinian deposits from summit eruptions and interbedded with mafic lavas from smaller flank eruptions (Wallenstein 1999). Because of prolific vegetation, there is a scarcity of outcrops of the older deposits, and so a detailed stratigraphy has only been reconstructed for the last 40,000 years. Along the southern flank of the volcano, the deposits of three trachytic Plinian eruptions are exposed: Roída da Praia (ca. 15 ka), Ribeira Chã (ca. 8–12 ka) and Fogo A (ca. 4.6 ka) (Wallenstein 1999). In 1653, a sub-Plinian eruption was recorded (Booth et al. 1978), which represents the youngest explosive eruption at Fogo volcano.

Fogo A eruption sequence

Summary of previous studies

The Fogo A deposit (4435–4672 radiocarbon years, Shotton et al. 1968, 1969) is the last major trachytic Plinian eruption related to the pre-historic activity of Fogo Volcano. This sequence was first documented by Walker and Croasdale (1970), who undertook a detailed study of the voluminous basal pumice fallout deposit (Fig. 1). Their study was one of the first to quantify a Plinian fallout deposit and is a famous reference in modern volcanology.

The Fogo A eruption produced an eruption column between 21 km (Bursik et al. 1992) and 35 km high (Walker and Croasdale 1970), and the material emplaced on land has an estimated minimum volume between 1.2 km3 (Walker and Croasdale 1970) and 1.37 km3 (Burden et al. 2013). Capaccioni et al. (1994) first noted the presence of pyroclastic flow deposits, which overlie the basal fallout sequence. Wallenstein (1999) identified two members and 17 units within the stratigraphy of the Fogo A sequence.

Stratigraphy

The excavation of new freeway outcrops in 2011 and 2012 enabled us to reconstruct a more detailed stratigraphy of the Fogo A sequence (Fig. 2), which can be divided into three members: FGA1, FGA2 and FGA3 (Pensa et al. 2015b).

Fig. 2
figure 2

a Representative stratigraphic log of Fogo A sequence, b representative pumice clast samples showing the change in colour from the base to the top of the sequence; c vertical variation of lithic clasts, pumice and ash content in the three ignimbrites

Full size image

FGA1—The first member represents the initial phase of the Fogo A eruption, involving the establishment of a high stable Plinian eruption column. The deposits are recognisable all around Fogo volcano (Fig. 1) and are trachytic in composition (~63 % SiO2). From the base, the first member is composed of 50 cm of alternating thin layers of grey ash (units FGA1a1–a4) and well sorted, vesiculated (65–70 %) white pumice lapilli (max 3 cm in diameter) fallout layers (units FGA1l1–l3), overlain by ~3.5 m of a moderately sorted, massive to diffusely stratified, highly vesicular (~70–75 %), pumice lapilli and bombs (max 40 cm diameter) fallout deposit (units FGA1l4a/4b–l5). The phenocryst content of the pumices varies between 1 and 3 % in volume. The upper unit FGA1l5 is characterised by the appearance of syenite accessory lithic clasts and some grey, banded pumice clasts, although white pumice still dominates.

FGA2—The second member is interpreted to represent the beginning of instability in the Plinian column, marked by partial collapses, and the appearance of ignimbrite deposits inter-bedded with fallout deposits. This sequence consists of two valley-ponded, confined intra-Plinian ignimbrites and laterally contiguous veneer surge deposits (units FGA2i1–i2), inter-layered within further Plinian fallout pumice lapilli layers (units FGA2l6–l7). These two ignimbrite deposits are only observable along the south flank of Fogo volcano. Both fall and pyroclastic flow deposits constituting the second member have a less evolved composition than the first member; the silica content varies between ~61 and ~59 % (Pensa et al. 2015b). Both intra-Plinian ignimbrites are ash matrix-supported, massive and poorly sorted, but their components are visibly different. The first intra-Plinian ignimbrite, FGA2i1, reaches a maximum thickness of 10 m in the paleo-valleys, and a maximum of 1 m on the nearby paleo-ridges as veneer surge deposits. It is pinkish in colour, and therefore it has been called the “pink intra-Plinian ignimbrite” (Fig. 2). It is mainly composed of highly vesiculated (74–75 %), pink trachytic-trachydacitic pumice and grey-banded pumice lapilli (~61 % SiO2), and also dense (<50 % vesicles), crystal-rich pinkish pumice clasts (5 % in volume). The phenocryst content of the juvenile clasts varies from ~6 to 25 % (vesicle free) in the most vesiculated pumices and ~54 % (vesicle free) in the densest pumice clasts and is principally represented by sanidine. No stratification is recognisable but there are lenses of rounded pumice lapilli at the top of the deposit and irregularly distributed at different stratigraphic heights. The lithic content reaches 10 % by volume and consists of trachytic lava and syenite, with reddish hydrothermally altered clasts. The ash content represents 60 % by volume.

The pink intra-Plinian ignimbrite is overlain by a pumice lapilli fallout deposit (FGA2l6), which is 75-cm thick, consisting of angular, pink, well-sorted, highly vesiculated (~70–75 %) trachytic and grey-banded pumice lapilli (average maximum diameter 9 cm).

This fall-out deposit is overlain by the second intra-Plinian ignimbrite (FGA2i2), which reaches a maximum thickness of ~ 5 m along the paleo-valleys, and a few centimetres on top of paleo-hills, where it is represented by a veneer surge ash deposit. It mainly consists of black and grey-banded, highly vesiculated (up to 75–77 %) pumice lapilli, and by black dense highly porphyritic pumice clasts (5–10 % in content). Because of the very dark colour of the pumices, we refer to this unit as the “black intra-Plinian ignimbrite” (Fig. 2). The composition of the black intra-Plinian ignimbrite is slightly less evolved than the pink intra-Plinian ignimbrite, with a SiO2 content of ~59 % (Fig. 3). Occassionally, in the outcrops most proximal to the vent, the juvenile clasts show evidence of plastic deformation and weak welding. The phenocryst content varies from ~10 % (vesicle free) within the most vesiculated pumices to ~40 % (vesicle free) within the densest pumices, with the phenocryst population mainly represented by euhedral sanidine crystals. The lithic clasts consist of trachytic lava fragments, constitute less than 5 % in volume, and their dimensions do not exceed 2 cm in diameter. The ash content constitutes 60 % by volume.

Fig. 3
figure 3

Dispersal area of the three ignimbrites of Fogo A eruption sequence and distribution map of the sampled paleomagnetic sites: pink intra-Plinian ignimbrite (pink squared pattern and paleomagnetic sites in green squares), black intra-Plinian ignimbrite (purple horizontal lines pattern and paleomagnetic sites in red circles), dark brown ignimbrite (yellow obliques lines and paleomagnetic sites in orange triangles)

Full size image

The black intra-Plinian ignimbrite is overlain by another pumice fall deposit (FGA2l7), that is 65-cm thick, composed of dark brown, grey and banded, well-sorted, highly vesiculated (~70–75 %) angular pumice clasts (average maximum size 13 cm). The presence of fallout deposits between and above the pink and black intra-Plinian ignimbrites indicates that, despite the partial collapses, the eruption column was still buoyant and sustained. The similar average pumice size of the FGA2 fallout deposits compared to those of FGA1 also suggests that the eruption column maintained its height during the eruption, even during the partial collapse phase.

FGA3—The third member, FGA3, represents the deposits of the final collapse of the eruption column, which produced the thickest and climactic ignimbrite (FGA3i1). This deposit is dark brown in colour and for this reason, it is named the “dark brown ignimbrite” (Fig. 2). The ignimbrite is ash supported, generally massive and poorly sorted and shows some stratification defined by trains of brown small rounded pumices and ash layers. The juvenile content is composed of dark brown and grey-banded, very highly vesiculated (77–80 %) pumice lapilli that constitute 15 % by volume of the entire deposit; the composition is the least evolved of the entire eruption sequence (~58 % silica). The sanidine phenocryst content varies between 5 and 10 % (vesicle free), and the maximum crystal size is around 2 mm. The ignimbrite lithic content is 20 % by volume, represented by trachytic lava and syenitic clasts, with an average size of ~10 cm. Matrix-supported, lithic clast lenses with blocks up to ~1 m in diameter are occasionally present at various heights. The thickness of this final unit is significantly greater than the pink and black intra-Plinian ignimbrites, reaching ~20 m in outcrops 4 km from the vent, along the south flank of Fogo volcano, where it is mostly confined into the paleo-valleys. On the northern flank, it is ~50-m thick, ~2 km from the vent.

Methods

Partial thermal remanent magnetization (pTRM) of the lithic clasts

This technique is based on the acquisition of a partial or total thermal magnetization (TRM) of lithic clasts that were incorporated into a pyroclastic flow during an explosive eruption. A full description of this geophysical technique can be found in Bardot and McClelland (2000) and McClelland et al. (2004). In a simplified model, if a pyroclastic deposit is emplaced above ambient temperature, the lithic clasts incorporated in the deposit are heated. During the heating process, part of or the entire magnetization of the clasts will be thermally unblocked. At the moment of emplacement, the clasts cool to ambient temperature of the deposit and acquire a new partial magnetization which will be oriented parallel to the Earth’s magnetic field. If the emplacement temperature is higher than the Curie temperature of the clasts, then all the clasts will have one single component of magnetization parallel to the geomagnetic field. If the emplacement temperature is lower than the Curie temperature, then the clasts will acquire a new partial thermal magnetization (pTRM) oriented parallel to the geomagnetic field. In this latter case, the clasts will have two components of magnetization: a low temperature (LT) component acquired during the last heating event and a high temperature (HT) randomly oriented component that was previously acquired during its thermal history.

The emplacement temperature of the lithic clast can be estimated through progressive stepped, thermal demagnetization in the laboratory. An increment of the total remanence is unblocked at each demagnetization step. The estimated emplacement temperature of a clast with two components of magnetization is given by the temperature interval between the LT and HT components. By contrast, in the case of a single component oriented parallel to the expected geomagnetic field, it is possible to estimate the minimum emplacement temperature of the clast, which is given by the Curie temperature (e.g. T = 580 °C for magnetite) of the clast.

Paleomagnetic sampling

The collection of lithic clasts for pTRM analysis involved sampling lithic fragments from different paleotopographic localities within the ignimbrites, characterised by thickness variations, to determine if these factors are reflected also in emplacement temperature variations.

In a first sampling step, according to the paleomagnetic method and sampling strategy described in Pensa et al. (2015a), the paleomagnetic sites were chosen where possible in proximity to carbonised wood fragments in the ignimbrites. This allowed the comparison of the data obtained with the pTRM analysis with data from reflectance analysis of the charcoal fragments, reported in Pensa et al. (2015a). After this first sampling at the scale of the outcrop, we performed more complete sampling of the all three ignimbrite units.

A total of 15 paleomagnetic sites were selected for the lithic clast sampling for TRM analysis from the three ignimbrites (Fig. 3; Table 1). A total of 44 lithic clasts were collected from five paleomagnetic sites (FO-06, FO-07, FO-09, FO-20 and FO-21) from the pink intra-Plinian ignimbrite (FGA2i1), widely distributed along the south flank of the volcano, where the ignimbrite is present, at a distance between 4.5 and 5.5 km from the summit. A total of 29 lithic clasts were sampled from the black intra-Plinian ignimbrite (FGA2i2) from three paleomagnetic sites (FO-10, FO-22 and FO-26), again distributed along the south flank of the volcano at distances between 4 and 4.5 km from the vent, matching the distribution of the ignimbrite. A total of 67 lithic clasts were collected from the dark brown ignimbrite (FGA3i1) from seven paleomagnetic sites: five from the north flank (FO-04, FO-05, FO-15, FO-16 and FO-18) at distances between 3.5 and 6.5 km from the vent and two from the south flank (FO-11 and FO-23) at a distance between 4 and 4.5 km from the vent (Fig. 3, Table 1).

Table 1 Summary of the paleomagnetic localities sampled for each ignimbrite of Fogo A eruption sequence, number of lithic clasts collected in each locality, analysis of magnetic components and emplacement temperatures of each paleomagnetic site
Full size table

The lithology of the fragments collected is mostly fresh trachytic lava and, secondarily, syenite; all altered clasts were discarded. The size of the lithic fragments collected ranged from 2 to 10 cm in diameter. In the field, all the selected clasts were oriented by a magnetic compass, marking the dip and the strike in situ on the flattest exposed surface. Subsequently, the oriented clasts were set in white non-magnetic plasticine (a type of modelling clay) and then, following the procedure indicated by Cioni et al. (2004), positioned into a rigid plastic cylinder with standard dimensions (diameter Ø = 25 mm, height h = 22.5 mm).

Eleven to 13 thermal steps, at increasing temperatures (40–50 °C), were chosen for the demagnetization of the samples. All the fragments were heated in an appropriate oven, equipped with a fan, and the magnetization intensity was measured after every step using a JR-6A Spinner Magnetometer at the Paleomagnetic Laboratory of University of Roma Tre, Italy. The software used to process the data is REMASOFT by Agico (Chadima and Hrouda 2006), and the paleomagnetic mean directions for each site are calculated applying Fisher statistics (Fisher 1953).

Electron microprobe analysis on Fe-Ti oxides

In order to estimate the original temperature of the magma prior to eruption, major oxide geochemistry data for Fe-Ti minerals from a total of 33 Fe-Ti oxide grains were selected within juvenile clasts representative of the entire Fogo A eruption sequence. For each mineral, the core and the rim were analysed with the purpose of detecting possible differences. In order to identify magnetite-ilmenite oxide couples, which can be used as a geothermometer (Powell and Powell 1977; Spencer and Lindsley 1981; Andersen and Lindsley 1985), we mainly selected mineral grains containing both magnetite and ilmenite. Suitable oxide couples were only found in the black intra-Plinian ignimbrite (Table 2). Determinations were made using a JEOL JXA8900R Wavelength Dispersive (WD) and Energy-Dispersive (ED) combined micro-analyser Super-probe, operating at 15 kV accelerating voltage, 20 nA beam current, and using a 2-μm defocused beam. The analysis was performed at the CSIRO Research Laboratories at Monash University.

Table 2 Geochemical analysis of the composition of magnetite and ilmenite minerals and calculated temperatures of the magma using the Powell and Powell (1977), Spencer and Lindsley (1981) and Andersen and Lindsley (1985) equations for magnetite-ilmenite geothermometer
Full size table

SEM analysis

Scanning electron microscopy (SEM) analysis was undertaken on the very fine ash (63 μm) of all three ignimbrites (FGA2i1, FGA2i2, FGA3i1), at Roma Tre University, Rome, Italy, to investigate the fragmentation process and to detect evidence of possible magma-water interaction during the eruption. The instrument used for morphological, compositional and structural characterisation of particles is a FEI SEM XL30. A total of 76 images were taken and studied (Fig. 5).

Results

Thermal remanent magnetization (TRM) results

The natural remanent magnetization (NRM) of the samples is in the range 1.06–0.924 A/m. Most of the samples are completely demagnetized at temperatures between 400 and 620 °C. These unblocking temperatures suggest that magnetite and titano-magnetite are the main magnetic minerals.

The results of the TRM analysis of all lithic clasts collected show that 101 samples out of 140 have a single magnetic component, while 21 samples have two magnetic components, one stable at high temperatures (HT component) and one at low temperatures (LT component). The remaining 18 samples were discarded because they display unstable behaviour (Table 1). The demagnetization process revealed different types of paleomagnetic behaviours. We identify five different types (Type A, B1, B2, C and D) of paleomagnetic behaviours (Fig. 4), which build on previous interpretation by Cioni et al. (2004) and Zanella et al. (2007). Type A is characterised by lithic clasts showing a single component, the direction of which remains fixed but it is not close to the Azorean Geocentric Axial Dipole (GAD) field during the Fogo A eruption. This indicates that the original clast magnetisation was not overprinted, neither partially or totally. The demagnetization pattern displays an anomalous trend, with very stable to even increasing intensity with increasing temperature, which indicates a high coercivity magnetic mineralogy (Fig. 4b). This type of clast may either indicate a cold depositional environment or a clast the mineralogy of which prevented the crossing of the unblocking temperature during the new thermal event.

Fig. 4
figure 4

Thermal demagnetization data representative of the five different paleomagnetic behaviours: a schematic reconstruction of the different origin of lithic clasts and their thermal histories (see text for further explanation). b Equal area stereonets of mean remanence directions, orthogonal plot and demagnetization curve (solid dots, lower hemisphere; open dots, upper hemisphere) of type A lithic clasts showing single of magnetization oriented far from the Azorean magnetic field direction. These lithic clasts carry their original TRM, completely not affected by the reheating from the flow. c Equal area stereonets of mean remanence directions, orthogonal plot and demagnetization curve (solid dots, lower hemisphere; open dots, upper hemisphere) of type B1 lithic clasts showing single component oriented parallel to the Azorean magnetic field direction. The original TRM of these clasts was deleted by the reheating due to the eruptive event. The new TRM records emplacement temperature above the Curie temperature. d Equal area stereonets of mean remanence directions, orthogonal plot and demagnetization curve (solid dots, lower hemisphere; open dots, upper hemisphere) of type B2 lithic clasts showing single component, which is almost totally demagnetised at lower temperature in respect to type B1. The direction is close to the Azorean magnetic field direction; the original TRM of these clasts was deleted by the syn-eruptive reheating, but the stereonet diagram display a zigzag trend. e Equal area stereonets of mean remanence directions, orthogonal plot and demagnetization curve (solid dots, lower hemisphere; open dots, upper hemisphere) of type C lithic clasts showing double component: a low-temperature (LT) magnetic component (Tb max), parallel to the Azorean magnetic field direction; and a high-temperature (HT) magnetic component randomly oriented. The high-T component corresponds to the original TRM which has been partially deleted by the eruptive reheating. The emplacement temperature of the deposit corresponds to the separation temperature between LT and HT components. f Equal area stereonets of mean remanence directions, orthogonal plot and demagnetization curve (solid dots, lower hemisphere; open dots, upper hemisphere) of type D lithic clasts showing double component, where the low-T magnetic component is parallel to the Azorean magnetic field direction; while the high-T magnetic component is randomly oriented and the high-T corresponds to the original TRM which has been only partially deleted by the syn-eruptive reheating. In type D, the breakpoint of the two magnetic components is not well defined. As displayed in the Zijderveld diagram, the path of the two components is not straight but curvy. Thermal zone shapes are based on Neri et al. (2003) and Dartevelle et al. (2004) models

Full size image

Type B is also characterised by lithic clasts with a single component, but different from type A, the orientation is close to the local GAD field at the time of the eruption. Within type B group, we discriminate: type B1, which includes lithic clasts where the demagnetization pattern shows a typical decrease in intensity with increasing temperatures and a fixed orientation (Fig. 4c) and type B2 lithic clasts, where the intensity of magnetization decays to less than 20 % at temperature much lower than the Curie temperature for magnetite and consequently show large scatter in orientation at the highest temperature steps of demagnetization (Fig. 4d). Type B clasts indicate that any original magnetization was totally overprinted by a new thermal event; nevertheless, type B1 is much more reliable in providing such information, in respect to type B2.

Type C lithic clasts display two magnetic components, which are clearly distinguishable as two differently oriented segments on the Zijderveld diagram, and the low-temperature (LT) magnetic component is oriented parallel to the local magnetic field direction as shown in the stereonet diagram (Fig. 4e). The high-temperature (HT) magnetic component is randomly oriented (Fig. 4e). The HT component relates to the original TRM which has been only partially deleted by the new thermal event. The temperature of the deposit corresponds to the intersection between the LT and HT segments on Zijderveld diagram (Fig. 4e).

Type D lithic clasts have two components, as for type C, but rather than a sharp corner on the Zijderveld diagram, the two segments display a curvy pattern (Fig. 4f). This can be attributed to overlapping blocking temperature spectra (or coercivity spectra) of the ferromagnetic grains carrying the two components (Butler 1992). This means that the clast experienced a LT thermal event like type C, but that the temperature of that event is less constrained and only a temperature interval could be approximated corresponding to the curve that connects the HT and LT segments (Fig. 4f).

The principal result of this study on the Fogo A lithics is that the paleomagnetic analysis gives reliable results as the single magnetic component of types B1 and B2 clasts, and the LT component of Type C and Type D clasts are oriented parallel to the Azorean GAD field at the time of the eruption; whereas the HT magnetic component of samples with two magnetic components (type C and type D), is randomly oriented (Fig. 4). Type A clasts have not been used for the estimation of the emplacement temperatures.

The paleomagnetic data will be presented separately for each ignimbrite unit, and the detailed orthogonal plots with single magnetic component and two magnetic components are available as supplementary material (SM - A1 – 4; SM - B1 – 3; SM - C1 – 5). Also see Table 2.

Pink intra-Plinian ignimbrite: paleomagnetic results

In the five paleomagnetic sites for the pink intra-Plinian ignimbrite, we found clasts with both single and double components of magnetization. In particular, 31 out of 44 clasts have a single component, whereas 8 clasts have 2 components of magnetization (5 samples have been discarded because they displayed unstable behaviour). Amongst the clasts with a single component, 4 samples correspond to type A, 11 samples to type B1 and 16 samples to type B2.

All the samples belonging to type B1 display magnetization stable from 480 to 620 °C, whereas the samples with type B2 magnetization are stable from 300 to 480 °C. The orientation of the magnetization in the type A samples is random, whereas that in the type B1 and type B2 samples is close to the expected GAD direction, with a mean declination of 348.1° and inclination of 52.6°.

In the second case (two components), the magnetization of the LT component of type C and type D samples (respectively five and three samples) is stable from 350 and 400 °C (only one lithic clast stable is at 300 °C). The orientations of the LT components are close to the expected GAD direction, with a mean declination of 354.3° and inclination of 49.0°, similar to the orientation of the single component of type B1 and type B2 samples. The orientation of the HT component is random (Table 1; supplementary material SM - A 1 –A 4 and Pensa et al. 2015a).

Black intra-Plinian ignimbrite: paleomagnetic result

At the three paleomagnetic sites of the black intra-Plinian ignimbrite, we found clasts with only a single component of magnetization. In particular, five samples belong to type A, 21 to type B1, and one to type B2. The magnetization of type B1 and type B2 samples is stable from 480 to 620 °C. The orientation of the single component is close to the expected GAD direction, with a mean declination of 348.5° and inclination of 54.1° (Table 1; Supplementary material SM - B 1 –B 3 ).

Dark brown ignimbrite: paleomagnetic results

At the seven paleomagnetic sites of the dark brown ignimbrite, we found clasts with both single and two components of magnetization. In particular, 43 out of 67 clasts have a single component, whereas 13 samples have two components of magnetization (HT, LT). A total of 11 samples were discarded because they display unstable behaviour. In particular, of the samples with a single component, 9 samples belong to type A, 5 to type B1 and 26 to type B2. All the samples belonging to type B1 display magnetization that is stable from 480 to 580 °C, whereas for the type B2 samples, magnetization is stable from 350 to 520 °C. The orientations of the single components in the B1 and B2 lithic clasts are close to the expected GAD direction, with a mean declination of 347.5° and inclination of 50.1°.

In the second case, the magnetization of the LT component is stable from 250 to 370 °C (only one lithic clast was stable at 400 °C and another at 180 °C). The orientations of the LT components are close to the expected GAD direction, with a mean declination of 350.1° and inclination of 52.4°, similar to the single-component orientation of type B1 and type B2 samples. We note that the LT component of lithic clasts from the south and north flanks of the volcano indicate the same range of emplacement temperatures. The orientation of the HT component is random (Table 1; supplementary material SM - C 1 –C 5 ; and Pensa et al. 2015a).

Ilmenite-magnetite geothermometer analysis

The electron microprobe analysis of the Fe-Ti oxide grains revealed that the population is composed exclusively of Ti-magnetite and ilmenite. The Ti-magnetite crystals occur throughout the Fogo A sequence, whereas the ilmenite crystals were found only within the black intra-Plinian ignimbrite.

As shown in Table 2, the Ti-magnetite oxides are characterised by FeO content that varies from ~73.66 to ~76.04 % in the pink intra-Plinian ignimbrite, ~75.60 to ~78.19 % in the black intra-Plinian ignimbrite and ~74.07 to ~75.55 % in the dark brown ignimbrite. The TiO2 component of the Ti-magnetite varies slightly between the ignimbrites (from 15.22 to 20.02 %). Ilmenite crystals were found coupled with Ti-magnetite only within the black intra-Plinian ignimbrite. They are characterised by TiO2 content between 48.97 and 48.61 %, and the FeO content varies between 43.59 and 44.17 %. The analysis of the core and the rim of the single oxide did not show variation in FeO and TiO2 content (Table 2).

By entering the oxides data obtained from the electron microprobe analysis in the worksheet created by Lepage (2003), we could extrapolate the equilibrium temperature by using different geothermometers. Using the Powell and Powell (1977) calculations, the equilibrium temperature reached by the Ti-magnetite–ilmenite couple is ~911 °C. According to Spencer and Lindsley (1981) calculations of the equilibrium temperature, it is ~912 °C, whereas according to the Andersen and Lindsley (1985) calculations, it is ~884 °C. All the three results can be approximated at 900 °C.

SEM analysis

The fine ash of all three deposits is composed of glass shards, with subordinate amounts of crystal fragments and lithics. The three ignimbrites display a greater abundance of plate-like glass shards showing tube vesicles, as well as cuspate bubble-wall shapes, and subordinate to negligible content of moss-like and blocky shards. The overwhelming majority of ash particles represent an original highly vesiculated magma, as the larger pumice clasts do. The well-preserved bubble shape and the low content of blocky fragments indicate that fragmentation was dominantly magmatic (Fig. 5; Wohletz 1983; Büttner et al. 1999).

Fig. 5
figure 5

Scanning electron microscopy images of glass shards examples. Microphotographs of the pink and black intra-Plinian ignimbrites and the dark brown ignimbrite

Full size image

Discussion

The preliminary emplacement temperature estimations of the Fogo A sequence have been presented in Pensa et al. (2015a), based on the comparison between TRM and Ro% analysis of charcoal fragments of only three key sites. Here, we discuss the whole paleomagnetic dataset obtained by analysing a total of 140 sampled lithic clasts at 15 localities.

As described by Pensa et al. (2015a), using TRM and Ro% methods together gives a more accurate and validated temperature assessment than applying each method individually. pTRM results are, in fact, strongly dependent on the number and origin of samples collected in the field, and the estimation of the emplacement temperature, based on the pure statistical evaluation of the pTRM data, can be over- or underestimated.

The nature of the lithic clasts analysed for emplacement temperature needs to be taken into consideration, as during the sampling, it is not always easy to discriminate between accessory and accidental clasts, making it possible to have a predominance of one or the other. The different source of the clasts, wall conduit/vent or flanks (accessory and accidental clasts, respectively) may render it difficult to interpret pTRM results (Fig. 4a).

Accidental lithic clasts picked up from the volcano flanks (e.g. pentagon symbol in Fig. 4a) easily equilibrate with the temperature of the pyroclastic flow deposits, depending on their size and thermal properties (e.g. Martí et al. 1991; Bardot and McClelland 2000; Cioni et al. 2004), and unambiguously record the deposit temperature as the clasts were cold at the time of incorporation into the flow and deposit. For accidental clasts, it is consequently expected to find only type B (Fig. 4c, d) for deposits above the Curie temperature and types C or D for deposits below Curie temperature, but still hot (Fig. 4e, f).

Accessory lithic clasts eroded from the conduit wall and vent (e.g. rectangle symbol in Fig. 4a) may instead have a much more complex heating history and resulting pTRM signature (Cioni et al. 2004). This is because accessory lithic clasts have an unknown history of pre-heating before being incorporated in the pyroclastic deposit, which depends largely on the depth at which they have been eroded and the time and extent of exposure to the heating at high temperature during magma ascent, and eventually pyroclast-gas mixture ascent through the conduit and plume (Fig 4a); such unknown pre-heating history makes it possible for an accessory lithic to land at a higher temperature than the host deposit, depending on its thermal behaviour during flow. However, it is therefore possible for an accessory lithic clast to be of type B if the deposit temperature was above the Curie temperature; however it could also be of type B, or C or D, if incorporated in a deposit below the Curie temperature depending on its thermal path (Fig. 4a); for types C and D of accessory clasts, the retrieved emplacement temperature may also be higher than the actual of the deposit, again depending on its thermal path.

Pensa et al. (2015a) reconstructed such complex and different thermal histories of individual lithic clasts at sites FO-04, FO-05 and FO-07. There, the concomitant occurrence of charcoal fragments allowed determination of a deposit temperature in agreement with the LT component of a sub-population of the lithic clasts sampled just around the charcoal. In contrast, the rest of the population was characterised by a single component, though parallel to the LT component. The authors then concluded that for the Fogo A ignimbrites, the lithic population is composed of a mixture of pre-heated accessory clasts and originally cold clasts (probably accidental). However, because their lithology is very similar, it is likely that different proportions of the two sub-populations are present at each site. It is impossible on a visual inspection to discriminate between the two origins, but the LT component parallel to the GAD provides an accurate measure of the emplacement temperature.

Based on the considerations above, we analyse pTRM results displayed in Table 1 and notice that only the black intra-Plinian ignimbrite has only clasts belonging to types A, B1 and B2, with a single component in all sampled sites. In comparison, both the pink intra-Plinian ignimbrite and the dark brown ignimbrite clasts from the same sites may show either a single component or double components, the latter being almost always less represented at each site. In particular, the five paleomagnetic behaviour typologies are present only in the pink intra-Plinian ignimbrite, where the double component is represented by both type C and type D groups, whereas in the dark brown ignimbrite the double magnetic component is only represented by type C group.

Based purely on statistics, the conclusion should be that the most reliable estimate is a temperature always higher than the unblocking temperature (variable across samples between 400° and 620 °C), as most of the analysed clasts show a single component. However, in light of the results presented in Pensa et al. (2015a), we here note that the LT component of type C (and type D) is parallel (or almost parallel) to the single magnetic component of clasts belonging to types B1 and B2 at each site (Table 1). Analysed clasts are all in the range of 1–3 cm in diameter and similar in lithology. Therefore, according to the experimental data of Cioni et al. (2004), the time required for thermal equilibration between the clast and the ignimbrite should not differ at all, being relatively rapid, in the order of minutes or tens of minutes, compared with the cooling rate of the deposit. We therefore interpret the presence, at the one sampling site, of clasts with either a single or double pTRM components as an effect of differing pre-heating histories of different clasts. According to this interpretation, we therefore suggest that at sites where all the clast types (B1, B2, C, D) are present, the LT component of those characterised by a double pTRM component (types C and D) provides the best estimate of the emplacement temperature of the ignimbrite.

According to Cioni et al. (2004), a continuous flowing mixture at 850 °C can reheat a thin layer of lithic clasts a few millimetres across after only a few seconds, reaching a temperature close to the temperature of the erupting mixture. In their studies, Cioni et al. (2004) stated that lithic clasts eroded from the conduit wall with a radius smaller than 1 cm reach the same temperature as the flow after 300 s. This time is considered to be the total time taken for the lithic clast to travel from the conduit to the eruption column (atmospheric path) and to the depositional location within the pyroclastic flow (Fig. 4). In this trajectory, the lithic clasts spend time passing through different thermal zones at decreasing temperatures (Fig 4). According to the numerical modelling of Neri et al. (2003) and Dartevelle et al. (2004), lithic clasts eroded from the conduit remain within the thermal domain characterised by very high temperature (similar to the magma temperature, that is, above the Curie T) not only within the conduit but also within the first 2 to 4 km of the eruption column. This provides the necessary time for the lithic clast to fully equilibrate with the surrounding gas-pyroclast mixture, before being incorporated into the pyroclastic flow and passing through the other lower thermal zones (Fig. 4).

From the pTRM data of the five paleomagnetic sites within the pink intra-Plinian ignimbrite (Table 1), we select the best quality LT component (type C) clasts to indicate the average emplacement temperature, providing an estimate between 300° and 400 °C. The SC clasts with a single component represent pre-heated clasts that have been eroded from the conduit wall.

The pTRM data of the lithic clasts collected within the black intra-Plinian ignimbrite indicate the presence of clasts with a single component (types A, B1, B2). The overwhelming dominance of type B1 clasts makes it more straightforward to interpret an average emplacement temperature above the unblocking temperature, i.e. between 580 and 620 °C. The very high temperature of the black intra-Plinian ignimbrite is also validated by the presence of incipient welding in proximal settings, as well as the absence of charcoal fragments, which instead characterise both the pink and the dark brown ignimbrites (Pensa et al. 2015a). We interpret this absence of charcoal to be due to the very high temperature of the deposit that would have led to the complete incineration or all woody material. At all sites sampled in the dark brown ignimbrite, results show the presence of clasts with either single component or double component. The analysis of the LT component of type C lithic clasts provides a coherent and relatively narrow temperature interval, between 250 and 400 °C in the north and between 250 and 370 °C in the south (Table 1). These data match the range of temperature suggested by Pensa et al. (2015a) using both pTRM and reflectance data.

Factors that affect the emplacement temperature of Fogo A ignimbrites

Our results show the presence of deposits with different dispersal patterns and emplacement temperatures. As discussed by Cioni et al. (2004) and Zanella et al. (2007; 2008 and 2014), the emplacement temperature of an ignimbrite is the sum of the contributions of different factors that operated during the eruption. The degree of mixing of air into the eruption column or pyroclastic density currents (Sparks and Wilson 1976; Bursik and Woods 1996; Lesti et al. 2011), the height of collapse from the eruption column (Wilson et al. 1980; Woods and Bursik 1991, Cioni et al. 2004; Shea et al. 2011), the degree of magma-water interaction (Koyaguchi and Woods 1996; Porreca et al. 2006, 2008), the degree of paleotopography confinement, local increase of turbulence due to roughness of the substratum (Gurioli et al. 2005, 2007; Zanella et al. 2007; Caricchi et al. 2014), the temperature of the magma at vent (Dobran 1992), the dispersal area, and the size, nature and abundance of the components (Martí et al. 1991) are all considered to be significant factors that affect the final emplacement temperature of pyroclastic flow deposits. Evaluating the role of each factor and which of these has determined the emplacement temperature of the deposit is not simple. In the case of the Fogo A eruption, all the listed factors may have played a role in generating ignimbrites at different temperatures.

Pink intra-Plinian ignimbrite versus black intra-Plinian ignimbrite

The two intra-Plinian ignimbrites have both been generated by partial collapses of the eruption column, separated by perhaps a few hours or less between each collapse, based on the continuity of the intervening fallout deposits. The two fallout layers (FGA2l6 and l7), interbedded between the pink and the black intra-Plinian ignimbrites, are well-sorted and do not show any grain-size gradation that could be related to a recovery of the eruption column following the collapse events. The presence of two fallout deposits indicates that the unsteadiness did not affect the entire eruption column, which remained buoyant and continuously produced fallout deposits (Fig. 6).

Fig. 6
figure 6

Reconstruction of the possible evolution of the Fogo A eruption: a initial phase with steady Plinian eruption column and emplacement of the trachytic fallout deposit (first member); b, c shift to unsteady eruption column and the generation of two partial collapses from different levels. The column is still present during the emplacement of the pink and black intra-Plinian ignimbrites; d total collapse of the eruption column and emplacement of the dark brown ignimbrite. Modified after Pensa et al. 2015b

Full size image

Compared with the white trachytic basal fallout deposit (FGA1l3–l5), with an average of 64.6 wt.% in silica content (trachyte), the pink and the black intra-Plinian ignimbrites display a slight change in composition towards less evolved compositions, characterised respectively by 61 to 59 wt.% in silica content (trachydacite).

The magma temperature at the initiation of the eruption has been estimated, using the ilmenite-magnetite geothermometer, to have been around 900 °C (Table 3). Considering the very similar composition of the trachytic and trachydacitic magma batches, we consider that magma temperature at the vent did not change during the eruption.

Table 3 Comparison of the characteristics of each ignimbrite of Fogo A eruption sequence and analysis of the possible causes of the eruption column collapse
Full size table

In spite of the minimal difference in composition, the colour of the juvenile clasts of the two intra-Plinian ignimbrites is remarkably different from the basal fallout pumice (FGA1l3–l4), which is white in colour, and to the two inter-stratified fallout deposits (FGA2l6–l7) in which the pumices are light and dark grey banded, though they have a similar composition to the intra-Plinian ignimbrites. The juvenile clast vesicularity also remains very similar, slightly increasing from 70 % in the basal trachytic fallout to 75–77 % in the intra-Plinian ignimbrites. This may possibly be due to post-fragmentation bubble expansion but suggests that degassing style did not change significantly throughout the eruption (Table 3).

The intra-Plinian ignimbrites are generally similar in facies characteristics, both being matrix supported, massive and poorly sorted, with no evidence of interaction with external water during the eruption (Fig. 5). Proportions of juvenile and lithic lapilli relative to ash content are similar (Fig. 2, Table 3). The pink intra-Plinian ignimbrite has a slightly higher content of lithic clasts (10 %) compared to the black intra-Plinian ignimbrite (5 %). The dispersal area of the two intra-Plinian ignimbrites is also similar; they are confined along the same paleo-valleys along the south flank of Fogo Volcano. In some areas, the pink intra-Plinian ignimbrite reached more distal areas (Mateus, Terra de Carvalho, Villa Franca do Campo north) compared to the black intra-Plinian ignimbrite, which occurs only in a more proximal area (Fig. 3).

Despite the similarities, the estimated emplacement temperature for each ignimbrite shows a substantial difference, 300–400 °C for the pink intra-Plinian ignimbrite and >580 °C for the black intra-Plinian ignimbrite (Table 3). This implies that, given similar conditions of topographic confinement, component content, areal distribution and original magma temperature, other factors influenced the emplacement temperature of the two intra-Plinian ignimbrites.

The maintenance of such a high temperature for the black ignimbrite was facilitated by the very small size (<2 cm) and low content in lithic clasts (5 %), and by the presence of dense crystal rich pumice clasts (5–10 %) with large diameter (>15 cm). According to Woods and Bursik (1991), Martí et al. (1991) and Cioni et al. (2004), the amount and the size of lithic clasts in a pyroclastic flow have a great influence on the temperature of the flow during transport and of the resulting deposit. The main reason is the thermal disequilibrium between the hot gas mixture and the volcanic constituents. The different grain sizes of the pyroclasts within the eruption column exert a crucial role on the dynamics of the eruption. Woods and Bursik (1991) showed that only clasts smaller than 3–4 mm in diameter can rapidly reach thermal equilibrium with the gas. These particles are transported by eddies within the column, towards the upper part of the umbrella region. All the coarser fragments, instead, are not in thermal equilibrium and are transported to their maximum height and pushed to the plume margins. An increase of grain size during the eruption, due to increase in discharge rate or to the enlargement of the volcanic conduit, implies increased thermal disequilibrium and the consequent instability of the eruption column (Woods and Bursik 1991; Thomas and Sparks 1992).

Martí et al. (1991) highlighted in particular the influence of accessory and accidental lithic clasts, within the hot gas mixture, on the dynamics of eruption columns and pyroclastic flows. According to their conduction model, populations of clasts with Fourier number (τ = kt/a 2 where k is the thermal diffusivity of lithic clast, t is the residence time and a is the clast radius) greater than 1, can reach thermal equilibrium within the deposit. This means that only clasts with a radius smaller than 1 cm can reach the same temperature within the flow before deposition (Cioni et al. 2004). Cold fragments with dimensions >1 cm need more time to reach the average temperature of the deposit.

They concluded that clasts with diameter less than 5 cm and constituting a volume fraction of 12 % can cause the cooling of the hot-gas mixture by 300 °C. Cioni et al. (2004) estimated that only after a few hours would large lithic clasts be fully thermally equilibrated with the surrounding deposit. During this time, cold coarse lithic clasts absorb heat from the gas-ash mixture while the juvenile clasts release heat within the deposit.

Given the relatively low lithic clast content of both intra-Plinian ignimbrites, their markedly different temperatures of emplacement indicate that lithic clast content played no role in this.

In the literature, the emplacement of very hot and confined pyroclastic flows was observed during the Mt. St. Helens eruption on the 22nd of July (Banks and Hoblitt 1996; Paterson et al. 2010). As reported in Banks and Hoblitt (1996), the direct temperature measurements of the ignimbrite indicated values >600 °C, which is similar to the black intra-Plinian ignimbrite of the Fogo A eruption sequence. In Cas and Wright (1988), the eruption of the 22nd of July of Mt St. Helens is described as an example of ignimbrite generated by instantaneous collapse from very low height (~500 m). According to Rowley et al. (1981), this type of fountaining ignimbrite is called “pot boiling over.”

In the case of Fogo A eruption, during the collapse that produced the black intra-Plinian ignimbrite, the eruption column was present and buoyant, so we cannot consider the black intra-Plinian ignimbrite a proper “boiling over” ignimbrite However, the very high emplacement temperature, the traces of welding, the paleotopographic confinement and the distribution to only proximal areas, indicate a low-level collapse from the eruption column. As already described, juvenile clasts of the black intra-Plinian ignimbrite display traces of welding. This indicates that juvenile clasts were emplaced very hot, slightly above the glass transition temperature, and did not have time to cool down from the moment of the eruption to the emplacement. This also suggests that the black intra-Plinian ignimbrite did not experience sudden or major cooling, which we interpret to be due to a low level of collapse from the eruption column. This then would have ensured minimal ingestion of cold air into the collapsing column.

By contrast, the juvenile clasts of the pink intra-Plinian ignimbrite do not display evidence of welding. The emplacement temperature estimated by pTRM is between 300 and 400 °C. This estimation indicates high emplacement temperature but not enough to exceed the glass transition temperature and to develop welding texture. The lower emplacement temperature of the pink intra-Plinian ignimbrite cannot be attributed to the cooling by the contact with a substrate paleosoil, as 4 m of basal fallout pumice fallout deposits underlie the ignimbrite. The lithic clast content (10 %), though twice that of the black intra-Plinian ignimbrite, does not support a temperature difference of 400–500 °C between the two ignimbrites.

The loss in temperature of the pink intra-Plinian ignimbrite can therefore be attributed to a higher level of collapse from the eruption column, which implies longer residence time of the juvenile clasts within the column, and consequently, a greater level of ingestion of cold air and cooling of the pyroclasts.

According to Shea et al. (2011, 2012), studies of the deposits of the A.D.79 eruption of Vesuvius show that the increase in the abundance of dense juvenile clasts and lithic blocks from the conduit wall are the trigger of partial collapses. As suggested by Wilson et al. (1980), an increase of eruption rate can lead to a decrease of ingestion of air and consequently can affect the mixing efficiency, causing the shift from a steady buoyant column to an unsteady, partially collapsing column. In the case of Vesuvius, the fluctuating magma discharge rate led to the emplacement of six pyroclastic density currents before the total collapse of the eruption column (Shea et al. 2011).

In the case of Fogo A eruption, the two partial collapses correspond to the moment of change in magma composition from trachytic to trachydacitic composition. Despite the silica content shift from the trachytic basal fallout composition to the trachydacitic intra-Plinian ignimbrites, the pyroclast vesicularity remained almost the same.

The general pumice size within the pink and the black intra-Plinian ignimbrite are noticeably large compared with the fallout deposits, and both ignimbrites contain dense juvenile clasts with vesicularity <50 %. There is no evidence of a change in the magma discharge rate during the compositional change, but the ejection of pyroclasts with larger diameter could cause thermal disequilibrium within the eruption column and its consequent instability (Woods and Bursik 1991).

The dark brown ignimbrite

The dark brown ignimbrite represents the final collapse of the eruption column and the conclusion of Fogo A eruption, and compositionally, it is the least evolved unit of the entire sequence. The juvenile content is composed of black and dark brown-banded pumices. The pTRM results indicate an average emplacement temperature between 250 and 370 °C, which is in agreement with the average emplacement temperature obtained with the comparison between charcoal reflectance analysis and pTRM data by Pensa et al. (2015a) (Tables 1 and 3).

This low emplacement temperature could be due to different factors: high levels of air ingestion during the final collapse due to the collapse of a very high column, interaction between magma and water, little paleotopographic confinement and high lithic clast content. To determine the main causes of a low emplacement temperature, we analysed the characteristics of the dark brown ignimbrite along the north and the south flank.

As already described, the dark brown ignimbrite is spread over the wide Ribeira Grande Graben in the northern sector of Fogo Volcano, whereas in the south, it is more paleo-topographically confined along a paleo-valley. However, by the time this ignimbrite was emplaced, much of the paleotopography in the south had been filled in by the earlier ignimbrites, and the dark brown ignimbrite also spread out over subdued topography to the sides of the paleo-valleys there.

Spreading of the pyroclastic flow over the flat wide Ribeira Grande Graben should have facilitated a fast heat loss due to the larger surface area of the flow and its buoyant ash cloud (Figs. 1 and 3 and Table 3).

The comparison between the pTRM analyses of the lithic clasts collected within the dark brown ignimbrite in the north and in the south revealed that, despite the different paleotopographic setting, the range of estimated emplacement temperatures is similar suggesting that neither valley confinement nor the greater thickness of the deposit contributed significantly to heat retention. As displayed in Table 1, the two paleomagnetic sites (FO-11 and FO-23) along the south flank indicate an emplacement temperature between 300 and 370 °C, whereas five paleomagnetic sites (FO-04, FO-05, FO-15, FO-16 and FO-18) along the north flank display an average range of temperature between 250 and 350 °C.

The similarity of the temperature data from both sides of the volcano, and the fact that in both sectors, there is no evidence of water-magma interaction (Fig. 5), suggests that the low emplacement temperature of the dark brown ignimbrite has to be due principally to the efficiency of the cooling by air entrainment during the eruption and the collapse, and to the high content (20 % by volume) of lithic clasts. The multiple lithic breccia lenses at different heights within the deposit and the big blocks (>1 m in diameter) at the top of the deposit suggests that part of the conduit walls collapsed and that the magma discharge rate increased during the eruption.

The enlargement of the vent or the conduit can lead to increased instability of the base of the eruption column (jet region) (Wilson et al. 1978, 1980; Sparks et al. 1997; Ogden et al. 2008). As suggested by Shea et al. (2011), the increase in the density of the eruption column due to the incorporation of dense lithic blocks and the consequent inability of the volatiles to transport these particles upwards within the column can lead to decrease in the eruption column height and to column collapse (Fig. 6). Given that there are no subsequent fallout deposits, the entire column collapsed at this final stage, including the relatively cooled upper part of the eruption column.

Heat dissipation during Fogo A eruption

The thermal history of the Fogo A eruption is complex. From the data collected in the field, coupled with the evidence from the pTRM and Ro% analysis in the laboratory, the three ignimbrites of Fogo A eruption seem to have experienced a different thermal evolution. From the data obtained using the Ti-magnetite-ilmenite geothermometer, the average temperature of the magma at the vent has been estimated at around 900 °C (average temperature from Powell and Powell 1977; Spencer and Lindsley 1981 and Andersen and Lindsley 1985; Tables 2 and 3). This result is in accord with the temperature estimations of trachytic magmas in the literature of ~900 °C (Mukherjee 1967).

The heat loss from the moment of eruption to the time of ignimbrite emplacement is different for each of the ignimbrite deposits. The heat loss, for the pink intra-Plinian ignimbrite, was substantial and is estimated to have caused a drop in temperature of between 400 and 500 °C. For the black intra-Plinian ignimbrite, the heat lost during the eruption and the emplacement caused a lower temperature drop of ~300 °C. The lower heat loss of this deposit is corroborated by the almost welded aspect of the juvenile clasts, which suggests that the pumice clasts were still hot enough to be plastically deformed during sedimentation, close to the glass transition temperature for trachyte (Tg peak 721 °C and Tg onset 658 °C with 0.03 wt.% water content and with cooling/heating rate of 20 K/min (see Giordano et al. 2005); higher magmatic water volatile content would lower the Tg.

The dark brown ignimbrite is the unit that experienced the greatest heat loss of the three Fogo A ignimbrites, involving a temperature drop of between 500 and 600 °C between eruption and deposition (Tables 2 and 3).

Considering the high sedimentation rate of the ignimbrites inferred from the massive, poorly sorted texture, the small dimensions of São Miguel Island (70 km × 15 km) and the limited dispersal area of the Fogo A ignimbrite deposits (4 to 5 km from the vent), the considerable heat dissipation from the pink intra-Plinian ignimbrite and especially the dark brown ignimbrite was very rapid.

This indicates that the high content in lithic clasts and big blocks incorporated from the crater walls contributed significantly to the heat loss during the emplacement of the climactic dark brown ignimbrite. Moreover, the collapse of the entire column, including the cooled upper parts, suggests that the great heat loss experienced by the dark brown ignimbrite was also attributable to the ingestion of cold air into the eruption column during the final collapsing phase.

Conclusions

The estimation of emplacement temperature of Fogo A ignimbrites performed with thermal remanent magnetization (pTRM) revealed a complex thermal history:

  • The detailed study of the thermal history of individual lithic clast allowed us to identify five different paleomagnetic clast behaviours (A, B1, B2, C and D). This result allows us to discriminate between accessory and accidental clasts and increases the accuracy and precision of emplacement temperature estimates for the three ignimbrites.

  • The two intra-Plinian ignimbrites were emplaced at a higher temperature than the final ignimbrite. The pink intra-Plinian ignimbrite had an emplacement temperature between 350 and 400 °C. The LT component of the population of lithic clasts with two magnetic components indicates that the depositional temperature did not exceed the Curie point. The black intra-Plinian ignimbrite was deposited at the highest temperature of 580 to 620 °C. This estimation is corroborated by the presence of only lithic clasts with a single magnetic component and incipient welding textures of juvenile clasts. This means that the Curie point has been exceeded. The final dark brown ignimbrite displays the lowest emplacement temperature amongst the ignimbrites of the Fogo A sequence and was deposited at an average temperature of 250−370 °C.

Analysis of different factors that could lead to different emplacement temperatures suggested that:

  • Despite showing the same topographic confinement along narrow valleys on the south flank of the volcano and similar areal dispersal, the two intra-Plinian ignimbrites experienced dissimilar thermal histories. This is attributed to the high content of juvenile clasts and low content of lithic clasts in the black intra-Plinian ignimbrite compared with the pink intra-Plinian ignimbrite and the different areal distribution of the two deposits (distant from vent for the non-welded, pink ignimbrite and proximal for the partially welded black ignimbrite). Moreover, different collapse heights from the eruption column and consequently different amounts of air ingested into the column must be considered in the evaluation of the emplacement temperature (high collapse height for the pink ignimbrite, low collapse height for the black ignimbrite).

  • The juvenile particles of the final, dark brown ignimbrite experienced the greatest decrease in temperature. The comparison of the pTRM data of deposits on the south and north flanks of the volcano revealed the same emplacement temperature. This means that the low emplacement temperature is not due to the different topographic confinement conditions but rather to the higher abundance of lithic breccia lenses and big blocks not in thermal equilibrium with the ash-gas mixture. This played a significant role in the instability and collapse of the eruption column and in absorbing much of the heat in the eruption column and pyroclastic flow, thus influencing the emplacement temperature of the ignimbrite. Moreover, a high collapse level for the dark brown ignimbrite would have facilitated more mixing of air into the column.

  • The estimation of the magma temperature at the moment of the eruption obtained with the magnetite-ilmenite geothermometer revealed an original pre-eruption temperature around 900 °C. During the eruption the heat lost resulted in significant temperature reduction of between 400 °C and 600 °C.