Throughout its geologic history, the island of Pantelleria (Figure 1), the type locality of peralkaline rhyolitic (pantelleritic) magmatism, has been the setting for dominantly explosive volcanism. The relatively low viscosity of these magmas make the pyroclastic deposits particularly prone to rheomorphism and welding, which obscures primary textural and architectural features. This peculiarity, coupled with discontinuous field exposures and a remarkably complex evolutionary history (viz., overlapping explosive events), has contributed to the difficulty in the advancement of our volcano-stratigraphic knowledge of the island. For these reasons, Pantelleria has been paradigmatic in the establishment of integrated field and geochemical techniques in unraveling the volcanic evolution of such active centers.
In order to appreciate fully the evolution of the volcanic stratigraphy at Pantelleria, we briefly review the history of geologic studies on this island that began with Gemmellaro , who reported the results from his field surveys that he integrated with reports from contemporary naturalists. More than fifty years elapsed before Förstner  published the next study of the island, which included a generalized geologic map (1:100.000) and, most significantly, a few chemical analyses of minerals and of a very peculiar and previously undescribed alkali- and silica-rich effusive rock, for which he proposed the name pantellerite. In October 1891, a submarine basaltic eruption 5 km west-northwest offshore of Pantelleria town attracted the attention of naturalist and astronomer Riccò , who precisely described the evolution of the eruption, including exploding scoria at sea now known as lava balloons. Significantly, he paid great attention to pre- and syn-eruptive phenomena, such as earthquake swarms, bradyseisms, and other macroseismic effects. These early studies [including Bergeat 1907] sparked the interest of the American petrographer H. S. Washington, who conducted a field campaign in 1905 and subsequently published three papers [Washington 1913a,b, 1914] that are now considered the foundation for all later studies, not only for the detailed geological descriptions, but especially for the in-depth discussion of the petrography of pantellerite rocks that much improved the early analytical efforts of Förstner.
Comparative nomenclature of Pantelleria ignimbrites and their ages
|Jordan et al. ||40Ar∕39Ar age||Rotolo et al. ||40Ar∕39Ar age (ka)||Speranza et al. ||Mahood and Hildreth ||K/Ar age (ka)||Wright |
|Green Tuff||46(∗)||Green Tuff||45–50||B|
|Mordomo||D = Z||85||D = Z||Z||78–84||C, D, E|
|Acqua||F = Q||107||F||F||101–110||F, G|
|Cinque Denti||128||P||123||P||P||124–133||G, g, U|
|(locally = Q)|
|Capre||138||Welded breccia (Br 1)||140–146||Welded breccia||Breccia||104–127|
There was renewed interest in the geology of Pantelleria during the 1960s, with most studies focused on the volcanology of the island. Borsi et al.  published a general overview, and Rittmann  provided the first detailed study of the volcanic stratigraphy in which he distinguished several ignimbrite and pumice fallout deposits, recognized the central caldera structure, and produced a detailed geological map (1:25.000) of the island. His conclusions were confirmed and further developed in several publications by his collaborator Villari [1969, 1970, 1974], who proposed a basic stratigraphy that recognized several ignimbrite units. In particular, Villari focused on the “Green ignimbrite”, the youngest and most widespread pyroclastic unit on the island, whose ignimbritic nature was confirmed by Schmincke . Wright  was the first to attempt to unravel the complex stratigraphy and lateral correlations of the older pyroclastic units, some of which he interpreted as welded air fall deposits instead of ignimbrites. Geochronological (K–Ar) data were first presented by Barberi et al.  and Bigazzi et al.  and these were later followed by Cornette et al.  and Civetta et al. [1984, 1988]. Cornette et al.  recognized two partially overlapping caldera structures that they termed the Lago Caldera and Monastero Caldera (50 ka), the latter of which they associated with the eruption of the Green Tuff (GT). Contemporaneously, Mahood and Hildreth  also described two “nested” calderas on the island, which they termed the La Vecchia Caldera (93 ka) and Cinque Denti Caldera (55 ka), which differed from those proposed by Cornette et al.  in terms of location, size, and significance. The remarkably thorough and insightful paper of Mahood and Hildreth  was ahead of its time and provided the first (and, until now, only) comprehensive and integrated description of the structure and volcanological history of the island. Mahood and Hildreth also put a major focus on the stratigraphy of the pre-GT welded and rheomorphic ignimbrites, supported by a large number of K–Ar ages.
Most of what was published over the next two decades focused on specific petrological and volcanological aspects of the geology of Pantelleria [Avanzinelli et al. 2004; Behncke et al. 2006; Bonaccorso and Mattia 2000; Civetta et al. 1998; De Guidi and Monaco 2009; Esperança and Crisci 1995; Ferla and Meli 2006; Fulignati et al. 1997; Kovalenko et al. 1994; Lowenstern 1994; Lowenstern and Mahood 1991; Mattia et al. 2007; Perugini et al. 2002; Prosperini et al. 2000; Stevenson and Wilson 1997; Wallmann et al. 1988; White et al. 2005, 2009], with a sharp increase in studies produced over the past ten years [Arzilli et al. 2020; Avanzinelli et al. 2014; Baginski et al. 2018; Campagnola et al. 2016; Conte et al. 2014; Di Carlo et al. 2010; Di Genova et al. 2013; Fouré et al. 2012; Gioncada and Landi 2010; Giuffrida et al. 2020; Jordan et al. 2021; Hughes et al. 2017; Kelly et al. 2014; Lanzo et al. 2013; Liszewska et al. 2018; Neave 2020; Neave et al. 2012; Richard 2015; Romano et al. 2018, 2019, 2020; Romengo et al. 2012; Rotolo et al. 2013, 2017; Scaillet et al. 2011, 2013; Speranza et al. 2010, 2012; White et al. 2020]. Notably missing until 2007 were stratigraphic and geochronological studies, implicitly suggesting to some extent that the great effort made during the 1980s was still considered valid and that methodological improvements in geochronological methods had yet to come. The first 40Ar∕39Ar study of a pantelleritic enclave in an ignimbrite produced the oldest age reported on Pantelleria [517 ± 19 ka, Rotolo and Villa 2001], which defined a lower limit for the onset of pantellerite magma production and sparked new interest in the geochronology and volcanic stratigraphy of the island [Jordan et al. 2018; Rotolo et al. 2007, 2013; Scaillet et al. 2011, 2013; Speranza et al. 2010, 2012].
The principal aim of this paper is to review and summarize the existing volcano-stratigraphic knowledge of Pantelleria, which is otherwise scattered between pre- and post-GT papers. In doing so, we try to highlight how methodological improvements have had a decisive impact on developing a fully integrated stratigraphy and time-integrated evolutionary history of the system, portraying a volcanological scenario that has far-reaching implications for, but not strictly limited to, other peralkaline volcanoes.
2. The pre-GT volcanological evolution
Rittmann  published the first pre-GT stratigraphy based on eight measured sections and distinguished several ignimbrite and fall units, which he labeled A (younger) to G (older). Villari , on the basis of field evidence and petrographic analyses, recognized several ignimbrite sheets (uncorrelated between different sections), identified a single caldera and distinguished the units as either pre- or post-caldera ignimbrites. The field-based stratigraphic study of Wright , correlated seven welded tuffs across the island (without defining their type localities), plus eight other uncorrelated (minor) members. These were named using non-sequential capital letters, a rather confusing and non-intuitive nomenclature that would continue to be used until 2018. From older to younger these were: I, H, G, F, E, D, C, and B, with the youngest one corresponding to the GT. On the basis of 37 K/Ar ages coupled with extensive field work, Mahood and Hildreth  described the first detailed stratigraphy of the pre-GT eruptive cycles and units, and recognized nine informal units of welded tuffs/ignimbrites, most of which blanket the entire island. Although their stratigraphy was substantially different in many critical ways from Wright’s  (Table 1), they maintained and adapted his convention for naming the units (from older to younger : I, H, M, S, P, Q, F, D, Z and GT), but also did not define any type sections. They did, however, attribute a lithic-rich welded pyroclastic unit visible in the vertical scarp at Cala delle Capre as the caldera-forming unit of the La Vecchia caldera (Figure 1).
Many years later, Speranza et al.  applied paleomagnetic methods to the pre-GT ignimbrites and simplified Mahood and Hildreth’s  stratigraphy by merging some ignimbrite units (D and Z) previously considered distinct. They also correlated two pyroclastic breccia units on the opposite sides of the island (NE: Cala Cinque Denti, SW: Cala delle Capre, Figure 1) to the La Vecchia caldera collapse and constrained it to 160–130 ka. These new interpretations were integrated with a new geochronological (40Ar∕39Ar laser-ablation) study of these units [Rotolo et al. 2013] with the following results: (i) the age of the La Vecchia caldera collapse was more tightly constrained to 146–140 ka by dating the juvenile material in the same pyroclastic breccias studied by Speranza et al. ; (ii) the age of eruption of other five ignimbrites was substantially refined as follows [following Mahood and Hildreth 1986]: M = 181 ± 1.2 ka, S = 171 ± 1.7 ka, P = 123 ± 1.6 ka, F = 107 ± 1.4 ka, Z = 85 ± 1.5 ka; and (iii) the conclusion by Speranza et al.  on the correlation of the D and Z units was confirmed and dated at 85 ka and units P and Q were also found to be correlative.
Comparison of ages determined with different methods for four key post-GT eruptions
|Gallo||7.09 ± 0.8||5.9–6.2||3.03 ± 0.3||5.7 ± 0.1|
|Randazzo fall||8.2 ± 1.7||5.42 ± 0.22|
|5.75 ± 0.08|
|Khaggiar lava flow||8.0 ± 0.8||5.9–6.2||5.5 ± 5.0|
|8.5 ± 4.5|
|11.0 ± 3.0|
|Fastuca||9.70 ± 0.6||6.2–6.8||6.0–5.8 ± 0.03||6.1 ± 0.1|
It was only after a combined field, 40Ar∕39Ar, and petrographic study [Jordan 2014], later integrated later with paleomagnetic and additional 40Ar∕39Ar data [Jordan et al. 2018], that each of these ignimbrite units (and related pumice fallout, when present) were finally tied to a clearly defined type locality and renamed accordingly (Table 1). This was a major step forward over the poorly defined and counter-intuitive letter-based scheme. The resulting stratigraphic reconstruction clarified the structural and volcanic dynamics of the Pantelleria edifice by better delineating paroxysmal events and their recurrence through time; they also emphasized the occurrence of an indefinite number of active local centers producing lower explosivity eruptions between each ignimbrite [inter-ignimbrite activity, Jordan 2014; Jordan et al. 2018]. The erupted (onshore) volumes of these ignimbrites, although limited by the impossibility of knowing the amount of tephra deposited at sea, varied between 0.15 to 0.64 km3 (D.R.E.), the largest of these belonging to the Polacca Fm. (187 ± 2 ka, Figure 2a) [Jordan et al. 2018]. An important observation of Jordan  and Jordan et al.  is the occurrence of lithic breccias in five different ignimbrites, strongly suggesting that at least five caldera collapses occurred, although their morpho-structural remnants are now totally buried with the exception of the two clearly visible, though discontinuous, La Vecchia and Cinque Denti caldera scarps.
3. The Green Tuff
The GT, the ninth and the youngest of the known and exposed Pantelleria ignimbrites, is undoubtedly the most studied eruptive unit at Pantelleria. Although the age of the GT has long been correlated with the peralkaline distal ash Y-6 [Keller et al. 1978; Anastasakis and Pe-Piper 2006; Margari et al. 2007], its age was only loosely constrained by a few low-resolution K/Ar dates varying between 46.9 ± 2.0 ka and 50.8 ± 3.6 ka [Cornette et al. 1983; Civetta et al. 1988] and 45 ± 4 ka to 50 ± 4 ka [Mahood and Hildreth 1986]. It was only recently that this age was refined to a higher resolution estimate at 45.7 ± 1.0 ka by 40Ar∕39Ar laser-ablation dating [Scaillet et al. 2013].
The emplacement dynamics of the GT have been variably interpreted through time. Once viewed as (i) a lava flow [Washington 1914] or (ii) a welded fall [Mahood 1984; Wolff and Wright 1981a,b; Wright 1980], it later became clear that it involved more high-energy dynamics, consisting of either (iii) a compound ignimbrite [Mahood and Hildreth 1983, 1986; Orsi and Sheridan 1984; Villari 1970], (iv) a diluted ash-flow [Schmincke 1974], and/or (v) a welded low aspect-ratio ignimbrite [Williams 2010; Williams et al. 2014]. The frequent and variable-scale rheomorphic folding was interpreted either as post-depositional [Mahood and Hildreth 1986; Villari 1970; Wolff and Wright 1981a,b] or syn-depositional (with a diachronous emplacement of the pyroclastic currents during essentially three eruptive phases, [Catalano et al. 2007, 2014]). The GT was correlated to the younger caldera collapse, whose remnants are recognized in the Monastero scarp [“Monastero caldera” of Cornette et al. 1983] and the scarp at Cala Cinque Denti [“Cinque Denti caldera”, of Mahood and Hildreth 1986].
The innovative study by Williams  focused on high-resolution chemostratigraphy of multiple sections of the GT. The whole ignimbrite deposit, preceded by a basal unwelded pumice fallout member that crops out in only a few places, was divided into eight time slices [“entrachrons” of Branney and Kokelaar 2002], each representative of an eruptive timing of a few minutes. Time slices were chemically correlated across the island, providing diachronous snapshots of the areas progressively inundated by the pyroclastic density current (PDC), each with different runout distances influenced by the pre-existing topography: some pulses by-passed topographic barriers whereas others were instead partially or totally halted by the terrain, during the brief time interval (⩽1.5 h) during which the current was emplaced [Williams et al. 2014]. Williams  questioned whether the GT was the actual eruption that created the Cinque Denti caldera, citing a lack of clear evidence for large-scale caldera collapse during the GT eruption and suggesting that the Cinque Denti scarp could, in many places, be older. The only evidence found for syn-eruptive collapse is a small-volume pyroclastic breccia within the GT section at Monastero, which is proposed as the type locality for the GT ignimbrite (except for its basal fallout member, which was defined at the Khattibucale scarp; [Williams 2010]).
The (onland) volume of the GT has been recently re-evaluated at 0.28 km3 (D.R.E.) [Jordan et al. 2018], much lower than previous estimates of 0.49 km3 and 0.60 km3 D.R.E. [Mahood and Hildreth 1986; Wolff and Wright 1981b, respectively].
4. The post-Green Tuff volcanological evolution
Volcanism following the GT was distributed throughout a large number (>40) of different centers closely spaced in time and space and characterized by mildly explosive and/or effusive activity. Mafic local centers are largely confined to the northwest of the island and are not found within the caldera, felsic centers are commonly found on or near caldera faults [Jordan 2014].
The first post-caldera event was the eruption of Monte Gibele–Montagna Grande trachyte lava (ca. 3 km3 according to [Mahood and Hildreth 1986]) in the center of the young caldera. Cornette et al.  and Mahood and Hildreth  both supported Washington’s view regarding the uplift of the M.gna Grande trachyte lava block [first proposed by Förstner 1881], with these latter authors providing detailed descriptions of the bordering faults, the hinge, and the offset (“trapdoor uplift”) that caused the displacement of the M. Gibele source vent. Cornette et al.  and Civetta et al. , provided the first K/Ar ages (nine in total) for the reconstruction of the recent eruptive history. These were followed by twelve more K/Ar and six 14C ages by Mahood and Hildreth , who proposed a general stratigraphic scheme framed as a “morpho-structural” subdivision of eruptive centers based on their position between or along the caldera rim and radial intra-/extra-caldera faults. Civetta et al.  added 39 new K/Ar ages on lavas and tephra, and based on field evidence (e.g., paleosols) proposed a more precise interpretation involving six eruptive cycles, the first being the eruption of the GT and the second the eruption of trachyte from the M. Gibele vent. Several vents were active during the most recent (sixth) cycle, such as the composite Cuddia Randazzo center (K/Ar age, 8.2 ± 1.7 ka), which consists of a pumice ring surrounding a coeval pantellerite lava dome (Figure 2b), whose rupture produced the Khaggiar lava field, which reaches the sea at Punta Spadillo, 2.5 km away from the source vent. The most energetic (strombolian/sub-Plinian) event was the Fastuca eruption, whose source vent is located on the northern slope of M. Grande [Orsi et al. 1991; Rotolo et al. 2007].
The poor age resolution of the K/Ar method (with ±2–6 ka 2𝜎 errors) was not precise enough to allow discrimination between young eruptive units in order to place tighter constraints on the youngest eruptions. To distinguish these, Speranza et al.  performed the first application of paleomagnetic methods to three key eruptions. Their youngest documented deposit is the Cuddia Gallo agglutinate (Figure 2c), which they bracketed between 5.9 and 6.2 ka, contemporaneous with the Khaggiar lava flow. For comparison, previous K/Ar ages yielded 8.2 ± 1.7 ka (Civetta et al. , Cuddia Randazzo pumice, cogenetic with Khaggiar lava), whereas 14C gave 5.4 ± 0.2 ka, on an early Khaggiar lava flow [Mahood and Hildreth 1986] (Figure 2b, Table 2).
Scaillet et al.  conducted the first 40Ar∕39Ar high-resolution laser-ablation dating on anorthoclase phenocrysts from lavas and pumices erupted during the last 20 ka. Due to the very young ages (7–15 ka) and rather K-poor feldspar composition resulting in low 40Ar∗ yield, this proved a more efficient approach at resolving the occurrence of 40Arxs and/or contamination with feldspar xenocrysts that plagued earlier K/Ar attempts (conducted on large-size aliquots). Scaillet et al.  were successful in obtaining new and more precise ages for the recent intra-caldera activity. In particular, the Fastuca eruption was dated at 9.7 ± 0.6 ka. Scaillet et al.  concluded that the eruptive pace is on a long-term wane and coupled this to an overall (though minor) tendency in decreasing differentiation of pantellerite magmas. Further, these new and better-resolved 40Ar∕39Ar data and field observations [see discussion in Scaillet et al. 2011] do not support the five cycles subdivision of post-GT activity proposed by Civetta et al. .
Table 2 summarizes the ages of the three youngest eruptions at Pantelleria obtained by different methods: the 40Ar∕39Ar ages are seen to be systematically older than paleomagnetically derived paleosecular variation-tied estimates, which are in turn comparable or slightly older than 14C ages. K/Ar ages appear rather dispersed with relatively high uncertainties. The bias between 40Ar∕39Ar and 14C estimates is well known [Civetta et al. 1998; Mahood and Hildreth 1986] and was interpreted by Scaillet et al.  as intrinsic to the 14C method, possibly due to inefficient insulation of the tephra/soil horizon hosting the charcoal and consequent contamination with modern carbon.
5.1. Methodological improvements: application of paleomagnetism to assess correlation of rheomorphic ignimbrite scattered outcrops
Paleomagnetism addresses the direction and intensity of magnetization measured in volcanic rocks that in turn may reflect the local characteristics of the geomagnetic field (direction and intensity) at the time of eruption and cooling [e.g., Butler 1992]. A volcanic rock acquires its magnetization when it cools below the Curie temperature (Tc) of its dominant ferromagnetic mineral (typically magnetite, with a Tc of 575 °C) to the ambient temperature. After the volcanic rock has cooled, its magnetization is “frozen” and does not change even when the characteristics of the geomagnetic field change. Direct observations during the last five centuries have shown that both direction and intensity of the geomagnetic field change quite rapidly (up to 6° direction change per century in Europe, e.g., [Lanza et al. 2005]), a phenomenon known as secular variation (SV) of the geomagnetic field.
The fast temporal change of paleosecular variation (PSV) of the geomagnetic field along with routinely achieved accuracy of paleomagnetic direction determination in volcanics (directions defined with confidence cones of 2°–4° of radius) imply that two volcanics emplaced only 100–200 years apart can be paleomagnetically discriminated. Such high resolution of paleomagnetic age correlation can fully complement radiometric dating: the latter yields absolute ages, while the former—with a dating resolution of about a century—may definitely establish whether two scattered volcanic outcrops belong to the same eruption or not. The best resolution of K/Ar dating is on the order of few millennia, while for laser-based 40Ar∕39Ar techniques can resolve down to 1 ka or better (up to a century for sanidine, [Renne et al. 1997]). For lower-K anorthoclase crystals it is close to 0.5 ka [Scaillet et al. 2011] and PSV-based paleomagnetic dating can provide valuable temporal and cross-correlative constraints to complement existing 40Ar∕39Ar ages.
Paleomagnetism had been previously used to assess whether individual lava flows belonged to the same lava field [Bogue and Coe 1981; Coe et al. 2005; Hagstrum and Champion 1994; Speranza et al. 2008], whether different welded scoriae were produced by the same eruption [Zanella et al. 2001], and—concerning ignimbrite correlation only in the pioneer works by Grommé et al.  and Ort et al.  on the mid-Tertiary ignimbrites from the western US and the Campanian ignimbrite from the Phlegraean Fields, respectively. Pantelleria seemed to be an ideal target to use the paleomagnetic correlation method, as the pre-GT ignimbrites were exposed only at isolated sea coves, with many difficulties in their correlation, given also the rheomorphism and the frequent in-depth lateral lithofacies variations. K/Ar dating by Mahood and Hildreth  had not solved the question, as age error bars of ignimbrite units Z, D, F, and Q overlapped. Speranza et al.  collected 23 paleomagnetic sites from the aforementioned ignimbrites plus ignimbrite P and the so-called “welded lithic breccia” considered by Mahood and Hildreth  to relate to the La Vecchia caldera collapse.
The paleomagnetic analysis of Pantelleria ignimbrites was successful due to their highly welded character allowing easy coring of compact yet relatively soft ignimbrite matrix. Also, scatter in paleomagnetic direction (Figure 3) proved to be smaller than in lavas, probably due to: (1) the lack of post-emplacement tilt in ignimbrites, compared to lava field sectors with continuous lava supply over weeks or months that can push and tilt already solidified lava blocks; (2) much smaller magnetization intensity of ignimbrites with respect to lavas, implying much smaller magnetic anomalies generated by buried deposits or the volcanic unit itself; local magnetic anomalies in fact represent one of the most significant scatter source of paleomagnetic directions from volcanics [Baag et al. 1995; Speranza et al. 2006].
The main results reached by the paleomagnetic study of the pre-GT ignimbrites by Speranza et al.  are as follows:
- Ignimbrites D and Z (now called “Mordomo” after [Jordan et al. 2018]) are both characterized by high 20°–30° declination and 50°–60° inclination values suggesting a common emplacement. This result turned out to be fully consistent with the new 40Ar∕39Ar determinations by Rotolo et al. . Considered together these data indicate a pooled age of 84.7 ± 0.5 ka (2𝜎, MSWD = 1.15) for this depositional unit.
- A lithic-rich ignimbrite exposed below ignimbrite P (now called “Cinque Denti”) at Cala Cinque Denti (NE Pantelleria coast) paleomagnetically correlates with the “welded lithic breccia” of Cala delle Capre [SW Pantelleria, Mahood and Hildreth 1986], demonstrating that the La Vecchia caldera-forming eruption occurred at 130–160 ka, based on available K/Ar ages by Mahood and Hildreth . Such age window was later fully confirmed, and substantially refined at 140.0 ± 1.5 ka by 40Ar∕39Ar dating in Rotolo et al.  for the now called “Capre” ignimbrite, later 40Ar∕39Ar dated at a similar age of 137.9 ± 0.8 ka by Jordan et al. .
- The paleomagnetic directions of F and Q ignimbrites share similar directional data (within <10°), but Speranza et al.  did not realize that they could represent the same event. Relying on 40Ar∕39Ar dating, the F/Q ignimbrite correspondence was first proposed by Rotolo et al.  and integrated as such in the latest volcanic stratigraphy by Jordan et al. , who called it “Acqua” ignimbrite.
To sum up, the integration of paleomagnetism, 40Ar∕39Ar radiometric dating and stratigraphic analysis allowed solving the timing of pre-GT ignimbrite emplacement with a resolution that would have never been attainable using only one or two of those methods. Paleomagnetism of welded ignimbrites proved to yield very reliable results, more accurate than those obtained by lavas.
The work by Speranza et al. , along with pioneer papers by Grommé et al.  and Ort et al. , and more recent contributions by Ort et al. , Finn et al.  and Kirscher et al.  on the Bolivia, Idaho, and Armenia ignimbrites ,respectively, prove that paleomagnetism represents probably the best tool to assess the correlation of ignimbrite outcrops and proximal versus distal volcanic deposits. We expect that in future, paleomagnetism will be highly used on worldwide welded ignimbrites to solve similar volcanological problems, in combination with high-resolution 40Ar∕39Ar.
5.2. Methodological improvements: 39Ar∕40Ar dating of young and K-poor feldspar
Although PSV-tied directional data can provide an efficient high-resolution correlation tool in a tightly Earth-referenced coordinate frame, 40Ar∕39Ar remains essential in terms of absolute temporal control. This is especially so for progressively older units both because a master PSV curve has not been established worldwide, and none locally extends very far back in time. 40Ar∕39Ar ages are also insensitive to tectonic unrest and post-cooling tilting characterizing explosive volcanoes (sector collapse, caldera resurgence/foundering, bradyseism, etc.). Methodological progress in the past two decades has considerably improved the resolution of the 40Ar∕39Ar technique in terms of either target (lava flows, fall units, ignimbrite flows) and material processing (single vs. bulk crystal dating, single vs. multiple ion collection). As stated earlier, laser-based 40Ar∕39Ar techniques have reached a resolution narrowing to a century for sanidine [Renne et al. 1997]. Such a performance came about as the result of greatly improved blank levels permitted by laser-based extraction systems compared to older furnace setups. In such systems, the effective 40Ar∕39Ar resolution scales linearly with the K content of the sample such that a fivefold reduction in K results in a fivefold reduction in absolute precision [Scaillet 2000], practically affording a limiting precision five times bigger (±0.5 ka, 2𝜎) for lower-K anorthoclase crystals typical of peralkaline rhyolites [Scaillet et al. 2011].
Detailed 40Ar∕39Ar work at Pantelleria have shown that, beyond precision the major obstacle to building a tightly resolved chronostratigraphy is the occurrence of mixed 40Ar∕39Ar systematics characterized by internally discordant ages scattering beyond individual analytical errors. Except for the GT [Scaillet et al. 2013] and several lava flows and small-sized Strombolian eruptions [Scaillet et al. 2011, 2013; Rotolo et al. 2013], about a half of all units dated so far have proved to be affected by such systematics. Scaillet et al.  devised a dedicated two-step protocol combining multi-grain fusion (about 10–15 crystal at a time) with pre-degassing in vacuo to improve both the radiogenic yield and the removal of surface-bound volatiles potentially carrying excess 40Ar. By screening out low-T steps featuring anomalously old ages, this proved essential in producing the first high-resolution 40Ar∕39Ar ages ever produced on such youthful mildly K-enriched material. Systematic application of this approach to selected post-GT units showed this excess component to be dominantly derived from late-stage interaction with atmospheric or hydrothermal agent (i.e., secondary, non-magmatic, component).
A different picture emerged for the older, more explosive, ignimbrite deposits constituting the bulk of the volcano infrastructure prior to the deposition of the GT. Unlike the GT, which displayed well-behaved 40Ar∕39Ar data permitting an age of 45.7 ± 1.0 ka to be confidently resolved [Scaillet et al. 2013], 7 out of 13 pre-GT ignimbrite samples showed discordant 40Ar∕39Ar patterns with internal age variations as great as 400 ka (but more commonly <40 ka). Notably, such internal variations arise despite the application of the two-step protocol, pointing to the primary (syn-magmatic or syn-depositional) origin of the anomalous ages [Rotolo et al. 2013]. An age in excess of the depositional time may arise due to excess 40Ar contamination in the magma reservoir proper (i.e., via Ar dissolved in the melt, cf. Esser et al. ), or as a result of syn-eruptive incorporation of xenocrysts from older deposits, or entrained by the magma en route to the surface.
A characteristic feature of within-unit 40Ar∕39Ar variations is that they cover a relatively reproducible time span, from about zero to ∼50 kyr. When plotted as a probability density of the age in excess of the depositional age (Figure 4), such systematics reveal striking differences between pre-GT and post-GT units. While both distributions peak near the origin (i.e., no excess 40Ar), post-GT 40Ar∕39Ar ages are more sharply peaked and quite rapidly fade away past 5 kyr of the depositional age (Figure 4c). In contrast, Pre-GT units display a broader distribution with a tail extending well beyond the depositional age, up to 50 kyr and older (Figure 4d). A characteristic break in slope also occurs in the quantile distribution of the sorted post-GT data (Figure 4a) that is not apparent in the pre-GT data distribution which is more rounded and never achieves a linear trend across the data populating the peak (Figure 4b). This statistical difference hints at an intrinsically different mechanism of xenocrystic incorporation. While incorporation of reworked epiclastic/subvolcanic products is clearly accidental in both cases, it is more extensive in the pre-GT ignimbrites units than in the lower-energy post-GT fall deposits. In the latter case, syn-eruptive re-incorporation of older material can be understood to occur at shallower levels than in the higher-energy, highly disruptive, caldera-forming eruptions.
In this connection, it is remarkable that half of the pre-GT ignimbrites affected by xenocrysts locally display homogeneous (xenocryst-free) systematics [Rotolo et al. 2013; Jordan et al. 2018]. The coexistence in a single depositional unit of lateral variations in xenocryst contamination, with the local absence thereof, indicates that epiclastic mixing dynamics or within-reservoir melt/rock interactions were locally controlled. Should the latter apply, the local control on xenocrystic incorporation would imply sector zoning in the reservoir and, possibly, piecemeal caldera collapse to preserve such a zoning during deposition. With continuous progress in single-grain 40Ar∕39Ar resolution [Jordan et al. 2018] and data productivity, Pantelleria clearly will provide excellent opportunities to refine such time-spatial eruptive dynamics based on systematic 40Ar∕39Ar dating of anorthoclase.
6. Physical volcanology of a peralkaline volcano
6.1. Local eruption centers versus ignimbrite-forming eruptions
As a defining locality for peralkaline volcanism, it is worth considering the general eruptive styles of Pantelleria eruptions. Broadly, there are ignimbrite eruptions and local (monogenetic) centers, as portrayed in Figure 5 in a scheme that integrates the most recent geochronological, paleomagnetic, and field data for pre- and post-GT eruptive units [Jordan et al. 2018; Rotolo et al. 2007, 2013; Scaillet et al. 2011, 2013; Speranza et al. 2010, 2012]. Ignimbrite eruptions are large eruptions from the caldera. Local centers can be effusive or explosive and have occurred throughout the entire ⩾324 ka subaerial volcanic history [Jordan et al. 2018; Scaillet et al. 2011 and references therein]. They can produce small ignimbrites, pumice falls and lavas (sometimes spatter fed) and typically show transitions between explosive and effusive activity [Stevenson and Wilson 1997]. They may be cone- or shield-shaped with a limited dispersal, which in most cases does not exceed several hundred meters (⩾3 km for the Fastuca eruption, [Rotolo et al. 2007]) so they are best termed strombolian. Eruption style, magnitude, and periodicity are similar whether pre-GT or post-GT [Jordan et al. 2018].
Whilst the GT is an important marker horizon in that it is widespread and easily recognizable, it should not be used to mark any particular change in eruptive style in Pantelleria’s eruptive history. Pumice falls associated with the ignimbrite-forming eruptions are rare; only the GT, Cinque Denti, and Arco Formations have associated pumice falls and these are relatively thin and not widespread. Eruption columns were sub-Plinian at most [Williams 2010; Jordan 2014]. Other peralkaline volcanoes do not produce widespread Plinian precursor pumice falls either [e.g., Terceira: Self 1976; Kenya: Leat 1991; Gran Canaria: Schmincke and Sumita 1998]. This may be because of low magma viscosity and the fact that peralkaline eruptions do not form tall stable eruption columns [Mahood 1984]. This is a key difference between peralkaline and metaluminous eruptions.
6.2. Caldera-forming eruptions
Of eight ignimbrites pre-GT, five comprise of variably widespread significant lithic breccias (GT, Mordomo, Acqua, Arco, and Polacca Formations). Of these, the Acqua breccia is the thickest at ∼5 m. Lithic breccias are interpreted to represent an energetic phase of the eruption coupled with vent erosion or widening and are therefore commonly thought to indicate caldera collapse phases of an ignimbrite eruption [Walker 1985; Druitt and Bacon 1986]. Only the lithic breccias in the Mordomo Formation and in the Acqua Formation contain plutonic clasts, evidencing excavation from the vent or magma reservoir wall rocks and therefore are the two most likely candidates for caldera-forming eruptions.
There is no compelling evidence for single climactic collapse phases during the GT, Arco or Polacca eruptions. However, as they contain at least local breccias, it is likely that progressive and incremental collapse occurred along reactivated scarps [Walker 1984].
6.3. Welding and rheomorphism
Pantelleria is famed for its welded and rheomorphic pumice falls [e.g., Cala di Tramontana center, Stevenson and Wilson 1997] and ignimbrites [e.g., The Green Tuff, Orsi and Sheridan 1984]. The styles and features of Pantelleria rheomorphic ignimbrites are similar to those of other high-grade to extremely high-grade ignimbrites [e.g., the Greys Landing ignimbrite, Idaho, Andrews et al. 2008] including stretching of pumice blocks and lapilli, lineations and foliations, folds on the micro- to meter scale), ramp structures, pull apart structures and tension gashes, gas blisters, upper and basal autobreccias, and rotated clasts (Figure 6a–e). Features more particular to peralkaline rheomorphism include welding throughout the deposit despite thin deposit thicknesses (even <0.5 m) and on steep slopes, large gas cavities, small glass shards or “globules” showing spherical shapes, and round bubbles in previously deformed pumice particles [Schmincke and Swanson 1967; Gibson 1970; Korringa 1971; Schmincke 1974; Williams 2010; Jordan et al. 2018].
Rheomorphism in the ignimbrites is predominantly syn-depositional [Branney et al. 2004; Williams 2010; Jordan et al. 2018]; hot, sticky particles agglutinated as an aggrading deposit from a PDC. The low viscosities meant that the overriding current was able to cause shearing in the underlying deposit where analysis of kinematic indicators show a change of shear direction with height through the deposit, from which it is possible to infer flow direction of the overriding PDC [Williams 2010, following Andrews and Branney 2005; Sumner and Branney 2002]. In places, there is evidence for post-depositional flow, dominantly down slope [Wolff and Wright 1981b; Williams 2010]. The presence of spherical vesicles in strongly flattened fiamme records late-stage exsolution of a gas phase and revesiculation.
Low viscosities [Baker and Vaillancourt 1994; Di Genova et al. 2013; Mahood 1984] due to elevated water [Barclay et al. 1996; Lanzo et al. 2013; Lowenstern and Mahood 1991] and halogen contents [Aines et al. 1990; Carroll 2005; Gioncada and Landi 2010; Lanzo et al. 2013; Lowenstern 1994], despite the rather low pre-eruptive temperatures [Di Carlo et al. 2010; Romano et al. 2018, 2020] and low glass transition temperatures [418–552 °C, Di Genova et al. 2013] of peralkaline melts, are all thought to favor the welding of peralkaline eruptive materials [Dingwell et al. 1998; Quane and Russell 2005]. Eruption columns are inferred to be low, which minimizes cooling of pyroclastic particles during fountaining [Mahood and Hildreth 1986]. Early pre-eruptive temperature estimates for the GT assumed high temperatures (∼950 °C) based on low viscosities [Mahood 1984] and the scant data available from Fe–Ti oxide geothermometry, 933–960 °C [Carmichael 1967; Wolff and Wright 1981a,b]. But more recent work with silicate mineral equilibrium [White et al. 2005] and experimental petrology [Di Carlo et al. 2010; Romano et al. 2020] have shown these earlier values to be overestimated by >200 °C. The GT, the most rheomorphic of all the ignimbrites, is strongly peralkaline with very high chlorine concentrations (up to 1.1 wt%, [Williams 2010; Lanzo et al. 2013]; up to 0.7 wt% for groundmass glass of Mordomo Fm., [Jordan 2014]). It should be noted that pre-eruptive H2Omelt contents from melt inclusions in the GT [H2O⩽4.2 wt%, Lanzo et al. 2013] are comparable with contents from much lower explosivity pantellerite eruptions (H2Omax = 4.4 wt%, Gioncada and Landi ) and with H2Omelt inferred from phase equilibria experiments [Di Carlo et al. 2010; Romano et al. 2020]. This evidence may suggest that other than the pre-eruptive H2Omelt eruptive triggers must be considered; one might be related to an abrupt syn-eruptive viscosity increase due to nanolites growth [Càceres et al. 2020; Di Genova et al. 2020].
7. Did the eruptive pace and the erupted magma volume condition melt evolution and melt productivity?
The depth and the pre-eruptive conditions of pantellerite magma have been constrained at P = 0.5–1.2 kbar, T = 730 °C, H2Omelt = 4% [Di Carlo et al. 2010; Gioncada and Landi 2010; Lanzo et al. 2013; Romano et al. 2020], and at P = 1–1.5 kbar, T = 925 °C, H2Omelt = 2% for trachyte magma (top member of the GT sequence; Romano et al. [2018, 2019]). The presence of a long-lived magma reservoir at mid-crustal levels (∼8 km) that has efficiently homogenized basaltic melts and served as the source of parental magmas to the erupted trachytes and rhyolites was proposed by White et al.  and is supported by the nearly identical incompatible trace-element ratios and patterns on multi-element diagrams recorded in the ignimbrites [Jordan et al. 2021]. These ignimbrites have several common compositional characteristics in addition to similar trace-element ratios, including range of SiO2 (∼64–71 wt%; Figure 7a) and Zr (∼500–2000 ppm; Figure 7b) but differ in terms of peralkalinity, iron enrichment, and degree of silica oversaturation [Jordan et al. 2021]. Most of the felsic eruptive products at Pantelleria have been comenditic trachytes and comendites, with only the earliest (Zinedi Fm.) and latest (GT Fm.) ignimbrites having a pantelleritic affinity. Whether these differences are due to pressure, water content, oxygen fugacity, or something else is beyond the scope of this paper and will be the focus of future studies. However, similar range in both SiO2 and Zr strongly suggest similar degrees and rates of differentiation from trachyte to rhyolite. Considering both effusive and explosive products, Mahood and Hildreth  suggested a frequent rate (13 ± 6 ka) for felsic eruptions on Pantelleria over the past 190 ka. However, when one considers only the seven major ignimbrites to erupt over the past 186 ka, there has been a steady increase in time between eruptions accompanied by a decrease in eruptive volumes which seems to suggest an overall decline in magmatic activity at Pantelleria (Figure 8) with rates declining from 0.076 to 0.007 km3⋅ka−1. These calculated rates based on the data of Jordan et al.  are consistent with the estimates of 0.008–0.055 km3⋅ka−1 calculated by Mahood and Hildreth . Calculated rates using the volume estimates of Rotolo et al.  also show a similar, secular decline from 0.049 to 0.007 km3⋅ka−1. This waning tendency was also observed by Scaillet et al.  in post-GT mildly explosive activity.
The history of the Pantelleria peralkaline volcano is usually divided into an early history dominated by nine ignimbrite eruptions and at least two, but up to five, caldera collapses, followed by a later (post-GT) history, characterized by numerous (>40) low explosivity to effusive eruptions produced from several closely spaced (spatially and temporally) eruptive vents.
The ignimbrites have a remarkably complex stratigraphy due to discontinuous exposures, rapid lateral facies variations, and ubiquitous but highly variable welding and rheomorphism. Inter-ignimbrite activity was characterized by small to moderate explosive or effusive eruptions from scattered vents, whose remnants are now only in few places poorly visible along some coastal scarps. The periodicity of ignimbrite-forming eruptions alternated with lower magnitude events (in a geological scenario that could be similar to the present day one following the eruption of the GT), which raises many still-unanswered questions about triggering mechanisms.
For the above reasons, Pantelleria represents an emblematic case history about the approaches adopted to untangle the volcanostratigraphy and portray its evolutionary history. Early approaches, based on field studies supported by K/Ar ages, defined a basis for future studies, but suffered heavily from the obfuscation caused by the rheological peculiarities of peralkaline magmas and tephra that added to the intrinsic uncertainty of the K/Ar dating. Only in the past ten years has the pre- and post-GT stratigraphy been investigated with multiple absolute/correlative chronological methods (40Ar∕39Ar geochronology, paleomagnetism) coupled to accurate field and petrographic characterization of pyroclastic products. Despite the possible drawbacks about the application of paleomagnetic methods on peralkaline rheomorphic ignimbrites (in principle still plastic below the closure temperature of principal ferromagnetic minerals) and the ability to date K-poor feldspars in young rocks by 40Ar∕39Ar, the congruence of these two methods, within a well-defined field and petrographic context, provided answers to the majority of the open problems with a resolution otherwise unattainable.
Combining 40Ar∕39Ar ages with petrochemical data of seven major ignimbrites erupted during the last 186 ka reveal a slight stretching of ignimbrite recurrence time, coupled to a decrease in erupted magma volumes. This tendency apparently holds also for the post-GT eruptive activity (violent strombolian at most).
Recommendations for future research include additional 40Ar∕39Ar dating of the post-GT trachyte lavas and some still undated minor ignimbrites and local centers, and additional paleomagnetic studies of the emplacement temperatures of some key ignimbrites. Also, the variations in degrees of rheomorphism, and a structural study with detailed analysis of the viscosity, pre- and syn-eruptive halogen content and emplacement temperature, could be a valuable avenue for future study.
We wish to thank two anonymous reviewers for their thorough and helpful comments. The Ar/Ar facility at ISTO is supported by the LABEX project VOLTAIRE (ANR-10-LABX-100-01), the Région Centre project ARGON, and the project EQUIPEX PLANEX (ANR-11-EQPX-0036). RW acknowledges Natural Environment Research Council studentship grant NER/S/A/2006/14156. NJJ gratefully acknowledges funding from the German Academic Exchange Service, Geological Society of London, Mineralogical Society of Great Britain and Ireland, Geologists Association, Quaternary Research Association, Volcanic and Magmatic Studies Group and the Department of Geology at the University of Leicester.