The Pyrenean basement rocks, NE of the Iberian Peninsula, southwestern Europe, include evidence of several pre-Variscan magmatic episodes which indicate the complex geodynamic history of this segment of the northern Gondwana margin from late Neoproterozoic to Early-Palaeozoic times. One of the most significant magmatic episodes was late Mid-early Upper Ordovician (Darriwilian-Katian) age that produced several granitic bodies and volcanic rocks interbedded with Sandbian-Katian sediments. This magmatism is well represented in the Ribes de Freser area (Freser valley, Bruguera and Campelles localities, eastern Pyrenees), where these Ordovician magmatic rocks were affected by an irregularly distributed Variscan deformation and mainly by severe Alpine tectonics, which originated the superposition of several structural units. We present a palinspatic reconstruction of this Alpine deformation (80-20 Ma), that permitted us to infer the geometry, facies distribution, original position, thickness, and significance of these volcanic rocks. This reconstruction allows us to interpret the volcanic rocks cropping out at the Freser valley, Bruguera, and Campelles areas as intra-caldera deposits representing a minimum preserved volume of the order of 100 km3. This may confirm the existence of super-eruptions of Upper-Ordovician age in that sector of the eastern Pyrenees and emphasizes the extent of the Upper-Ordovician felsic volcanism in this sector of the northern Gondwana margin, probably developed in an extensional scenario linked to the development of the Rheic Ocean during Gondwana margin breakup.

Large caldera-forming eruptions, sometimes called super-eruptions [e.g.,1, 2], are among the most catastrophic events that may occur on Earth, comparable to meteoritic impacts. The extrusion of hundreds to thousands of cubic kilometers of molten rock may cause severe direct impacts on the areas surrounding the caldera area and in the longer term may also affect the climate at a global scale. In historical times, none of these eruptions has been recorded, despite other similar eruptions of smaller size (Santorini, 3600 y b.p.; Tambora, 1815; Krakatoa, 1883; Katmai, 1912) have demonstrated their potential destructive capacity. In the recent past, the Toba eruption, in Indonesia, probably the largest eruption of the Quaternary with about 2500 km3 of erupted magma, is thought to have caused a significant imprint on global climate for several years after the eruption [3]. In the geological record, a considerable number of large caldera-forming eruptions have been identified, particularly during the Miocene, in places like the Basin and Range, in USA [e.g.,4], Sierra Madre Occidental, in México [5], Central Andes in Argentina, Chile and Bolivia [6, 7], or Europe [8]. In older areas, the state of preservation of the rocks and structures derived from these eruptions precludes to establish good stratigraphic correlation and, therefore, to identify the size of the corresponding eruptions. However, still there are a good number of examples in Paleozoic and Mesozoic [e.g.,9-13], and even in the Archean [14] to be examined.

The Pyrenees is an E-W trending Alpine belt formed between the Upper Cretaceous and the Miocene by the convergence between the Iberian and European plates [15]. Their pre-Variscan basement rocks record several well-developed episodes of calc-alkaline volcanism, associated with the Cadomian, Ordovician, and Variscan tectono-thermal events [16-23]. These volcanic episodes are marked by a profuse development of mainly felsic pyroclastic deposits associated with coeval plutonic bodies [19-22]. Despite these volcanic rocks have been affected by the Variscan (370, 290 Ma) and Alpine (80, 20 Ma) deformational episodes and further erosion, the current exposure of some of them still allows to recognize the existence of large volcanic centers that sourced large volume eruptions, as for instance the collapse calderas associated with the Permo-Carboniferous (Variscan) volcanism already suggested by Martí [9, 18]. Although the volcanic rocks of the Freser valley were the first described Upper-Ordovician volcanic rocks in the Pyrenees, after the pioneering works of the 1970s and 1980s [17, 24-27], little attention has been paid to these rocks and more recent works were mainly devoted to their equivalent plutonic counterparts [19, 28-31]. Recently, new data arose about the presence of subvolcanic felsic Upper-Ordovician rocks interbedded within the pre-unconformity strata close to the base of the Upper-Ordovician succession in the Pyrenees (ca. 453 Ma,[32]; ca. 457 Ma, [33]) and in the neighboring Mouthoumet massif (ca. 455 Ma, 33). This volcanism was probably widespread in other western Pyrenean massifs, such as the Pierrefitte dome [34], although geochronological data are needed to confirm the Ordovician age attributed to the rhyolitic sills of this massif. Upper-Ordovician metavolcanic rocks have also been described in the Catalan Coastal Range (ca. 455‒452 Ma;22, 30]), south of the Pyrenees.

In this paper, we propose the existence of an Upper-Ordovician collapse caldera in the Freser valley, eastern Pyrenees, through the reconstruction of the complex Alpine tectonics affecting this area (Figure 1). This interpretation is based on the establishment of precise stratigraphic correlations, on the facies analysis of the volcanic materials, focused in particular on the changes in thickness they show and on the presence of different lithologies that evidence this type of volcanic process, and on the palinspastic reconstruction of the studied terrains, strongly affected by Alpine tectonics. These data have allowed us to reconstruct the original structure of the area and the original distribution of the volcanic materials, which can only be explained by the presence of this collapse caldera. The example presented here constitutes the first example of Ordovician caldera described in the Pyrenees and may provide some keys to understanding the extent and significance of the felsic Ordovician magmatism well exposed in the Pyrenees and surrounding areas.

1.1. Structural and Stratigraphic Framework

In the Pyrenes, the stacking of several Alpine thrust sheets originates from a complete pre-Variscan succession, ranging in age from Ediacaran to pre-Variscan Carboniferous, crops out extensively in the central part of the chain (15) (Figure 1). The study area is in the eastern Pyrenees, in the Freser valley, on the southern slope of the Canigó massif, where a kilometric scale Alpine antiformal stack is cropping out (Figure 2). The Freser valley antiformal stack is truncated to the north by the Ribes–Camprodon out-of-sequence thrust [27] that separates this structure from the Canigó massif and the rest of the Alpine chain (Figure 2).

1.1.1. The Canigó Massif

The Canigó massif exhibits one of the most complete pre-Variscan succession of the Pyrenees. The lowermost part of this succession is a thick (up to 3000 m) unfossiliferous metasedimentary sequence that has been subdivided, from bottom to top, into the Canaveilles Group (Ediacaran to Cambrian Series 2) and Jujols Group (Cambrian Series 2 to Early-Ordovician) (35) (Figure 3(a)). An Upper-Ordovician succession forming a broad fining-upward siliciclastic package and ranging from 100–1000 m in thickness [36, 37], unconformably overlies the Canaveilles and Jujols rocks [38]. In this succession Hartevelt [37] distinguished five formations, from bottom to top: the Rabassa Conglomerate, Cava, Estana, Ansobell, and Bar Quartzite formations (Figure 3(a)). The basal Upper-Ordovician alluvial-to-fluvial conglomerate (Sandbian? La Rabassa Conglomerate Formation) is up to 100 m thick, and made up of heterometric clasts of vein quartz, quartzite, and slate [37]. The overlying Cava Formation, 0–850 m thick, consists of sandstones and shales rich in Katian brachiopods and bryozoans [37, 39]. The Estana Formation, 0–200 m thick, is composed of limestones and marly limestones with abundant late–Katian brachiopods, bryozoans, echinoderms, and conodonts [39]. The top of the carbonate succession is capped by the black-grey shales of the Hirnantian Ansobell Formation, 20–320 m thick, subsequently overlain by the 8–18 m thick Hirnantian Bar Quartzite Formation [40, 41].

A set of subvertical NNE–SSW-trending normal faults abruptly controlling the thickness of the basal Upper-Ordovician formations can be recognized on the southern slope of the Canigó massif, west of the study area [42]. Faults cut the Upper-Ordovician unconformity, the Rabassa conglomerates and the lower part of the Cava Formation and their displacement diminishes progressively up-section and vanishes in the lower part of the Cava Formation, indicating that the faults became inactive before deposition of the overlying late Katian Estana strata [43]. Moreover, sharp variations in the thickness of the Upper-Ordovician strata have been documented by Hartevelt [37] in other neighboring areas. According to Puddu et al [43], drastic variations in grain size and thickness can be attributed to the development of palaeotopographies controlled by faults and subsequent erosion of uplifted palaeoreliefs, with subsequent infill of depressed areas by alluvial fan and fluvial deposits.

The Canigó massif also contains the most complete record of the Ordovician magmatism in the Pyrenees, as magmatic rocks ranging in age (from U-Pb radiometric dating in zircons) late Early- to Upper-Ordovician (ca. 475–450 Ma), from felsic to mafic and from plutonic to volcanic crop out extensively in this massif (Figure 3(a)). The lowermost Canaveilles Group is cut by a voluminous orthogneissic body, about 2000 m thick, derived from late Early- to Mid-Ordovician intrusives, the Canigó gneiss, (ca. 472–462 Ma, [28, 31]). Equivalent plutonic bodies are well represented in the Aston (ca. 467–470 Ma; [44, 45]), Hospitalet (ca. 472 Ma; [44]), Roc de Frausa (ca. 477–476 Ma;[ 19, 28]) and Albera (ca. 470 Ma; [46]) massifs (Figure 1). Equivalent felsic volcanics have been documented in the Albera Massif, where subvolcanic rhyolitic rocks have yielded similar ages to those of the main gneissic bodies (ca. 474–465 Ma; [46]); coeval mafic rocks are clearly subsidiary [31]. In the Canigó massif, small gneissic bodies with Sandbian–Katian protolithic ages have been also recognized, indicating that the magmatism continued until Upper-Ordovician times yielding another set of magmatic rocks that constitute the protoliths of the Cadí (ca. 456 Ma; [29]), Casemí (446, 452 Ma; [29]), Núria (ca. 457 Ma; [30]) and Canigó G1-type gneiss (ca. 457 Ma; [31]). The lowermost part of the pre-Upper-Ordovician succession hosts meter-scale thick bodies of metabasite (ca. 460 Ma, [31]) and metadiorite sills (ca. 453 Ma; [29]).

1.2. The Freser Valley Antiformal Stack

The Freser valley antiformal stack [27] crops out in a kilometric scale culmination at both sides of the Freser valley (Figure 2). It is a second-order structure related to the main Alpine stack of basement rocks forming the Pyrenean Axial Zone. The antiformal stack involves several thrust sheets bearing different pre-Variscan stratigraphic successions (Figure 3(b), (c) and (d)) (26, 27, 47) and developed during the Early Eocene (Ypresian, according to 48). It is truncated to the north by the Ribes–Camprodon out-of-sequence thrust [27] that separates these thrust sheets from the Canigó massif and the rest of the Pyrenees (Figure 2). The different thrust sheets that form the antiformal stack are, from bottom to top: the Montagut, El Baell, Bruguera, Ribes de Freser, and Serra Cavallera thrust sheets (27) (Figure 2). They show an E-W trending antiformal geometry with the higher structural units deformed by the lower ones, demonstrating a piggy-back thrust sequence. Uppermost units in the southern limb have been completely overturned showing downward-facing folds. Thrusts climb up section from Variscan basement into Cenozoic cover rocks and the upper detachment of the antiformal stack is located at the bottom of the Eocene marly succession. One of the main characteristics of this antiformal stack is that every thrust sheets exhibits a different pre-Variscan stratigraphic succession and internal structure, that do not match with those exhibited by the rest of the Canigó massif.

The Montagut thrust sheet is the lowermost unit cropping out in a tectonic window along the Freser valley [27]. It is made up of undated felsic volcanic rocks cropping out at the bottom of the Freser valley, and unconformably covered by the uppermost Cretaceous-Lower Paleocene continental deposits (Garumnian red-beds) (Figure 2).

The El Baell thrust sheet is entirely made up of pre-Variscan basement rocks (Figure 3(d)). It comprises a 500 m-thick succession composed of limestones, marly limestones, shales with centimeter-thick, parallel bedding nodules, and shales [26, 27] in which three limestone-dominated levels can be distinguished (the El Baell Formation, 47). Conodonts and echinoderms allowed Robert [26] to attribute an early Katian (former Caradoc) age to the strata forming this unit. Sánz-López and Sarmiento [49] attributed conodont faunas to a late Katian age and correlated them with that described from the Estana Formation cropping out in the Canigó massif [37]. Black shales of probably Hirnantian age unconformably cap the former Katian beds [47]. Neither the base of the Upper-Ordovician succession nor the contact with Silurian beds is exposed in this unit. According to Muñoz [27] and Puddu et al [47], the rocks that form the El Baell unit are affected by E–W oriented, asymmetrical, south-verging folds with wavelengths ranging from a few to several meters. Folds are open to tight, with axial surfaces dipping 15°–30° toward the N and are associated with a subhorizontal or gently north dipping well-expressed axial plane cleavage. These folds should be Variscan in age because they do not affect the Alpine thrust that bounded these units or the Upper Cretaceous-Paleocene rocks of the adjoining tectonic units. Puddu et al [47] suggest that the drastic changes in thickness and facies of the El Baell and Estana formations, point to extensional tectonics that controlled the development of (half-) grabens, although the initial orientation of the faults controlling these grabens is difficult to establish due to late Variscan and Alpine deformations.

The Bruguera thrust sheet lies on the top of the El Baell unit and exhibits a 300 m-thick unfossiliferous slate-dominant succession, pre-Variscan in age and attributed to the Cambrian–Lower Ordovician by Muñoz (27) (Figures 2 and 3(c)). This succession is subsequently overlain by a volcanic unit [26, 50, 51] that consists of at least 300–400 m of strongly welded rhyolitic ignimbrites, variegated in color, that show a marked flow banding texture, which overlies a basaltic andesite lava flow [27, 51]. Recent U–Pb data on these volcanic rocks [52, 53] suggest a radiometric age of ca. 460–459 Ma (late Darriwilian–early Sandbian) (Figures 3(c) and 4). To the south, the pre-Variscan succession of the Bruguera unit is unconformably covered by uppermost Cretaceous-Lower Paleocene continental deposits (Garumnian facies) (27) that, in turn, form a third-order antiformal stack in the southern limb of the Freser valley antiformal stack (Figures 2 and 3(a)). The slate-dominated succession and the volcanic rocks exhibit different mesostructural features. The slate-dominated succession is weakly deformed, and the bedding surfaces display a marked dispersion resulting from the presence of two main fold systems with axes oriented E–W and N–S [27, 47]. According to these authors, at the outcrop scale, their relative chronology has not been established and both fold system produced decameter- to meter-sized open folds with subhorizontal axes with local development of a cleavage-related fold. In contrast, the Upper-Ordovician volcanic rocks display a well-developed flow structure that is used as surface reference while no folds are recognizable at outcrop scale in these rocks. Flow structure is affected by a kilometric E–W trending fold that can be correlated to the E–W Alpine fold affecting the basal thrust between the Bruguera and El Baell units as well as all the antiformal stack [27, 47]. It can be assumed that both fold systems affecting the underlying slate-dominant succession are pre-Alpine in age, because earlier Upper Cretaceous-Palaeocene rocks cropping out in neighboring units are not affected by these deformations. The E–W folds of the Bruguera unit can be correlated with the E–W folds of the El Baell unit, thus being probably Variscan in age, whereas a pre-Upper-Ordovician age can be proposed by the N-S ones, as they have not been recognized in the Upper-Ordovician volcanic rocks. If we assume a Cambrian-Ordovician age for the slate-dominated succession, the contact between the slates and the overlying volcanic sequence may be equivalent to the Upper-Ordovician (“Sardic”) unconformity described close to the study area [38, 47].

The Ribes de Freser thrust sheet is separated from the Bruguera unit by a set of north-dipping normal and reverse faults. The Ribes de Freser thrust sheet exhibits a 200–600 m-thick pre-Variscan basement succession, mainly composed of Upper-Ordovician volcanic, subvolcanic, and volcano-sedimentary rocks interbedded in the Katian sediments (Fig, 3B) (17, 24-27). The volcanic activity was mainly explosive [17] and a granitic body with granophyric texture, the Ribes de Freser granophyre dated at 458 ± 3 Ma [30], intrudes into the lower part of the succession. Martí et al [17] described these rocks as members of a calc-alkaline suite composed of basaltic andesites, dacites, and rhyolites that spread out throughout the Upper-Ordovician succession.

Finally, the uppermost Serra Cavallera thrust sheet is mainly made up of Silurian black shales, Devonian limestones, detrital Carboniferous marine “Culm,” and Permian continental deposits [27]. The basal Serra Cavallera thrust exhibits an out-of-sequence reactivation cutting the previously formed third-order antiformal stack located on the southern edge of the underlying Bruguera thrust sheet (Figure 2). The internal structure of this unit results from the superposition of Variscan and Alpine thrusts, both cutting previous Variscan cleavage-related folds [27].

Apart from the decameter to hectometer size thrusts and folds, a pervasive pressure–solution cleavage is the most extensive Alpine mesostructure preserved in the Garumnian marlstones and marly limestones. In the Garumnian detrital layers and massive limestones, cleavage is weakly developed, and small-scale thrusts structures are well preserved [53]. These Alpine mesostructures are not recognized in the pre-Variscan rocks of the El Baell, Bruguera, Ribes de Freser, and Serra Cavallera thrust sheets.

1.3. Field and Petrographic Characteristics of Volcanic Rocks

For what concerns the identification and characterization of the volcanic rocks we conducted detailed fieldwork covering all the study areas and elaborated several stratigraphic columns (Figures 3 and 4), which permitted to establish the stratigraphic correlations between the main volcanic units (Figure 4). This was crucial to determine facies and thickness variations along them, which were used as reference in the structural reconstruction of the area and in the volume estimates of the volcanic rocks. Samples of all volcanic rocks were collected at different locations and later studied in the petrographic microscope for textural and mineralogical characterization (Figure 5). The petrology and geochemistry of these rocks were already studied by Martí et al [17], who established that they correspond to highly altered and transformed primary calc-alkaline volcanic rocks forming a series that goes from basaltic andesites to peralkaline rhyolites.

Our field study concentrated on the volcanic rocks that form the Upper-Ordovician volcano-sedimentary succession of the Bruguera thrust sheet. These rocks crop out at different localities in the Campelles, Ribes de Freser, and Bruguera zones (Figures 3 and 4). The bottom of the succession corresponds to the Upper-Ordovician (“Sardic”) unconformity, well recognized near the study area [38, 54]. The succession includes at the base, and only locally a reddish colored breccia deposit that contains irregular blocks and smaller fragments of Ordovician and older rocks, all cemented by an interclast fine-grained clay matrix. This breccia deposit is discontinuous and does not appear in all outcrops. In this sense, they mark the transition between the external footwall and the internal hangingwall (whose top represents the basin floor). Above this breccia or, more frequently, directly above the Upper-Ordovician unconformity, there is a thick (200, 260 m) package of lava flows of basaltic andesite composition, which shows a similar aspect from massive to highly vesiculated in all outcrops where it is found. Conformably above the basaltic andesite lavas there is a succession of lithic and pumice-rich ignimbrites (150, 250 m thick), where pumices have been flattened and converted into fiammes. The content and size of lithic fragments increases

progressively toward the top of these ignimbrites becoming a breccia-like deposit with rounded lithic fragments, with sizes ranging from few centimeters to half a meter, and flattened pumices. These lithic fragments are mostly of Cambro-Ordovician, Upper-Ordovician rocks, and previous volcanic rocks. Conformably overlying this unit of lithic-rich ignimbrites appear a thick succession, up to 1 km thick, of lithics-poor, strongly welded to rheomorphic ignimbrites, which are eroded at the top by the uppermost Cretaceous-Lower Paleocene red beds (Garumnian, 27).

The best exposed and thickest succession appears at the north of Campelles (Figure 5(a)). Here, the basaltic andesites lie directly on the Upper-Ordovician unconformity. With a total thickness of up to 260 m, these rocks of dark greenish color are massive at the base and highly vesiculated toward the top, with all the vesicles now filled by secondary minerals (Figure 5(b)). Under the microscope these basaltic andesite lavas show a porphyritic texture (Figure 6(a) and (b)). Phenocrystals are generally small in size (< 2 mm). They correspond to altered plagioclase, partly replaced by chlorite, iron calcite, and epidote, as well as ferromagnesian minerals completely pseudomorphosed to chlorite, iron oxides, and carbonates, among which, however, some original sections of pyroxene and olivine have been recognized. The groundmass is microcrystaline and sometimes contains microlites of sodium plagioclase and small crystals of iron oxides and chlorite.

These lavas are unconformably overlain by the first package of ignimbrites (Figure 5(c)), which is formed of several units of incipiently to partially welded, lithics-rich ignimbrites, with

abundant pumice fragments rich in phenocrysts of plagioclase and quartz. These ignimbrite units also show planar contacts between them, without any intercalation of other pyroclastic or sedimentary deposits. Their thickness varies from 1 to 10 m and all of them are characterized by the presence of elongated pumices (fiammes), ranging in size from a few to 20 cm, which are always grouped in planes parallel to stratification. The amount of pumice fragments tends to increase toward to the top of each unit. The pumice fragments are completely devitrified and transformed into phyllosilicates (clay minerals, chlorite, and muscovite). Nonetheless, it is still possible to recognize some aspects of the original eutaxitic texture (e.g., frayed edges, elongated vesicles, flow bands, and rotation of phenocrysts, etc.) of these ignimbrites (Figure 6(c)). The content of crystals (mostly quartz, and strongly altered plagioclase and biotite) is similar in the pumices and in the matrix, where most of them are fragmented, and may range from 15% to 45%. The content in lithic fragments is above 20% in all cases and increases progressively toward the top of the succession where it can be as high as 60% or more, then becoming a pyroclastic breccia (Figure 5(d)). Most of the lithic fragments are Upper-Ordovician and pre-volcanic rocks, some of them similar to the underlying basaltic andesites and rhyolites. The size of the lithic fragments ranges from a few centimeters to half a meter.

The uppermost unit of this volcanic succession corresponds to a single ignimbrite unit, up to 1000 m thick, which conformably overlies the previous ignimbrites and breccias. This thick pyroclastic deposit corresponds to a strongly welded ignimbrite, very rich in pumices and poor in crystals and lithic fragments (Figure 5(f)). Pumice fragments are all stretched on a plane parallel to the base of the deposit, being this deformation more intense upwards. In the upper half of the deposit, the extreme stretching of pumice fragments acquires a flow banding texture (Figure 5(g)) similar to that of silicic lava flows. Pumice fragments are completely devitrified and transformed into crypto and micro-crystalline quarts and clay minerals, but textural features characteristic of these extremely welded rocks are still recognizable (Figure 6(d) and (e)). The crystal content is lower (up to 15%) than in the previous ignimbrites and corresponds to quartz, plagioclase, and biotite. They appear as phenocrysts in the fiammes or as fragments of these phenocrysts in the matrix. The content in lithic fragments, always of small size (< 1 cm), is very low (up to 2%), and they correspond to schists, and volcanic lithics with rhyolitic and andesitic compositions. The matrix of the ignimbrite is highly silicified and, in some places, it appears partially recrystallized to monocrystalline quartz. However, in some parts of the lower half of the deposit it is still possible to recognize pseudomorphs of original vitroclasts (Figs, 6d, e). The characteristics of this upper ignimbritic unit, particularly for what concerns the upper half of the deposit with flow banding and flow folds (Figure 6(f)), is proper of the extremely welded rheomorphic ignimbrites [see 55].

Completeness of the volcanic succession described above varies across the study area (Figure 4). At the base, the volcanic rocks either appear deposited on the Upper-Ordovician unconformity or are in contact by fault with Cambro-Ordovician rocks (Figures 2 and 4). Also, the top is either faulted and in contact with Cambro-Ordovician or Upper-Ordovician rocks or corresponds to an erosive contact with the Garumnian (uppermost Cretaceous-Lower Paleocene) red beds (Figures 2 and 4). The degree of erosion affecting the upper part of the volcanic succession is unknown, so its total original thickness cannot be estimated. The three described volcanic units or only some of them are distinguishable in the studied stratigraphic logs, showing a similar lithology, color, and textural aspects. However, some facies and thickness variations exist between these three main units of the volcanic succession when comparing the studied stratigraphic logs, suggesting irregularities and differences in the areal distribution and thickness of these rocks.

The reconstruction before the Pyrenean deformation of the pre-Variscan units forming the different thrust sheets of the Freser valley antiformal stack can be established according to their age, lithology, and internal structure (Figure 7(a)). In order to achieve this purpose, the different thrust sheets have been restored following the preservation of the bed length of the Garumnian layers. Restoration has been made using a new cross-section along the study area, that has been constructed based on previous information [27, 47] and new collected field data (Figure 7a).

Alpine structures have been restored from the youngest to the oldest, thus starting with the restoration of the out-of-sequence Serra Cavallera and Ribes-Camprodon thrusts (Figures 2 and 7(a)). From the map and the cross-section the displacement of the former can be estimated at ca. 1km (Figure 7(b)). In contrast, the out-of-sequence displacement of the Ribes-Camprodon thrust cannot be precisely evaluated, although from the cross-section a minimal displacement of several hundreds of meters can be proposed (Figure 7(b)). After that, we sequentially restored the Montagut, Baell, Bruguera-Ribes de Freser, and Serra Cavallera thrust sheets (Figure 7(c) and (d)). In the case of the Bruguera-Ribes de Freser and Serra Cavallera ones, we have used the cut-offs of the uppermost Cretaceous-Lower Paleocene beds as piercing points, but in the El Baell unit, the absence of the post-Variscan cover makes the restoration more speculative. The minimum calculated displacements are 2 km for the El Baell unit, 4.2 km for the Bruguera unit, and 4 km for the Serra Cavallera unit, which show an accumulated southwards displacement of ca. 10.5 km for the units cropping out in the Freser valley (Figure 7(e)). We cannot estimate the amount of displacement of the Montagut units as its footwall does not crop out.

Restoration of the Alpine deformation places the uppermost Serra Cavallera unit in a pre-Alpine northernmost position (Figure 7(e)), probably near the culmination of the big antiform shown by the Canigó massif, where similar Devonian succession crops out northwards (Figure 1). Restoration also situates the Bruguera and Ribes de Freser units in a pre-Alpine northern position, the El Baell unit in an intermediate setting, and the lowermost Montagut unit lying originally in a southernmost position (Figure 7(e)).

4.1. The Freser-Bruguera-Campelles Caldera

The studied Upper-Ordovician volcanic succession represents an exceptionally large eruptive event that occurred coevally with Upper-Ordovician extensional fault development. This volcanic succession forms part of a wider and longer magmatic episode that generated abundant explosive and effusive volcanism and intrusion of calc-alkaline silicic rocks, lasting in time for about 20 Ma from ca. 472 to 452 Ma. However, in the studied volcanic succession the absence of any discontinuity among the deposits suggests that all of them form part of the same eruptive event or at least of successive events that were not separated so much longer in time and occurred in the same area. The number of deposits was still preserved and their stratigraphic correlation permits to propose an interpretation of their significance in terms of eruptive activity.

Proximal sedimentary breccias, locally observed at the base of the volcanic succession and unconformably overlying the pre-Upper-Ordovician succession, indicate that sedimentation started in a tectonic basin that was later totally infilled with volcanic material. The characteristics of these breccias (restricted distribution, presence of angular blocks of different compositions, absence of primary volcanic components, etc.) suggest that they correspond to proximal deposits related to the erosion of a paleorelief, probably formed of uplifted blocks, before the onset of volcanism, and probably deposited by alluvial fans at the foot of the main fault scarps rimming the basin.

The deposition of a thick (up to 260 m) succession of basaltic andesite lava flows on the sedimentary breccias represents the first volcanic episode in the area. The fact that these lava flows appear sometimes directly on the Upper-Ordovician unconformity indicates that they covered a larger extension than the breccias. Volcanism might have been connected to the faults controlling the basin subsidence. The characteristics of the lava flows do not provide information on their emplacement environment, but the reddish color of the breccias and the abundance of large vesicles in most of the lavas account for a subaerial emplacement.

The deposition of the basaltic andesite lava flows was immediately followed by the emplacement of different units of pumice-rich ignimbrites with a variable number of lithic fragments of different compositions. The eutaxitic texture, defined by the flattening and elongation of pumice fragments, may be an original feature due to an emplacement at high temperatures [e.g., 56], or a secondary feature originated by alteration and transformation of original pumice clasts into clay aggregates [57]. In this case, the recognition of sintered vitroclasts in the matrix and of fluidal textures marked by rotation of individual phenocrysts and angular lithic fragments, suggest this is an original feature, so these ignimbrites would have been emplaced at a relatively high temperature, well above the glass transition temperature of silicic glasses [58]. The presence of lithic fragments of volcanic nature indicates that volcanism started earlier in other parts of the basin or in other adjacent basins. The progressive increase of the size and abundance of lithic fragments toward the top of these ignimbritic units suggests a progressive enlargement of the eruption conduits, which may be related to an increasing displacement of the faults that controlled the subsidence of the basins and which were also acting as eruption feeders. In fact, the last units correspond to pyroclastic breccias that may be interpreted as co-ignimbrite lag breccias, which are a distinctive feature in caldera-forming successions [e.g., 59-61].

Immediately above this sequence of lithic-rich ignimbrites appears a continuous (i.e., lack of any internal discontinuity) thick (1000 m) succession of strongly welded ignimbrites, which in a progressive way show toward the upper half part of the deposit with a clear rheomorphic texture. This is defined by the extreme elongation of the pumice fragments, giving rise to the appearance of a marked flow banding that makes these ignimbrites appear as lava-like rocks. It is remarkable the absence of lithic fragments in these ignimbrites, thus indicating that they erupted through large vent fractures that did not suffer erosion during the extrusion of the magma. The main feature of this ignimbritic package, apart from its texture, is the preserved thickness, which is up to 1000 m despite having been truncated at the top by the deep erosion at the bottom of the Garumnian sediments.

All the features described above, account for the presence of a caldera collapse structure formed during the eruption and emplacement of the ignimbrites, and which probably developed along the same or part of the same faults that were controlling the subsidence of the tectono-sedimentary basin. The onset of the caldera subsidence would have started following the eruption of the first ignimbrites, which would have been responsible for the decompression of the associated magma chamber and of the achievement of the caldera collapse conditions in the volcanic system. The presence of collapse calderas that totally or partially use the tectonic structure of volcano-tectonic basins is not uncommon and has already been described in the Permo-Carboníferous successions of the Catalan Pyrenees [9] but also in modern analogues in other regions such as the graben calderas of the Sierra Madre Occidental in Mexico [62]. The formation of this collapse caldera would explain the preservation of this thick volcanic succession but also the main features that it presents. The emplacement of the andesitic lava flows was probably not related to the caldera formation but indicates that volcanism had already started in the basin after the deposition of the first sediments and presumably was controlled by the same faults system responsible for the subsidence of the basin. In this sense, these faults would have facilitated not only the ascent of deep magmas and their eruption at surface, but also their accumulation at shallow levels, allowing for the development of a large magmatic body beneath the basin. The eruption of the first ignimbrites marked the opening of this magmatic reservoir or chamber and its progressive decompression until conditions for caldera collapse were reached. This means when the resistance threshold of the chamber roof was exceeded and it started to collapse inside the magma chamber along the faults that were controlling basing subsidence and probably also along newly formed normal faults, all together defining a ring-fault system. The moment in which the caldera collapse started is marked by the eruption of the pyroclastic breccias (co-ignimbrite lag breccias) that can be observed on top of the lithic-rich ignimbrite units. Immediately after started the massive eruption of the remaining magma in the form of high-temperature ignimbrites through highly dense pyroclastic density currents (PDCs) inside and outside the caldera depression. The characteristic of the deposits that can be now observed, including a total thickness up to 1000 m, absence of paleosols and/or other temporal discontinuities, presence of co-ignimbrite lag breccias, and the high degree of welding of the deposits, suggest that they mostly correspond to the intra-caldera succession, while the deposits extra-caldera were eroded out by the Garumnian transgression, although some remains can still be observed locally in the study area (Figure 2). The rapid intra-caldera deposition of magmatic material facilitated the preservation of a high temperature, so the pyroclasts could sinter and weld till the extreme of becoming rheomorphic and undergoing a secondary flow once they were already emplaced [see 10, 12, 63]. The facies and thickness variations observed in this volcanic succession suggest the existence of an irregular paleotopography on which these rocks deposited. This could correspond to an original feature of the Upper-Ordovician unconformity or to the active extensional faults occurring during caldera collapse, which could have developed in a piecemeal fashion, as it occurs in many large caldera [e.g., 4, 62].

This evolution in three steps matches with the proposal of Huppert and Sparks [64] of a progressive heating of the crust originated by the arrival of mafic magmas. After deposition of sedimentary breccias, and according to these authors, in a first step cold crust favors basaltic magma emplacement. Basalts can reach the surface, in our case through a preexisting or coeval fracture system, as basaltic differentiates or andesites. However, melting of the crust induced by the emplacement of successive batches of mantle-derived magmas at the base of the crust would be a plausible mechanism to explain the development of silicic magmatism. This is suggested rather than magma differentiation due to the peraluminous character of the rhyolite ignimbrites [9]. Volume of silicic magmatism would have increased with time giving rise to the formation of a collapse caldera and the subsequent emplacement of a large volume of ignimbrites. Other silicic bodies could not reach the surface and were emplaced at shallow levels giving rise to subvolcanic granites (Ribes Granophyre) not far from the caldera location.

The subsequent Variscan and Alpine tectonics have modified the original caldera, splitting its original structure into pieces of different sizes. However, we have reconstructed the pieces of this puzzle back to their original position, once restored a representative cross-section of the study area. Based on this reconstruction and the map distribution of the different units we can estimate the extension and volume of the deposits accounting for the proposed caldera-collapse structure and provide its minimum dimensions (Figure 8). So, the caldera collapse structure would have had a polygonal shape, as it corresponds to this particular type of calderas [e.g., 62], and minimum external dimensions of 12 km in the East-West direction and 7.5 km North-South direction, with an averaged depth of 1000 m. This accounts for a minimum preserved volume of at least 100 km3, if we also consider the other outcrops of volcanic rocks present in the study area (see Figure 2). However, the total volume of volcanic products emitted in this eruption should be much higher, probably of several hundreds of km3, if it is compared with caldera-forming eruptions responsible for calderas of similar sizes [see 65] and if the eroded and fault-truncated parts would be considered. These dimensions would classify this Upper-Ordovician caldera-eruption as a large caldera-forming eruption or super-eruption [see 1, 2].

4.2. Regional Considerations

The Upper-Ordovician caldera deposits described here emphasize the wide distribution of the felsic late Mid-early Upper aged-Ordovician volcanism in this sector of the northern Gondwana margin. As stated, this volcanism has been recently described in other areas of the Pyrenees (ca. 457–453, [32, 33]), Mouthoumet massif (ca. 455 Ma, [33]), Catalan Coastal Range (ca. 455‒452 Ma; [22, 30]) and in Sardinia, where Ordovician felsic volcanic rocks are especially well developed [66-74]. Nonetheless, in Sardinia the felsic volcanism started earlier: it ranged from the Furonginan or Tremadocian to the Katian [67-69, 71, 73], and in contrast with the Pyrenees, Ordovician felsic plutonic rocks are subsidiary. We interpret that the Upper Darriwilian-Sandbian volcanism reported here represents the final pulse of an Ordovician tectonothermal event, which in the Pyrenees started with the Floian–Darriwilian emplacement of plutonic rocks [75]. This event belongs to the Cambrian-Ordovician voluminous magmatism that occurred along the northwestern Gondwanan margin of continental Europe [see review in 76]. Several genetic models have been proposed to explain this magmatism, from subduction-related melts [69, 73, 77-80], to post-collisional decompression melting without significant mantle involvement [81], or partial melting of the lower continental crust in response to either mafic underplating or mafic intrusion [34, 43, 75, 82-84] or asthenospheric upwelling [31, 85, 86]. The example described here, with thick volcano-related deposits controlled by fault activity, fits better with an extensional setting, also compatible with the extensional faults of the same age described in the Canigó massif [43] and with the drastic changes in thickness and facies of the El Baell and Estana formations. This extensional scenario could be linked to the break-up of Gondwana and the birth of intervening oceans, in a pivotal time interval between the end of Neoproterozoic-Early Cambrian (Cadomian) accretionary tectonics and the Cambrian-Ordovician opening of the Rheic Ocean [87-90].

We present a palinspatic reconstruction of the Alpine deformation affecting the Ribes de Freser area (Eastern Pyrenees), that permitted to interpret the volcanic rocks cropping out at the Freser valley, Bruguera, and Campelles areas as intra-caldera deposits representing a minimum preserved volume of ca. 100 km3. This reconstruction permitted to infer the geometry, facies distribution, original position, thickness, and significance of these volcanic rocks that confirm the existence of super-eruptions of Upper-Ordovician age in that sector of the eastern Pyrenees. This also emphasizes the extent of the Middle-Upper Ordovician felsic volcanism in this sector of the northern Gondwana margin. Finally, we interpreted this super-eruption as being favored by an extensional scenario coeval to Gondwana break-up and the opening of the Rheic Ocean.

All data used in the preparation of this contribution is included in the manuscript.

The authors declare that there has been no conflicts of interest between them or with third parties in the preparation of this manuscript.

JM is grateful for the MECD (PRX16/00056) grant. This research has been partially funded by the Spanish grants MINECO CGL2017-84901-C2, CGL2017-87631-P, PGC2018-093903-B-C22, MICINN PID 2020-114273GB-C21 of the Ministerio de Ciencia, Innovación y Universidades–Agencia Estatal de Investigación– Fondo Europeo de Desarrollo Regional, UE and project PID2020-117598GB-I00, funded by MCIN/ AEI /10.13039/501100011033. We thank Hugo Murcia and Karly Nemeth for their constructive reviews

Exclusive Licensee GeoScienceWorld. Distributed under a Creative Commons Attribution License (CC BY 4.0).