Catching the Main Ethiopian Rift evolving towards plate divergence

Magmatism accompanies rifting along divergent plate boundaries, although its role before continental breakup remains poorly understood. For example, the magma-assisted Northern Main Ethiopian Rift (NMER) lacks current volcanism and clear tectono-magmatic relationships with its contiguous rift portions. Here we define its magmatic behaviour, identifying the most recent eruptive fissures (EF) whose aphyric basalts have a higher Ti content than those of older monogenetic scoria cones (MSC), which are porphyritic and plagioclase-dominated. Despite these differences, calculations highlight a similar parental melt for EF and MSC products, suggesting only a different evolutionary history after melt generation. While MSC magmas underwent a further step of storage at intermediate crustal levels, EF magmas rose directly from the base of the crust without contamination, even below older polygenetic volcanoes, suggesting rapid propagation of transcrustal dikes across solidified magma chambers. Whether this recent condition in the NMER is stable or transient, it indicates a transition from central polygenetic to linear fissure volcanism, indicative of increased tensile conditions and volcanism directly fed from the base of the crust, suggesting transition towards mature rifting.

www.nature.com/scientificreports/ In the NMER, we have also distinguished three Eruptive Fissures (EF; Fig. 2), which differ from the volcanotectonic lineaments giving monogenetic scoria cones: (a) the eruptive fissures form the most extended alignments of cones, 6 to 14 km long, belonging to the same eruptive event, recognized through remote sensing data and in the field; (b) the eruptive fissures are neither eroded nor partially buried and, in the field, their products are generally massive and mostly aphyric, without evidence of alteration or weathering; (c) the eruptive fissures are accompanied by a relatively larger (1-2 orders of magnitude more voluminous) amount of lava with respect to the monogenetic scoria cones.
The northernmost eruptive fissure (herein Metehara EF) is formed by two parallel and én-échelon segments (11 and 7 km long), extending from the middle (ca. up to 500 m above the rift floor) southern slope of Fantale volcano to the scoria cones inside Lake Besaka ( Fig. 2A). At its northernmost tip, the eruptive fissure does not affect the rim of Fantale caldera. Along the Metehara EF a chain of scoria cones and a small-volume (~ 53 × 10 6 m 38 ) lava field formed ( Fig. 2A). These products have been previously associated with the two post-caldera lava flows within the caldera (e.g. 33 ). Nonetheless, our survey reveals that intra-caldera lava flows are obsidian coulee, trachytic-rhyolitic in composition and without macroscopic evidence of magma mingling/mixing (Fa1 and Fa2 samples in ESM 1 and 2; cf. also 27 ), as also recently confirmed by 31 . Historical reports and local oral traditions date this eruptive event at 1810 CE 8,34,35 .
The eruptive fissure affecting the Kone volcanic complex (Kone EF) consists of 5 main én-échelon segments, for a total length of 25 km (Figs. 1, 2). The main fissure (EFKone-1 in Fig. 2B) is ~7 km long and passes through the rims of the two main Kone calderas (Fig. 2B). Our survey has highlighted that the most recent lava flows filling the Kone-Korke calderas were fed by three scoria cones developed along the EFKone-1 fissure system (Fig. 2B), which continues to the south forming two small scoria cones on the southern lower flank of Kone. The southernmost termination of the Kone eruptive fissure lies south of the calderas, in the Kokoro area (Fig. 2B), where an én-echelon segment with offset of ~3 km produced a modest lava flow (Kb1-3 samples in Fig. 2 and ESM 1 and 2), consistent with previous studies 27,29 . The Kone EF is dated at 1810-20 CE 8,23,29,36 .
The third identified eruptive fissure (Koka EF) lies just outside the southern rim of Gedemsa caldera (Fig. 2C). It extends for ~ 7 km and it is arranged in two én-échélon N10°E segments, feeding a mostly aphyric and nonaltered lava that flowed into Koka Lake (Fig. 2C). This eruptive event, coinciding with the Melkasa Unit of 37 , appears distinct from that generating the products emitted within the Gedemsa caldera. Although an absolute age is not available, according to historical reports this eruptive fissure may have formed during 1775 ± 25 CE, when this area was volcanically active 8 .
Whole rock geochemistry. Whole Fig. 5B,D,E; ESM 2). This variability is reflected by most of their trace element contents (e.g., Fig. 6; ESM 2) and their normalized patterns for rare earth and other incompatible trace elements (except for Sr; Fig. 6A,B; ESM 2). With respect to the rock classification for Ethiopian high-Ti volcanic products proposed by 38 , MSC basalts have an almost full fit with the whole rock range compositions of HT1 basalts, except for slightly lower Y values. No variability in whole rock compositions has been observed as a function of the MSC latitude (cf. ESM 2).
The EF products of Metehara and Kone-Kokoro present a more homogeneous composition, being basalts and hawaiites (Fig. 4), whereas Koka samples plot at the boundary between hawaiites and mugearites ( www.nature.com/scientificreports/  www.nature.com/scientificreports/  www.nature.com/scientificreports/ whereas the fissures of the EF may reach up to ~ 14 km; (b) the EF mark the most recent eruptive episode in the NMER; (c) the MSC erupt a lower volume of products (1-2 orders less) than the EF; (d) the MSC magmas typically present a whole rock signature with lower contents of TiO 2 , FeO tot , Na 2 O and P 2 O 5 and higher of Al 2 O 3 than the EF magmas; e) the MSC products are porphyritic, with the mineralogical assemblage being plagioclasedominated, whereas the EF products are almost aphyric (P.I. = 0-5 vol.%). These different features could be either attributed to small changes in the partial melting processes occurring at depth within the rift, producing two different parental magmas, or to distinct conditions experienced by the same parental magma during ascent. In order to discern between these two possibilities, the primary magma compositions for products of MSC and EF have been obtained through a mass balance calculation approach able to reset the modifications of whole rock composition due to crystal fractionation (cf. Methods). Resulting compositions show that MSC and EF basalts derive from parental magmas with similar petrological and geochemical features (overlapped star symbols in Fig. 5). This supports previous studies on primary magma at the MER 39,40 . To obtain primitive compositions for the MSC lavas, addition of at least 27 vol.% of a solid fraction (Pl 10%, Cpx 9%, Ol 6%, Ti-mt 2%) is required. Conversely, magmas feeding the recent EF are in equilibrium with the mantle adding a total solid fraction of at least 20 vol.% of a mafic assemblage (Ol 9%, Cpx 8%, Ti-Mt 3%). The main difference between the two crystallization processes is the presence of plagioclase (partially fractionated, and mostly erupted) only in the MSC magmas, whereas the other mineralogical phases are generally comparable.
Differences between MSC and EF should therefore be related to varying modes of magma storage and ascent. Petrography of MSC and EF products suggests different chemical and physical conditions and processes of magma storage and ascent. The plagioclase dominance in the mildly porphyritic (P.I. 15-35 vol.%) Monogenetic Scoria Cones products suggest storage at the favourable P-T conditions for this mineralogical phase, i.e. in the shallow crust. Conversely, the absence of crystals (P.I. 0-5 vol.%) and the lack of negative Eu anomalies in the www.nature.com/scientificreports/ spider-diagrams of Metehara, Kone-Kokoro and Koka EF (Fig. 6A,B) suggest that these magmas experienced direct ascent from a deep reservoir located below the nucleation depth of plagioclase (~ 10 km for NMER; cf. text below). In the MER, differentiation (mainly crystal fractionation plus assimilation) has been primarily used to explain the silica-rich peralkaline composition of central volcanoes and the "Daly-gap", which is the spatial coincidence of basaltic and rhyolitic products (e.g. 10,18,20,[41][42][43]. Conversely, here we consider the wider process of crustal magma differentiation (since the onset of magma generation), also including the deeper portion of the plumbing systems (19-25 km in the NMER 44 ), where geophysical investigations have detected up to 4% of melt 21,45 .
Simulations of crystal fractionation for EF and MSC magmas are performed through the thermodynamicbased algorithm of Rhyolite-MELTS (cf. "Methods" 46,47 ). Considering that only poorly-differentiated compositions are involved in the two sample suites, our Rhyolite-MELTS modelling is based on two different (and simplified) configurations of magma plumbing systems: (i) one step of deep isobaric crystallization, common to all products, and (ii) one step of intermediate-shallow crystallization. All the simulations start from the calculated primary liquid composition for NMER magmas, which is fairly the same for MSC and EF and is representative for a magma whose composition is not modified by differentiation processes (cf. text above and purple/yellow stars in Fig. 5; ESM 2).
Deep and isobaric crystallization is simulated through Rhyolite-MELTS by means of ca. 250 runs, within the ranges of parameters shown in Table 1. Pressures of crystallization are tested between 350 and 900 MPa, considering adiabatic magma ascent from the source of partial melting to a deep storage reservoir. Oxidation state is tested between QFM-2 and QFM + 1, representing a possible range for MER magmas at different depths 39 . For the initial H 2 O content, we consider a spectrum from a fairly undegassed to a poorly degassed basaltic magma (1-2 wt.%; cf. 42 ). Results show that the model characterized by the parameters listed in Table 1 presents a best-fit with the compositions of EF products (solid lines in Figs. 5B-H). The obtained pressure of 510 MPa, corresponding to ~ 19 km of depth (assuming an average density of 2800 kg/m 348 ) and the H 2 O content of 1.5 wt.% fully match the values provided by the deepest and undegassed olivine-hosted melt inclusions in the NMER 40 . The Quartz-Fayalite-Magnetite (QFM, as the log of fO 2 ) buffer resulted at QFM-2 (reducing conditions). Literature studies (cf. 42 for a review) have considered more oxidizing conditions for the oxygen fugacity (fO 2 ), with QFM between QFM and QFM + 1. Nevertheless, a deeper crystallization may require more reducing conditions. Also, to obtain a liquid line of descent fitting the previously unrecognized high values of TiO 2 (> 4 wt.%) of the EF products, we must inhibit the crystallization of Ti-oxides, trying to maintain Ti in the liquid phase. This occurs when the fO 2 is low, as confirmed by the best fit of EF samples with the liquid line of descent with the considered QFM-2 (Table 1; Fig. 5).
Results suggest that Metehara and Kone-Kokoro EF compositions are explained with an isobaric deep (~ 19 km) crystallization and fractionation of a mafic mineral assemblage (clinopyroxene 11-18%, olivine 10-12%, Ti-magnetite 1-3%), which increases at 43-51 vol.% (clinopyroxene 20-25%, olivine 19-21%, Timagnetite 3-5%) for Koka EF products ( Table 1). The higher fractionation degree of the southern EF of Koka with regard to the other EF products could be attributed to a higher evolution of magma, possibly related to the southward lithospheric thickening and decrease in extension 19 . Furthermore, samples of Koka EF do not perfectly fit with the obtained liquid line of descent (Fig. 5), a feature possibly resulting from their slightly higher crystal content (up to 5 vol.%) with respect to the Metehara and Kone-Kokoro EF products. For example, the lower P-content of Koka EF with respect to the liquid line of descents could be attributed to the fractionation of small amounts of apatite, conversely to the Metehara EF samples, which are strongly enriched in P 2 O 5 , probably for www.nature.com/scientificreports/ unfavourable conditions for the fractionation of apatite. In general, Fo 81 olivine starts to crystallize at 1253 °C, whereas augitic clinopyroxene appears at 1161 °C, decreasing MgO, CaO and FeO (Fig. 5). Therefore, this first modelled configuration of the plumbing system through Rhyolite-MELTS suggests that the scarcity of crystals in EF products can be attributed to an efficient fractionation of mafic crystals from the liquid phase at the base of the crust and direct ascent towards the surface. The second configuration of plumbing system modelled with Rhyolite-MELTS considers a further step of crystallization before reaching the surface. This should explain the extensive crystallization of plagioclase only in Monogenetic Scoria Cones products, allowed by favourable P-T-H 2 O conditions, and the distinct geochemical signature between MSC and EF. Indeed, geochemical whole rock data and the first step of crystal fractionation highlight that MSC compositions are not primary, but underwent a deep (ca. 510 MPa) fractionation of a mafic mineral assemblage before rising (Table 1). Therefore, after experiencing the previously modelled first step of deep crystal fractionation, magma rose adiabatically to mid-crustal levels and started to crystallize. Due to the plagioclase-dominated assemblage, this mid-crustal reservoir should be located above the depth of nucleation of this mineralogical phase. Results of this second step of simulations derive from ~ 200 runs (conditions in Table 1). The starting composition of this new simulation is the liquid composition after nucleation and growth of the three mafic mineralogical phases in the deep reservoir (olivine 12 vol.%, clinopyroxene 6 vol.% and Ti-magnetite 1 vol.%; ESM 2); the relative starting temperature is 1151 °C ( Table 1) Table 1). Therefore, simulations suggest that compositions of Monogenetic Scoria Cones are explained by the sum of a deep crystal fractionation of clinopyroxene, olivine and Ti-magnetite, plus shallow crystallization of a total mineral cargo of 9-42% (clinopyroxene 4-19%; plagioclase 2-12%; olivine 2-6%; Ti-magnetite 1-5%). The scattered compositions of MSC products with respect to the calculated liquid line of descent could be attributed to a non-efficient crystal fractionation in a deep reservoir, resulting in a crystal cargo between 15 and 35 vol.% and also in small whole-rock compositional differences ( Fig. 5; ESM 2). Therefore, this step of crystal fractionation suggests that magmas producing MSC result from an intermediate depth (~ 11 km) storage and crystallization.

Discussion
Monogenetic volcanoes along the MER have been generally considered as deriving from a homogenous tectonomagmatic context, being generally grouped as "basalts s.l. " (e.g. 1,19,27,31,35,37 ). Our results rather show that these volcanoes can be distinguished into two groups leading to different volcanic structures (Monogenetic Scoria Cones and Eruptive Fissures) deriving from the same primary magma but different ascent history and distinct pre-eruptive dynamics (Fig. 8). Our modelling shows that magmas of Monogenetic Scoria Cones (MSC) derive from the superimposition of deep crystallization and fractionation at lower crustal depths (~ 19 km), in common to eruptive fissure (EF) magmas, and a plagioclase-dominated storage at mid-crustal depths (~ 11 km).
Evidence for a mid-crustal magma reservoir for the MSC mainly results from interpreting plagioclase textures and their core-to-rim compositional profiles. Oscillatory zoning in plagioclase crystals is classically related to crystallization under quiet conditions, i.e. without significant variations of P-T-X-H 2 O. In this context, the ΔAn 1-3 oscillations in the MSC plagioclases result from crystallization kinetics at the crystal/melt interface, whereas ΔAn 3-7 oscillations can be attributed to the ponding and movements of plagioclase crystals within a rather chemically and physically homogeneous reservoir [49][50][51] . Conversely, the resorption textures at the core of the plagioclase could be attributed to high decompression rates in water-undersaturated portions of the feeding system, where considerable amounts of water are still dissolved in the melt [52][53][54] . The dominance of oscillatory-zoned plagioclases (> 90% of crystals in our dataset) suggests that MSC magmas crystallized in reservoirs with stable chemical and physical conditions or only with minor P-T gradients. The absence of crenulated or resorbed surfaces and of resorbed rims or sieve-textures in the analysed plagioclase of the MSC, together with After magmas have been generated down in the Astenosphere, during adiabatic ascent they are stored at different crustal levels before reaching the surface. This promoted magma differentiation towards rhyolitic/pantelleritic compositions (cf. 18,20,[41][42][43] ) associated with caldera-forming events. Such plumbing system configuration can be related to immature phases of rifting with lower extension rate and is probably currently represented by the configuration of the plumbing system in the Southern MER. The last caldera-forming event at Fantale is dated at 168 ka 27 . ACF assimilation + crystal fractionation. (B) Between at least 168 ka and approximately two centuries ago the deep magma recharge below the central volcanoes halted, with the shallow magma reservoirs beneath central volcanoes forming solidifying crystal mushes. Volcanic activity occurred only in the monogenetic scoria cones (MSC). The feeding magma, after having undergone crystal fractionation of the mafic mineralogical phases at the base of the crust, rose up and ponded at mid-crustal levels (ca. 11 km) within sill-like reservoirs. Here, plagioclase-dominated crystallization started, with the rise of small pulses of magmas (either through dykes or drops) producing the monogenetic basaltic scoria cones (MSC) in the NMER. (C) The most recent evolution of the NMER (approximately in the last two centuries) has been characterized by linear volcanism associated with eruptive fissures (EF) fed by dikes directly rising from the base of the crust. The shallow plumbing systems below the central volcanoes are completely solidified. www.nature.com/scientificreports/ flat-to-decreasing compositional profiles for An and FeO, indicates that the MSC derive from a "one-shot" intrusion from the crust base, not followed by further magma/gas recharges or rejuvenations. Therefore, MSC magmas rose, with limited crystal fractionation (< 12 vol.%), from the base of the crust and emplaced at ~ 11 km depth, likely feeding sub-horizontal intrusions (as sills; Fig. 8B), adding a further step of (plagioclase-dominated) crystallization. These sill-like intrusions would spatially mimic at depth the high-density MSC volcanic fields indicated in Fig. 1 (black dotted ellipses; Fig. 8B). These hundreds of scoria cones would then derive from the "one-shot" ascent of small magma batches from this mid-crustal reservoir (Fig. 8B). This scenario of stepwise storage is also consistent with previous studies suggesting a ~10 km depth rooting of the feeding structures of monogenetic scoria cones in the MER 55 . Below this depth, the crust is modified by dense gabbroic intrusions (cf. 21 ), likely the solidified magma reservoirs feeding the MSC. A simple and likely driving mechanism for this scenario may be lithospheric thinning and related decompression experienced by the rift axis 19,56,57 . Conversely, magmas producing the recent Eruptive Fissures in the Northern Main Ethiopian Rift do not show evidence of storage at mid/shallow levels of the crust. Their very low crystal cargo (P.I. 0-5 vol.%) suggests the direct ascent of a magma from a reservoir experiencing only recharge from depth and an efficient crystal settling of the denser mineralogical phases (e.g., olivine, pyroxenes, oxides; Table 1; Fig. 8). The lack of plagioclase and of negative anomalies for Eu in the spider-diagrams of Fantale, Kone-Kokoro and Koka EF (Fig. 6A,B) indicates that the main reservoir of EF magma was located below the depth of nucleation of plagioclase, that is here found at ~ 10 km. Geophysical data below the NMER have highlighted a low-velocity zone at 19-25 km, which is the approximate depth of the Moho for Afar-NMER region 58 . Our modelling envisages a plausible deep magma reservoir of the NMER magma plumbing system located at ~ 19 km depth, where magmas are first stored after adiabatic ascent from the source of partial melting (Fig. 8C). Here, magmas producing the EF would experience fractionation of 22-51 vol.% of a mafic solid assemblage (Table 1). These magmas would then rise directly to the surface through dikes influenced by the regional stress field. Rhyolite-MELTS simulations also show that the total amount of fractionated crystals in the EF increases southward, a feature possibly related to the lithospheric thickening and decrease in extension.
As concerns the polygenetic central volcanoes, no evidence of mixing with any magma, as that potentially left in the felsic chambers below them, has been detected in the EF products (Figs. 4,5,6,7). This indicates that the magma rising through dikes from the lower crust is not affected by the composition of these evolved magma reservoirs, even when erupting within a caldera, as at Kone. At the same time, our data confirm that the magma chambers below the polygenetic volcanoes have previously promoted magma differentiation, developing more evolved (i.e., rhyolitic) compositions ( Fig. 6; ESM 2; cf. 10,18,20,[41][42][43]. The apparent contradiction between the filtering role of magma chambers and the lack of differentiation of the EF magmas successively passing through them may be explained only assuming the lack of any residual mush in the former chambers, suggesting their current overall solidification. This suggests that the most recent activity along the MER consists of the sporadic but fast rise of deep and undifferentiated magma through dikes, in a context of overall solidification of the shallow plumbing systems below the central volcanoes (Fig. 8C). A recently detected rhyolitic dike intrusion occurred in 2015 a few km north-east of Fantale supports this scenario. In fact, although attributed to the Fantale magma chamber 33 , we suggest that this dike could represent the evolved shallow residual of the nineteenth century Metehara diking episode.
Our results suggest that the current volcanism along the NMER is mainly controlled by far-field stress related to plate motion, rather than any local contribution from near-field stresses induced by the magma chambers below the polygenetic volcanoes. The likely quiescence of the magma chambers of the polygenetic volcanoes is also consistent with the fact that, along the NMER, no central volcano has been observed to deform or experience unrest 7,8 . Indeed, the location of the EF suggests that the portion of magmatic system nearest to the dominant polygenetic volcano currently accommodates most of the extension, representing the present focus for the rise of magma. Only the slightly more evolved composition of the products of the southernmost EF (south of Gedemsa) suggests a lower amount of extension. Conversely, MSC fields focus in the less active peripheral parts of the magmatic systems, probably experiencing minor extension (Fig. 1).
The recent and rapid rise of transcrustal dikes feeding EF suggests a different tectono-magmatic behaviour with regard to that previously observed during the activity of the central polygenetic volcanoes and the monogenetic vents (i.e., MSC) along the NMER. The latter were both related to a stepwise rise and emplacement of magma. Whether magma was transferred through vertically or horizontally propagating dikes, it commonly emplaced at the base of magma chambers (below polygenetic volcanoes) or at mid-crustal depths (below monogenetic volcanoes). Conversely, the recent fissural volcanism (i.e., EF) is related to a rapid and continuous rise of magma from the base of the crust through vertically propagating transcrustal dikes. While the preferred location for the rise of magma has remained focused below the polygenetic volcanoes, the mode of rise of the Eruptive Fissures suggests enhanced conditions for dike propagation within the NMER. As parental magma feeding the monogenetic EF and MSC volcanism is similar, these enhanced conditions may be related to increased tensile stress promoted by plate motion, which enhanced the transition from central polygenetic to fissural volcanism in the NMER in the last centuries. Alternatively, or in conjunction, these enhanced conditions may be related to a general increase in magma supply recently experienced by this rift portion. Determining whether these possible tectonic and/or magmatic variations are the expression of a stable or a transient condition is challenging. On the one side, an evolving rift should increase its amount and rate of extension, as well as magma supply, promoting dike injection. On the other side, the Eruptive Fissures recently observed in the NMER may represent a transient fluctuation of a longer-term behaviour dominated by the stepwise rise of magma, as for example suggested by the decades-to thousands of years-long fluctuations in the amount of geodetic extension or erupted volumes along the MER 15,17,27,59 . Independent of any stable or transient condition promoting the rise and eruption of the EF magmas, our data suggest an at least medium-term (centuries) significant variation in the delivery of magma to the surface within www.nature.com/scientificreports/ the NMER. This interpretation is supported by the fact that the NMER has passed from dominant central volcanism (polygenetic volcanoes) and stepwise dike propagation (the MSC vents) to transcrustal dike propagation, developing linear fissure volcanism (the EF vents), modulated by the magma-assisted rifting process occurring at the base of the crust. The activity of transcrustal dikes and the solidification of the magma chambers below the central volcanoes explain the apparent contradiction between the widespread presence of magma at depth and its poor evidence at the central volcanoes. These features are in line with recent observations in the more mature transitional crust of Afar, where rapid dike emplacement drives plate opening along the rift axis, as in the Asal-Ghoubbet Rift in 1978, at Dallol in 2004 and at the Manda Hararo Rift from 2005 to 2009 2, [12][13][14]60,61 . This behaviour is in stark contrast with the tectonomagmatic features of the less-extending Southern MER, still largely related to the activity of polygenetic central volcanoes, also testified by their more frequent unrest episodes 7,8 and the much more limited appearance of monogenetic vents and related eruptive fissures. The NMER thus shows that a crucial step in the evolution of continental rifts is the transition from central towards linear transcrustal volcanism, a pre-requisite condition towards plate opening and continental breakup.

Conclusions
Analysis of the monogenetic volcanism along the NMER reveals two different processes of magma ascent from the lower crust. Most of the monogenetic vents (MSC) reveal a stepwise rise of magma to the surface, characterized by storage at middle crustal levels. A fewer longer eruptive fissures (EF) have formed in the last centuries and are characterized by the direct rise of magma, without evolution or contamination, even when passing through older polygenetic volcanoes, highlighting the rapid rise of transcrustal dikes through frozen plumbing systems. It is not possible to determine whether this variation marks the onset of a major change within the NMER or if it represents only a transient stress fluctuation during the reorganization of a rift. In any case, this variation suggests a transition from central polygenetic to linear fissural monogenetic volcanism possibly promoted by increased tensile conditions and modulated by the magma at the base of the crust, consistently with the behaviour of more mature rifts.

Methods
Field analysis and sampling. All data presented here derive from a field analysis performed along the Northern MER in February 2012. The axial portion of the rift has been surveyed from Fantale volcano to the southern termination of Koka Lake (Fig. 1). The survey was accompanied by sampling some representative outcropping volcanic products. On a total of 135 samples, 35  Using satellite images (Google Earth Pro; images acquired in November 2019), 283 monogenetic scoria cones have been identified in the study area, in agreement with previous studies 32 . Among these, 222 lie between Fantale and Boseti volcanoes, whereas 61 lie between Boseti and Gedemsa (cf. Fig. 1 and ESM 1). By using a clustering method similar to that used by 55 , at least three cones aligned along a fracture/fissure and horizontally offset no more than 100 m are considered coeval, or related to the same eruptive episode.
Geochemical and petrologic analytical techniques. Analyses for obtaining whole rock major element data were performed on powder pellets of all the collected samples at the Dipartimento di Biologia, Ecologia e Scienze della Terra of the University of Calabria (Italy) by means of a Philips PW2404 WD-XRF, taking into account the matrix effect corrections. Gravimetric methods were used for obtaining loss on ignition on each sample, and it was then corrected for Fe 2+ oxidation. Major elements for whole rocks are reported in ESM 2.
Whole rock trace element compositions were performed at the SGS Laboratories of Toronto (Ontario, Canada) by means of a Perkin Elmer ELAN 6100 inductively coupled plasma mass spectrometer. The first step of analysis is the fusion (within graphite crucible) of powdered rock samples by Na-peroxide, with consequent dissolution using dilute HNO 3 . Calibration of instruments was performed at the beginning and end of each batch of 5 samples by means of 4 runs on international certified reference materials. Precision and accuracy is better than 5% for all analysed elements. The complete major and trace element dataset for the analysed samples is available in ESM 2.
Polished petrographic thin sections have been prepared on a selected and representative number of 50 samples, on which petrographic and mineralogical analyses were conducted. In order to understand the relationships between the MSC and the more recent EF, on 12 selected plagioclase crystals textural and compositional observations were performed, the last also with the execution with step of analysis varying from 5 to 10 μm, depending on the crystal size (ESM 3). Textural observations were instead performed by means of high-contrast back-scattered electron (BSE) images. Both investigations were accomplished at the Dipartimento di Scienze Biologiche, Geologiche e Ambientali of the University of Catania (Italy) with a Tescan Vega-LMU scanning electron microscope coupled with an EDAX Neptune XM4-60 EDS micro-analyser equipped with an ultra-thin Be window and an EDAX WDS LEXS spectrometer (wavelength dispersive low energy X-ray spectrometer) calibrated for light elements. Operating conditions for obtaining high contrast BSE images were set at 20 kV (accelerating voltage) and ~ 8 nA (beam current) and 20 kV and 2 nA for the analysis of major element abundances. Analyses www.nature.com/scientificreports/ were repeated, during the analytical runs, on internationally certified standards (An-rich plagioclase and glass), giving a precision in the order of 3-5% and an accuracy of 5%.
Primary magma modelling. In order to obtain the potential primary liquid related to the products of the monogenetic activity, the PRIMELT3 MEGA.XLSM algorithm was used 62 . This provides a mass balance-like solution, based on the olivine addition to the composition of a basic lava until a primary magma composition is reached. As an input to PRIMELT3, the starting compositions have been chosen among the most basic and less porphyritic lavas of each dataset [FK3 (P.I. = 1 vol.%) for MSC and ME14A (P.I. = 12 vol.%) for EF; ESM 2]. Considering these samples as representative for the differentiated liquid phase, resulting compositions are shown in Fig. 5 (purple and yellow stars) and Table 1. In order to have a double check on the obtained primary compositions, a mass balance calculation approach has been also used. Any process of crystal fractionation into the magma reservoir has been reset by adding variable amounts of solid fractions of olivine, clinopyroxene, plagioclase, orthopyroxene and oxides to the same most basic and less porphyritic lavas within each investigated suite, until the resulting crystallizing olivine reaches a composition in equilibrium with the in mantle (with inferred Fo 88-90 ). The added mineral compositions are among the most basic of the entire dataset (ESM 4).

Rhyolite-MELTS modelling.
Rhyolite-MELTS is a thermodynamic-based software 46 Table 1. The obtained LLDs have been then plotted in the compositional diagrams together with the investigated sample suits, in order to choose the model for which the natural compositions are reproduced. The interval of temperature for the simulation has been set at 2 °C, as larger intervals would have influenced the resolution of fractional crystallization models 42,47 .