... (2016). The eruptive history and magmatic evolution of Aluto volcano: new insights into silicic peralkaline volcanism in the Ethiopian rift.

ThesilicicperalkalinevolcanoesoftheEastAfricanRiftaresomeoftheleast studiedvolcanoeson Earth.Herewe bring together new constraints from ﬁ eldwork, remote sensing, geochronology and geochemistry to present the ﬁ rstdetailedaccountof theeruptive historyofAluto, arestless silicicvolcanolocatedinadenselypopulatedsec- tion of the Main Ethiopian Rift. Prior to the growth of the Aluto volcanic complex (before 500 ka) the region was characterized by a signi ﬁ cant period of fault development and ma ﬁ c ﬁ ssure eruptions. The earliest volcanism at Aluto built up a trachytic complex over 8 km in diameter. Aluto then underwent large-volume ignimbrite erup-tionsat316± 19kaand 306 ± 12 kadeveloping a ~ 42 km 2 collapsestructure. Afterahiatus of~250 ka,a phase of post-caldera volcanism initiated at 55 ± 19 ka and the most recent eruption of Aluto has a radiocarbon age of 0.40 ± 0.05 cal. ka BP. During this post-caldera phase highly-evolved peralkaline rhyolite lavas, ignimbrites and pumicefalldeposits haveeruptedfrom ventsacrossthe complex.Geochemicalmodelling isconsistentwith rhyolite genesis from protracted fractionation ( N 80%) of basalt that is compositionally similar to rift-related basalts found east of the complex. Based on the style and volume of recent eruptions we suggest that silicic eruptions occur at an average rate of 1 per 1000 years, and that future eruptions of Aluto will involve explosive emplace- ment of localised pumice cones and effusive obsidian coulees of volumes in the range 1 – 100 × 10 6 m 3 .


Introduction
Quaternary volcanism in the Ethiopian Rift has been marked by the eruption of hundreds of cubic kilometres of highly evolved silicic magmas (e.g., Mohr, 1971;Di Paola, 1972;Barberi et al., 1975;Lahitte et al., 2003;Peccerillo et al., 2003Peccerillo et al., , 2007Field et al., 2013). Silicic volcanoes have produced extensive rhyolitic lavas as well as pyroclastic density current (PDC) and tephra fall deposits; many show evidence of large (N10 km 3 ), geologically young (b 500 ka), caldera-forming eruptions (e.g., Shala, Mohr et al., 1980;Gedemsa, Peccerillo et al., 2003and Kone, Rampey et al., 2010 and several show signs of unrest (e.g., Corbetti, Aluto, Bora and Haledebi; Biggs et al., 2011;. Many of these volcanoes are located in densely populated regions (Fig. 1A), and it is estimated that over 10 million people in Ethiopia live within 30 km of a Holocene volcanic centre (Brown et al., 2015). Our understanding of the eruptive histories of the silicic complexes is poor and hence there are large uncertainties as to the volcanic hazards they may pose in the future (Aspinall et al., 2011).
The focus of this study is the Aluto volcanic complex (Fig. 1B), a silicic peralkaline volcano which lies in the central sector of the Main Ethiopian Rift (MER), and for which few details of its eruptive past are known. Aluto is currently restless, having undergone episodic ground deformation for at least the past decade (Biggs et al., 2011;. This is of particular concern given that several thousand people live on the volcano as well as in the nearby towns of Ziway, Adami Tullo and Bulbula (each with populations in the range of 10,000-50,000, Central Statistical Agency of Ethiopia, 2012), and according to the Smithsonian Global Volcanism Program (Siebert and Simkin, 2002), 6.8 million people live within 100 km of the volcano. Aluto also hosts Ethiopia's only geothermal power plant (e.g., Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996), and while the plant is only producing 7 MW at present there are plans for significant expansion. Locating this major infrastructure on a volcanic edifice where eruptive frequency is unknown poses inherent risks (e.g., Wilson et al., 2012Wilson et al., , 2014. important resources (e.g., obsidian) that hominin populations have worked and transported since the earliest Palaeolithic, 1.7 million years ago (e.g., Piperno et al., 2009).
In this paper, field studies, remote sensing, geochronology and geochemical techniques are integrated to document the volcanic and magmatic evolution of the Aluto volcanic complex. We reconstruct the eruptive history of Aluto and present a new geological map (Section 5); we present the first dedicated whole-rock geochemical study of the volcano, and assess magma petrogenesis (Section 6), and finally, we develop a conceptual model of the eruptive history and magmatic evolution of the complex within the context of the evolving rift (Section 7).

Regional Tectonic and Volcanic Overview
The MER is an active continental rift that forms the northernmost section of the East African Rift System (see reviews by Ebinger, 2005;Corti, 2009). It is a classic example of an oblique rift  and is traditionally sub-divided into northern, central and southern segments that reflect prominent spatial variations in fault architecture and lithospheric characteristics (e.g., Hayward and Ebinger, 1996;Bonini et al., 2005;Keranen and Klemperer, 2008;Agostini et al., 2011). Aluto lies in the central MER (CMER) where the pattern of faulting is characterized by two distinct groups: (1) boundary faults located on the margins of the rift, and (2) localized internal faults, mapped as 'Wonji' faults (Agostini et al., 2011;Keir et al., 2015). The boundary faults formed during the initial stages of rifting (at 8-6 Ma in the CMER, WoldeGabriel et al., 1990;Bonini et al., 2005) and appear to facilitate very limited tectonic extension at present Pizzi et al., 2006;Agostini et al., 2011). The Wonji faults, located in the~15 km wide axial zone of the rift, are short, right-stepping en-echelon features that developed after~2 Ma (Boccaletti et al., 1998;Ebinger and Casey, 2001) and are the focus of present-day tectonic strain (Keir et al., , 2015Agostini et al., 2011).
There has been surface volcanism throughout the development of the MER (WoldeGabriel et al., 1990Abebe et al., 2007) but full understanding of the links between volcanic evolution and rift development is still hampered by a lack of detailed eruption chronologies for many individual volcanoes (Corti, 2009). Prior to development of the silicic volcanoes exposed along the rift axis today (including: Gedemsa, Bora-Berrecio, Aluto, Shala and Corbetti, Fig. 1B), several large silicic caldera-forming volcanoes were active in the MER during the Early Pliocene-Early Pleistocene (4-1 Ma, Fig. 1B). Tuff deposits from these volcanoes are often found in fault sections on the margins of the rift (e.g., WoldeGabriel et al., 1990WoldeGabriel et al., , 1992, while the associated caldera structures are deeply eroded or otherwise completely covered by younger basin-filling sediments (WoldeGabriel et al., 1990;Le Turdu et al., 1999). Three of the largest silicic caldera-forming volcanoes (Munesa, Awassa and Gademotta) formed within the axial region of the CMER (Fig. 1B), and are now extinct. At Munesa caldera climactic eruptions are dated at~3.5 Ma (WoldeGabriel et al., 1990(WoldeGabriel et al., , 1992Le Turdu et al., 1999), at Awassa caldera major eruptions took place at 1.85-1.1 Ma (WoldeGabriel et al., 1990) and at Gademotta (~20 km NW of Aluto) the final phase of activity is dated to~1.3 Ma (Vogel et al., 2006). Volcanism in the MER, prior to 1 Ma, was diffuse and encompassed a wide footprint of the rift zone (Bonini et al., 2005;Corti, 2009;Maccaferri et al., 2014), as evidenced by off-axis trachytic and basaltic volcanic centres (e.g., Chilalo, Hunkuolo, Kaka, Kubsa, Fig. 1B; Mohr and Potter, 1976;WoldeGabriel et al., 1990).
There are limited accounts of historical volcanic activity in the MER and the only reported eruptions occurred at Fantale and Kone volcanoes in the 1800-1830′s (Williams et al., 2004;Rampey et al., 2010). While satellite remote sensing has detected ground deformation at various silicic complexes, including Aluto (Biggs et al., 2011;, it is important to underscore that not one of the MER volcanoes is routinely monitored on the ground (Aspinall et al., 2011). Therefore, at present, to assess the likelihood of future eruptions in the MER, volcanologists must rely on remote sensing to detect magma movement in the subsurface, as well as geochronology to constrain the rates of past eruptive activity.

Previous Studies of Aluto
The earliest geological descriptions of Aluto come from regional scale mapping reports (e.g., Dakin and Gibson, 1971;Di Paola, 1972). These studies identified that the bulk of recent volcanic products of Aluto are rhyolitic PDC deposits and obsidian coulees. Di Paola (1972) recognized that various craters and lava flow deposits exist along predominantly NNE-SSW orientations, parallel to the Wonji fault structures, while Dakin and Gibson (1971) postulated that there had been a calderaforming event at Aluto based on field evidence for a~1 km long segment of the caldera wall visible on the north-east of the complex.
In 1970 the Aluto-Langano region was identified as a site for geothermal development and eight exploratory wells were completed between 1981(ELC Electroconsult, 1986. Geological mapping was undertaken at the same time (Kebede et al., 1985), and lithological descriptions were also completed for deep wells LA-4 and LA-8, see Tables S1 and S2, respectively (Yimer, 1984;Mamo, 1985). Using the cores and cuttings collected during deep well drilling, several previous publications (e.g., Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996), and drilling reports (Yimer, 1984;Mamo, 1985;ELC Electroconsult, 1986;Teklemariam, 1996), have established a deep stratigraphy of the complex. In Fig. 2 we summarize the Aluto Fig. 1. A) Map of Ethiopia and neighbouring countries overlain with population density. Red lines indicate major plate boundaries and grey triangles show Quaternary volcanic centres (after Siebert and Simkin, 2002). Population data are estimates for 2015 provided by Socioeconomic Data and Applications Center (SEDAC), http://sedac.ciesin.columbia.edu. Population values are extrapolated based on a combination of national growth rates from United Nations statistics and sub-national growth rates from census dates. The blue rectangle marked B, shows the area covered by the Digital Elevation Model (DEM) in Fig. 1B. B) Hillshade Satellite Radar Topography Mission DEM showing the location of the Aluto volcanic complex (marked A) within the central segment of the Main Ethiopian Rift (MER). Quaternary silicic volcanic complexes are shaded red, and include: Corbetti (marked C), Shala or O′a caldera (marked S); Bora-Bericcio (marked B) and Gedemsa (marked G). Recent mafic cinder cones and lavas of East Ziway (marked EZ) and the Butajira-Silti field (marked BSF) are shown by the dark grey shaded areas. Silicic volcanic complexes that became extinct prior to 1 Ma are outlined by the blue dashed lines, and include: Hawasa (marked HAW), Munesa (marked MUN), Gademotta (marked GAD), Kubsa (marked KU), Kaka (marked KA), Chilalo (marked CH), Hunkuolo (marked HK) and the Galama Range (marked GR). The beige shaded area identifies the maximum extent of the Ziway-Shala lake basin that developed in the Mid-Late Pleistocene (Le Turdu et al., 1999;Benvenuti et al., 2002Benvenuti et al., , 2013. The major present-day lakes are identified as: Ko: Lake Koka; Zw: Lake Ziway; Ln: Lake Langano; Ab: Lake Abijta, Sh: Lake Shala and AW: Lake Awasa. White stars identify centres of population with N10,000 inhabitants. Current extension of the MER, after Keir et al. (2006), Bendick et al. (2006) and Saria et al. (2014), is indicated by the black arrow with motion relative to a fixed Nubian plate. The black and blue outlined areas on the globe inset correspond to the regions covered in Fig. 1A and B, respectively. deep well stratigraphy and also overlay major structural features recognized by Hutchison et al. (2015). Lithological descriptions of the units are provided in Section 5.1 and Fig. 3.
The only previous interpretations of the eruptive history of Aluto were made by Kebede et al. (1985) andELC Electroconsult (1986). They each proposed a period of sub-aerial basaltic eruptions (Bofa Basalts, at 1.6 ± 0.5 Ma, Table 1) prior to the first silicic volcanism. Silicic eruptions at Aluto then built up an early dome complex, culminating in several large ignimbrite eruptions and caldera collapse. K/Ar dates (cited in Teklemariam et al., 1996) suggested that the climactic phases of eruption took place at~155 ka (Table 1). Subsequent investigations of Aluto have focussed on the geothermal field (e.g., Gebregzabher,  Hutchison et al., 2015). Note that the transect line is shown in Fig. 1B. Well data represent the synthesis of several publications (Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996) and drilling reports provided by the Geological Survey of Ethiopia (Yimer, 1984;Mamo, 1985;ELC Electroconsult, 1986;Teklemariam, 1996). The geological units shown have been correlated between the different wells on Aluto (Gizaw, 1993;Gianelli and Teklemariam, 1993) and indicate a prevailing mode of deposition rather than a single homogeneous unit (e.g., paleosols occur within the Bofa Basalt, see Section 5.1 for detailed descriptions). Red arrows indicate main upflow zone of geothermal reservoir along the Artu Jawe fault zone (Gianelli and Teklemariam, 1993;Teklemariam et al., 1996;Hutchison et al., 2015. Red 'F′ labels indicate major fumarole and degassing regions mapped at the surface of the volcano (after Hutchison et al., 2015).  . 4) with the lithological sequence reconstructed from deep well cores and cuttings (Yimer, 1984;Mamo, 1985;ELC Electroconsult, 1986;Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam, 1996;Teklemariam et al., 1996). Colours of the mapped surface volcanic units correspond with the shading in the geological map in Fig. 4. The relative thickness of each unit corresponds to the height of the coloured rectangle. Ages marked to the left of the stratigraphy are from our new 40 Ar/ 39 Ar determinations or radiocarbon ages unless indicated otherwise by numerical superscripts. Literature ages are from: 1 : Teklemariam, 1996 Teklemariam, 1993. 1986;Valori et al., 1992;Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996;Saibi et al., 2012) and the outline eruptive history proposed by Kebede et al. (1985) andELC Electroconsult (1986) has not been re-evaluated within the last forty years.

Sampling and Analytical Techniques
Our stratigraphy was established from a total of 8 weeks of field mapping conducted during 2012-2014. Further, the Natural Environment Research Council's Airborne Research and Survey Facility (NERC-ARSF) acquired high-spatial resolution topography (lidar), hyperspectral data and aerial photographs of the complex on 16th November 2012. Digital geological mapping was carried out in ArcGIS and involved analysis of orthorectified aerial photos, the lidar DEM as well as Google Earth and ASTER satellite imagery for regions beyond coverage of the lidar DEM. Volcanic and tectonic fault structures were mapped by Hutchison et al. (2015) using remote sensing techniques and soil-CO 2 degassing surveys. The volumes of recent volcanic deposits (obsidian coulees and pumice cones) were estimated using the lidar DEM following the methods of Nomikou et al. (2014). In brief, we use polygons to outline and mask out the deposit from the DEM. After removing this region from the DEM we then interpolated a smoothed near-flat pre-eruption surface across the masked area (processing was carried out in GMT software using a variable tension parameter, Nomikou et al., 2014). We then subtracted the present-day DEM from the interpolated DEM to calculate the residual volume between the surfaces, thus providing a volume estimate for the volcanic deposit. An example of this technique is given in Fig. S1. At Santorini volcano, lava flow maximum and minimum volumes were calculated by Nomikou et al. (2014) and typically varied by b 25%; we use this value as an estimate Table 1 Compilation of new and existing age constraints on volcanism at Aluto. 40 Ar/ 39 Ar ages shown are the weighted mean ± standard error of the mean (1σ), and in brackets the arithmetic mean ± one standard deviation (we quote the latter throughout the manuscript). Errors on other ages are quoted as originally published. Radiocarbon dates were calibrated with IntCal13 (Reimer et al., 2013) and OxCal v.4.2. Calibrated radiocarbon ages (2σ confidence interval) represent the age range of the deposits below and above the tephra layer, respectively. 11.12 ± 0.11 to 7.29 ± 0.06 cal. ka BP Benvenuti et al. (2013) pumice bracketed by gravels, youngest tephra at Haroresa of the volume uncertainty of this method. Finally, as the internal structure of the silicic domes on Aluto is poorly constrained we simply present the total volume of the deposit rather than a dense rock equivalent. We obtained 40 Ar/ 39 Ar age determinations at key points in the stratigraphic sequence (Fig. 3). Unweathered lava and bulk ignimbrite samples were collected in the field and pulverised in a jaw-crusher. Sanidine phenocrysts (250-500 μm in diameter) were separated from the crush using magnetic as well as lithium metatungstate heavy liquid density separation techniques. The phenocrysts were then leached ultrasonically in 5% HF for~3 min to remove adhering glass and groundmass. Grains were rinsed in distilled water and dried, followed by handpicking under a binocular microscope to remove any remaining contaminant phases (e.g., fluid/melt inclusions, quartz or granophyric textured crystals). Samples were irradiated in the Cd-lined facility at the Triga Reactor, Oregon State University, and subsequently analysed at the Natural Environment Research Council (NERC) Argon Isotope Facility, Scottish Universities Environmental Research Centre (SUERC). Five 40 Ar/ 39 Ar ages and one radiocarbon age determined for Aluto were reported by Hutchison et al. (in press), who used these ages to examine silicic magma fluxes in the MER. In this study we use these existing ages, plus an additional two 40 Ar/ 39 Ar ages, to examine stratigraphic relationships and the rates of recent volcanism at Aluto. A limited number of geochronological data have been presented previously for Aluto; these are shown alongside our new 40 Ar/ 39 Ar ages in Table 1.

Sample
A total of 36 lava and pumice samples were selected for whole-rock chemical analysis. Samples were trimmed of weathered material, and powdered in an agate ball mill. Major and selected trace element analysis was conducted at the Department of Geology at the University of Leicester by X-ray fluorescence (XRF) using a PANalytical Axios-Advanced XRF spectrometer (glass fusion beads were prepared for major elements, and powder pellets were prepared for trace elements). A complete suite of trace elements were also analysed by Inductively Coupled Plasma-Mass Spectroscopy (ICP-MS) at the Department of Earth Sciences, University of Oxford. For this, solutions of powdered samples were produced in a multi-stage method, involving initial hot, low-pressure digestion in HF and HNO 3 , followed by multiple drying and dissolution steps using HNO 3 , before final dilution in HNO 3 . Standard analyses for both XRF and ICP-MS methods were within 10% of their reference values. This project has also compiled previously published element data for the Aluto volcanic complex (from Di Paola, 1972;Yimer, 1984;Mamo, 1985;Teklemariam, 1996). Both previous and new bulk major and trace element data, a total of 45 samples, are presented in Table S3a-d.

Volcanic Stratigraphy
We have divided the stratigraphic relations as follows: Section 5.1 presents the deep well stratigraphy compiled from previous publications; Section 5.2 presents our field stratigraphy from surface mapping, and Section 5.3 presents age constraints and correlations between the surface and subsurface stratigraphy.

Deep well stratigraphy
Previous publications on the geothermal field of Aluto (Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996) and drilling reports provided by the Geological Survey of Ethiopia (Yimer, 1984;Mamo, 1985;ELC Electroconsult, 1986;Teklemariam, 1996) have been used to establish a composite deep stratigraphy for Aluto (Fig. 3, righthand column). In the following sections the unit name of previous authors is shown first, followed by our revised classification in brackets (justified in Section 5.3).
5.1.1. Neogene Ignimbrites (Neogene-Quaternary undifferentiated ignimbrites, Nqui) The base of all deep wells is represented by a sequence of rhyolitic ignimbrites and lavas at least 400-500 m thick (Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996;Figs. 2, 3). Welded crystal-rich ignimbrites and fine-grained rhyolites dominate the sequence but volumetrically subordinate basaltic lava layers are also observed (Yimer, 1984;Mamo, 1985). The ignimbrite horizons are generally reported to be vesicular and crystal-rich, with a mineral assemblage of sanidine, quartz, magnetite and amphibole . The ignimbrites represent the main reservoir for the geothermal field (Fig. 2) and have undergone extensive hydrothermal alteration . Note that previous authors commonly refer to this unit as 'Tertiary Ignimbrite'.

Lacustrine sediments (Quaternary deep well lacustrine sediments, Qdl)
Stratigraphically above the Bofa Basalts (Nqub) is a sequence of lacustrine sediments, predominantly mudstones and siltstones, that have been identified in all wells except LA-4 and LA-5 (Fig. 3, Teklemariam et al., 1996). The lacustrine sediments show spatial variations in thickness, becoming progressively thinner to the east of the Aluto complex (Gizaw, 1993;Gianelli and Teklemariam, 1993;Teklemariam et al., 1996;Hutchison et al., 2015). The maximum thickness of the lacustrine sediments is~400 m in well LA-2 (Teklemariam et al., 1996). In the centre of the complex they have a thickness of~80 m in well LA-8 (Mamo, 1985, Table S2) and 2 km east of this (traversing a major fault zone at Artu Jawe, Fig. 4) lacustrine sediments are absent in deep well LA-4 (Yimer, 1984, Table S1) and LA-5 (Gianelli and Teklemariam, 1993;Teklemariam, 1996). This pattern suggests considerable topographic offsets existed during lacustrine sediment deposition. Given the rift setting, it is assumed that deposition of the lacustrine sequences took place in a faulted environment and that the greatest thicknesses of sediments were established in the most downthrown blocks on the west side of the complex (Hutchison et al., 2015).
A number of wells on Aluto show trachytic units preceding and interbedded with the lacustrine sediments (Figs. 2, 3). Core and cutting samples reveal that these units are composed of grey, variably consolidated trachytes (including tuff, lava and breccia) with phenocrysts of  Kebede et al. (1985) as well as maps of Quaternary lacustrine sediments presented by Benvenuti et al. (2002Benvenuti et al. ( , 2013. Note the apparent lack of outcrop of units Qgei and Qcr is due to their limited exposure at the base of deep gorges, which are invariably covered by thick sequences of younger units. Outcrops labelled with a '?' were not visited during our field campaign and have been categorized on the basis of surface geological descriptions by Kebede et al. (1985); future mapping should aim to verify the relationship between these outcrops and the new stratigraphy. Local names that are referred to in text and subsequent figures are italicised. Faults have been mapped previously by Agostini  alkali feldspar (Mamo, 1985). Intense hydrothermal alteration has led to calcite, chlorite and smectite precipitation in these units (e.g., Teklemariam et al., 1996) greatly complicating interpretation of the original rock texture and their relation to the lacustrine sequences.
The lacustrine sediments and overlying ignimbrites (Section 5.1.4) are highly altered and rich in clay minerals (e.g., chlorite, illite and smectite) . It is likely that these units provide a low porosity and low permeability clay cap layer that seals the Aluto-Langano geothermal reservoir ( Fig. 2; Teklemariam, 1996).

Peralkaline rhyolites (Qcr-Qpoy)
The uppermost units found in the Aluto wells are peralkaline rhyolites (Figs. 2, 3), and include obsidian lavas, finely crystalline lavas and pumice breccias (Tables S1 and S2). There are no detailed descriptions of the sub-units within the peralkaline rhyolite sequence, and at present it is not possible to correlate sub-units (e.g., individual lava flow deposits) between the wells and assess lithological or thickness variations. Descriptions of well LA-4 by Yimer (1984) provide the most comprehensive account of the lithological sub-units within the peralkaline rhyolite sequence (Fig. 3, Table S1), but future work to systematically log these horizons is essential. Above the ignimbrite and trachytic tuff sequences (Section 5.1.4), Yimer (1984) identified flow banded comenditic rhyolites in a core cut at 256-266 m. Unfortunately, no cores were taken across the~30 m transition from the major ignimbrites to the comendites, Table S1, leaving considerable uncertainty as to the nature of the contact between these units (Fig. 3, discussed in Section 8.2). Cuttings suggest that superposed on the comenditic rhyolites are~50 m of pumiceous deposits and then~30 m of aphyric obsidian lava (Yimer, 1984). The aphyric obsidian is then succeeded by~40 m of sparsely porphyritic rhyolite lavas and finally, the upper most section comprises~30 m of pumiceous breccias (Fig. 3).

Field stratigraphy
In the following sections we establish the field relations between the volcanic units exposed at the surface. Section 5.2.1 describes units that were identified off the main volcanic edifice in faulted sections around the margins of the Aluto complex; representative of the deepest accessible stratigraphy in the study region. Section 5.2.2 describes the succession of units identified on the main volcanic edifice, providing a complete sequence from the base to the top of the edifice. Finally, Section 5.2.3 describes the main sedimentary units as well as undifferentiated volcanic deposits. The key accompanying figures are: the synthetic stratigraphy of the surface volcanic units (Fig. 3, left-hand columns); the new geological map of Aluto (Fig. 4); graphic logs of key stratigraphic sections (Fig. 5) and annotated field photographs (Figs. 6 and 7). In addition, representative geochemical data for each of the main units are given in Table 2.  (Kebede et al., 1985). Blocks of basalt are identified as rare lithic clasts within PDC deposits on the main edifice suggesting that these units underlie the Aluto complex.

Quaternary trachytic tuffs and lavas (Qdt).
Comenditic trachyte tuffs and lavas (Qdt),~100 m thick, are exposed in fault scarps east of Lake Langano (at Mt. Dima, Figs. 4 and 5M). The trachytes are composed of both porphyritic and aphanitic types, and display varying degrees of vesiculation. Bright red blocky breccias are found between the layers of trachyte lava (Figs. 5M, 7A), suggesting that there were considerable hiatuses and periods of reworking between emplacement of the major lava flow units.
5.2.1.3. Quaternary grey welded ignimbrites (Qgyi). The next unit in the stratigraphic sequence is a package of strongly welded grey ignimbrites (Qgyi) that contain abundant lithic fragments and elongate fiamme (composed of partially devitrified obsidian). The welded ignimbrites (Qgyi) overlie the trachytic lavas (Qdt) at Mt. Dima (Fig. 4), although the contact was not clearly exposed (Fig. 5M). For the majority of Qgyi exposures only the grey densely welded zone of the ignimbrite is visible at the surface. Kebede et al. (1985) reported that in more complete sections the ignimbrites have a moderately compacted and poorly welded pumiceous base that grades upward into the densely welded zone before transitioning back into a non-welded pumiceous top. The thickness of this unit in well-exposed sections near Haroresa ( Fig. 4) is~50 m (Kebede et al., 1985).

Quaternary green welded ignimbrites (Qgei).
Deposits of a welded green ignimbrite (map unit Qgei) are found in an uplifted fault block on the west of Aluto (Figs. 4,5B). Qgei is composed of fine-grained pistachio-green matrix with dark grey fiamme and accessory lithic rock fragments (Fig. 7B). The stratigraphic context of Qgei is difficult to resolve owing to limited exposure; however at the type locality west of the edifice ( Fig. 5B) it is immediately overlain by lacustrine sediments (siltstones, diatomite and reworked tephra) suggesting that it was emplaced in a lake basin. We frequently identified Qgei (and also Qgyi) as accidental lithic clasts within younger PDC deposits on the central edifice, suggesting that these two major ignimbrite units underlie the complex, and are considerably older than the volcanic units identified on the main edifice (Qcr-Qpoy, following section).

Quaternary aphyric obsidian (Qao).
Overlying Qcr in Gebiba gorge is a~10 m thick sequence of pumice, ash and breccia units (Fig. 6A). The contact between Qcr and the overlying volcaniclastic sequence is poorly exposed but taken to be unconformable (Fig. 5K). The volcaniclastic beds are themselves overlain by an obsidian lava flow deposit (Figs. 5K, 6A), which has a blocky aphyric obsidian surface and basal breccia (each~2 m thick), and a~7 m thick flow core comprised of finely crystalline rhyolite (Fig. 5K). The aphyric obsidian lavas and underlying bedded pumice are classified together as Qao. Qao lavas are dominantly exposed on the south-east slopes of the Aluto edifice where they cover an area of~15 km 2 and extend for N7 km away from the centre of the edifice (Fig. 4).

Quaternary porphyritic obsidian (Qpo)
. The next phase of activity is represented by sparsely porphyritic obsidian lavas (Qpo), identified in Fig. 5. Key stratigraphic sections used to constrain the eruptive stratigraphy (Fig. 3). Correlations are shown by the shaded colours that connect the sections. The location of each section is shown by the stars on the geological map (Fig. 4). Ages shown alongside the stratigraphy were determined by 40 Ar/ 39 Ar methods (Table 1 and Section 5.3). The bulk grain size of the volcaniclastic fragments in the deposit is given by the scale at the base of each section: A corresponds to ash, L to lapilli and B to blocks and bombs. Figs. 5 (sections C, D, G, H and I) and 6B. A major dome of the Qpo unit is developed on the northern rim of the complex, where lavas extend 2.5 km north of the volcanic vents (Fig. 4). Many of the Qpo lavas extend from breached craters suggesting that explosive phases preceded the lavas (although no underlying volcaniclastic units were exposed).
Obsidian lava flows Qpo and Qao are ubiquitously covered by soil horizons and tephra. Although we were unable to observe a direct contact between Qpo and Qao units it is evident from our field observations that soil horizons are thicker and more developed above Qao compared with Qpo (contrast Fig. 7C with Fig. 6B). We infer that Qao units were erupted earlier than Qpo lavas as shown in our stratigraphy in Fig. 3.

Quaternary west Aluto intermediate lavas (Qwai).
Eruption of intermediate lavas is restricted to the west of the edifice (Qwai, Fig. 4). The main occurrence of Qwai is a small trachyandesite scoria cone, 65 m high and~600 m diameter, south of the main access road to Aluto from Adami Tullo (Figs. 4, 5A and 6C). The scoria cone is covered by a thin layer of pumice (Figs. 5A, 6C) suggesting it developed prior to the most recent explosive eruptive phases (Qup, next section).

Quaternary undifferentiated pumice (Qup).
Covering all previously mentioned deposits of Aluto is an undifferentiated pale grey pumice deposit, Qup (Fig. 4, numerous sections in Figs. 5 and 6A-C). The thickness of the tephra varies across the complex from a few centimetres up to several meters where it has experienced reworking on steep slopes. The tephra comprises lapilli-pumice, and has rare chips of obsidian and hydrothermally altered material (b10 mm in diameter). Qup ubiquitously overlies palaeosols (e.g., Figs. 5, 7C) indicating that this unit marks the initiation of a major explosive eruption sequence. Field exposures (Figs. 6B,7C) suggest that this unit is a single pumice fallout deposit, and likely to be the product of a (sub) Plinian eruption, but further work needs to be done to verify this. 5.2.2.6. Quaternary pumiceous breccias youngest (Qpby). Pumiceous breccias (Qpby) are encountered above the Qup tephra at Humo gorge and Artu Jawe (Figs. 5F, G and 6B, D). Qpby units were erupted from discrete vents across the complex and were not erupted in a single event (although some ignimbrites appear to have closely followed Qup with little evidence for any substantial hiatus, e.g., Fig. 6B). Qpby units have been emplaced as both PDCs (Fig. 6B, D) with run outs of several kilometres and localized pumice cones (Fig. 5D). The pumice cones comprise coarse (up to 300 mm diameter), clast-supported pumice exposures that have a mantle-bedded morphology (e.g., Golba gorge, Fig. 5D). Pumice cones are typically between 10 and 100 m in height, and may extend to~1 km from the erupting vent; effusive obsidian coulees (e.g., Qpoy, next section) commonly follow, breaching the walls of the pumice cone.
Qpby PDC units, such as those identified at Humo gorge (Fig. 5F), show characteristic peralkaline ignimbrite textures with large pumice Photographs showing field relations of key units. A) Trachytic lava breccias and tuffs (Qdt) exposed in uplifted fault blocks south-east of the Aluto edifice (Fig. 5M). B) Green welded ignimbrite (Qgei, from section Fig. 5B) with collapsed pumiceous fiamme, hammer for scale 0.30 m in length. C) Contact between aphyric obsidian lava (regolith) and light grey pumice unit (Qup) with~1 m thick bright orange palaeosol developed (Fig. 5J). D) Section of explosive eruptions units, suggestive of phreatomagmatic processes (Fig. 5C). E) Hummocky 'blister' structures developed within a welded PDC deposit (Qpby, near Fig. 5F). F) Fluvially reworked tephras and volcaniclastic material deposited on the southern slope of the edifice, the large channel cutting down through the stratigraphy is typical of a lahar unit with a coarse lithic base load of obsidian cobbles. G) Detail inset from F), showing thin channels of reworked tephra (likely derived from units Qup and Qpby). Photographs F and G were taken~1 km south-west of stratigraphic section 5 K (see Fig. 4).
Overall the Qpby units represent a mixture of volcaniclastic deposits that were either built up around erupting vents (pumice cones), or formed density currents during particularly explosive phases. Their general evolution and morphology is comparable to the silicic lava cones and shields described by Mahood and Hildreth (1986) 6D) and so they are easily distinguished as the most recent eruptive products (Fig. 3). Qpoy lavas range from~200 m to~2.5 km in length, and show marked differences in surface weathering, suggesting that they were not emplaced in a single eruptive event. Qpoy vents commonly coincide with volcanic or tectonic structures, such as the caldera ring fault or the Artu Jawe fault zone that runs NNE-SSW east of the caldera floor (Fig. 4, Hutchison et al., 2015).

Undifferentiated volcanic units (Qutc and Nquw).
North of the Aluto edifice we identify numerous tuff cones (map unit Qutc), around the shore of Lake Ziway (Fig. 4), and to the east of Aluto, we have mapped scoria cones and basaltic fissure lavas occurring in the Wonji fault belt (Nquw, Fig. 4). Given their proximity and comparable surface weathering to the volcanic deposits of Aluto we assume that these formed contemporaneously to the Aluto edifice; however, their precise relations to the proximal volcanic stratigraphy have yet to be constrained.

Pleistocene and Holocene lacustrine units (Qlp and Qlh).
Lacustrine sediments (predominantly siltstone, sandstone and diatomite) deposited in the Ziway-Shala basin are found north, south and west of the main edifice (Fig. 4, Gasse and Street, 1978;Street, 1979;Le Turdu et al., 1999;Benvenuti et al., 2002Benvenuti et al., , 2013. While previous authors (e.g., Benvenuti et al., 2002) have defined several unconformity-bounded units that record major stages in the lake evolution, we simply delineate the main boundary between Pleistocene and Holocene sediments (Qlp and Qlh, respectively) in our geological map (Fig. 4). The Holocene high-stand of the Ziway-Shala basin (between 10 and 5 ka) was 1670 m above sea level Table 2 Geochemical overview of the main eruptive units. Major element concentrations are in wt.%, trace element concentrations are in ppm. Major elements were measured by XRF, trace elements were measured by either XRF or ICP-MS, as indicated by 1, 2 in the unit row, respectively. For the TAS subdivision row, P and C correspond to pantellerite and comendite respectively. TA stands for trachyandesite. The symbols represent ages from Teklemariam (1996) ( †), Le Turdu et al. (1999) (+) and this study (*). n.d. stands for no data. For SO 3 b.d. indicates below detection limits. LOI: loss on ignition. NK/A: is an indication of peralkalinity and was calculated as molar (Na 2 O + K 2 O)/Al 2 O 3 (e.g., Macdonald, 1974 (Benvenuti et al., 2002), and therefore sediments in the plains around Aluto that are below this height level are assumed to be Holocene in age. Conversely, sediments above 1670 m height are correlated to the Late Pleistocene Megalake phase deposited during the last glacial period,~100-22 ka (Benvenuti et al., 2002(Benvenuti et al., , 2013. The deep gullies west of Aluto that drain into the Bulbulla river are an exception to this, where~50 m thick sequences of Pleistocene lacustrine sediments have been described by Gasse and Street (1978), Street (1979), Benvenuti et al. (2002) and Ménard et al. (2014).
5.2.3.3. Reworked volcaniclastic sediments and colluvial-fluvial sediments (Qal and Qc). There is abundant evidence for fluvial reworking of volcaniclastic material on Aluto (map unit Qal). On the flanks of the main edifice, drainage networks are well developed and have reworked volcaniclastic material in both narrow streams (Fig. 7G) and large channels (potentially episodic lahars, Fig. 7F). The caldera floor of Aluto is a self-contained basin that has been infilled by reworked volcanic deposits derived from the rim of the complex (mostly rounded pumice lapilli as well as sand-silt size lithic fragments). It is possible that Aluto hosted a caldera lake during periods of humid climate, and this feature may explain the phreatomagmatic characteristics of PDC deposits reported in unit Qpby (Section 5.2.2). East of Aluto undifferentiated terrigenous clastic sediments (gravel, sandstone, siltstone and mudstone) were identified (map unit Qc). These units are representative of colluvial and fluvial sedimentary facies deposited on the faulted margins of the Ziway-Shala lake basin.

Age constraints and correlations between the surface and deep stratigraphy
New and existing age constraints on volcanism are given in Table 1, and shown alongside our stratigraphic interpretations in Fig. 3. These units are discussed from the base to the top of the pile. The ignimbrite units (Nqui) found at the base of the deep wells lack robust age constraints, as K/Ar analysis of altered feldspars yielded a minimum age of 1.4 Ma (WoldeGabriel et al., 1990). Nevertheless, previous authors (e.g., WoldeGabriel et al., 1990WoldeGabriel et al., , 1992Teklemariam, 1996) have correlated the Aluto deep well ignimbrites (Nqui) with petrologically and geochemically analogous ignimbrites found near Munesa (Fig. 1B), which yielded ages of 3.51 ± 0.03 Ma (Munesa Crystal Tuff), 2.9 ± 0.3 Ma (Munesa Vitrophyre) and 2.30 ± 0.03 Ma (Neogene Ignimbrite). These correlations should be viewed with caution until further geochemical evaluation and age analysis is undertaken.
The next unit is Nqub, and on the on the basis of their comparable geochemistry, thicknesses and emplacement style, we correlate the surface exposures of basaltic lava (Nqub) with the Bofa Basalt unit described in the deep wells (Fig. 3, Table S1,2 and Gianelli and Teklemariam, 1993). The age of Nqub is constrained at 1.6 Ma ± 0.5 Ma by a K/Ar determination from a porphyritic basalt lava on the eastern flank of Aluto (Table 1, Teklemariam, 1996).
No direct ages are available for lacustrine sequences encountered in the deep wells (Qdl). These sequences appear to represent the oldest lacustrine sediments deposited in the Ziway-Shala basin (Fig. 1B), and probably have an age of~570-330 ka (i.e., the maximum age range for the basin, Le Turdu et al., 1999).
Trachytic tuffs and lavas found at Mt. Dima (Qdt) are tentatively correlated with the trachytes identified in the deep wells (Fig. 3), on the basis of their similar deep position in the stratigraphy and their broad compositional similarity. Direct ages for trachyte lavas are still to be established, however their close association with lacustrine sediments (Section 5.1.3) would suggest that their ages are roughly equivalent (around~570-330 ka), and indeed they must have erupted beforẽ 310 ka, the age of the overlying ignimbrites. The welded ignimbrite sequences, Qgyi and Qgei, yield overlapping 40 Ar/ 39 Ar ages of 316 ± 13 ka and 306 ± 12 ka (Table 1) and we correlate these units with the thick ignimbrite and tuff sequences described in the deep wells (Section 5.1.4, Fig. 3).
For the units on the main edifice of Aluto, our 40 Ar/ 39 Ar results support ages of the comenditic rhyolites (Qcr) at 54 ± 13 ka, the sparsely porphyritic obsidian coulees (Qpo) at 62 ± 13, 22 ± 14 and 19 ± 5 ka, and the youngest obsidian coulees (Qpoy) at 16 ± 14 ka ( Table  1). West of Aluto, in the deep gorges of the Bulbula plain, Gasse and Street (1978) and Street (1979) identified numerous volcaniclastic horizons interbedded with lacustrine sediments and dated these using radiocarbon methods. The largest volcaniclastic sequence, informally known as the Abernosa pumice, was deposited between~26 and 14 cal. ka BP, and preserves~13 distinct tephra layers (Gasse and Street, 1978). The youngest tephra layer to be identified by Gasse and Street (1978) shows a broad age overlap with the youngest tephra found south-east of Aluto at Haroresa (Benvenuti et al., 2013, Fig. 4), be-tween~11 and~6 cal. ka BP ( Table 1). As these deposits represent the youngest widely dispersed tephra we correlate these with Qup, the most recent pumice fall unit that we could identify on the main edifice (Section 5.2.2, Fig. 4). This correlation would suggest that the youngest eruptive products (Qpoy and Qpby) are younger than~10 ka (Section 8.3), which is consistent within error of the 40 Ar/ 39 Ar for Qpoy. Finally, a radiocarbon age for charcoal found immediately beneath a thin bed of pumiceous pyroclastic deposits (20-30 cm thick) on the west of the Aluto edifice (Hutchison et al., in press) provides an age of 0.40 ± 0.05 cal. ka BP for the youngest explosive eruptions on the complex (unit Qpby, Fig. 3).
The ages also constrain Qao lavas between 74 and 49 ka and the Qwai scoria cone to N 10 ka (i.e., older than overlying Qup, Figs. 3, 6C). In Fig. 3 probable links between the surface volcanic deposits on the main edifice and the peralkaline rhyolites in deep wells are shown, and were made on the basis of comparable lithological descriptions.

Petrography
Basaltic lavas from Aluto (Nqub and Nquw) are vesicular and variably porphyritic (0-40 vol.% phenocrysts) with holocrystalline matrix textures (Fig. 8A). The phenocryst assemblage is dominated by plagioclase, with subordinate olivine, clinopyroxene and Fe-Ti oxides (Fig.  8A). The minerals found in the matrix are equivalent to the phenocryst assemblage. Trachyandesite lavas (sampled from the scoria cone, Qwai, Fig. 6C) are highly vesicular, and aphyric to scarcely phyric, with microphenocrysts predominantly represented by plagioclase feldspar (Fig. 8B). Within the trachyandesite scoria cone (Qwai, Section 5.2.2) small (b 10 mm) fragments of silicic rocks (possessing a granular texture, Fig. 8C) as well as partially resorbed xenocrysts of alkali feldspar (Fig. 8B, C) and aenigmatite are found mantled by the trachyandesite groundmass. The bulk geochemical composition of these silicic enclaves is comparable to the peralkaline rhyolites of Aluto (see Figs. 9-11, red star). The enclaves likely represent partially disaggregated xenoliths of intrusive silicic rocks or crystalline mush that was entrapped by the trachyandesite melt prior to eruption.

Major elements
Rocks were classified using the total alkalis-silica (TAS) diagram (Fig. 9A, after Le Maitre et al., 2002). Note that in all geochemical plots (Figs. 9-12) we have adopted a colour scheme that differentiates the main eruptive units following the interpretations that were made in Section 5, and which we expand on further in Section 7. The major element data shows that the majority of volcanic rocks sampled at the surface of Aluto are of rhyolitic composition (Fig. 9A). Overall, there is a large range in silica content across the sample suite (SiO 2 43-76 wt.%) and there are representatives of each major compositional division, although rocks of intermediate composition are least abundant.
The silicic rocks of Aluto are represented by trachytes (Qdt), and rhyolites (all units on the main volcanic edifice, Qcr-Qpoy, as well as the welded ignimbrite units, Qgyi and Qgei). Silicic rocks were further classified as pantellerites and comendites according to the FeO t -Al 2 O 3 diagram (Fig. 9B) after Macdonald (1974). Trachytes (Qdt) are While the rock appears green in field photographs (Fig. 7B) in thin-section the matrix appears orange-brown in colour and is composed of a very fine grained groundmass that has likely devitrified and altered significantly. Fiamme are visible in the lower centre of image and are composed of lenses of silicic material that are coarser than the groundmass. Alkali feldspar, the dominant crystal phase, show a large range in sizes and in some instances have trapped earlier crystallizing phases (note embayed alkali pyroxene inclusion top right). (E) Porphyritic obsidian (Qpoy, 01-02-13) sampled from the southern rim of the Aluto edifice (shown in plane polarised light). In the centre of the image the main phenocryst phases of aenigmatite (opaque to deep red or brown colour), alkali pyroxene (dark green) and alkali feldspar (low relief and colourless) can be identified. Note that the pyroxene is embayed and appears to be overgrown and replaced by the aenigmatite. The glassy matrix contains aligned microlites of alkali feldspar and alkali pyroxene. (F) Porphyritic obsidian (Qpo, 18-01-08) sampled from the north-west margin of the main Aluto edifice (shown in cross polarised light). A granophyric textured crystal (intergrowth of alkali feldspar and quartz) is visible on the left while a regular alkali feldspar crystal is shown on the right. The rectangular crystal form suggests that quartz is nucleating on the alkali feldspar. Granophyric textures are envisaged to form along the cooled walls and roof zone of the magma reservoir (e.g., Lowenstern et al., 1997). comenditic, welded ignimbrite units, Qgyi and Qgei, show pantelleritic and comenditic compositions respectively, and the bulk of post-caldera samples are pantelleritic rhyolites (the only exception is the comenditic rhyolite Qcr, found at the base of the main edifice, Section 5.2.2).
Two rhyolite pumice samples (17-01-01 K and 17-01-01G, Table  S3) showed anomalously low Na 2 O values (2.8 wt.%) and high loss on ignition (N5 wt.%) compared to other pumice samples (Table S3a); these samples are distinct outliers on the TAS diagram (Fig. 9A). Previous studies also identified anomalous low Na 2 O values in seemingly pristine pumice samples (e.g., Peccerillo et al., 2003;Fontijn et al., 2013). These authors suggest that post-emplacement alteration processes (e.g., leaching by surface water interaction) may be the causal factor. Aluto samples that showed Na 2 O loss were sampled from isolated pumice cones on the north-western periphery of the complex; these would have been located at the edge of the lake during lacustrine high stands (Gasse and Street, 1978;Le Turdu et al., 1999;Benvenuti et al., 2002, Fig. 4) and this may explain the alteration through interaction with surface waters. In all geochemical plots subsequent to Fig. 9A these altered samples have been excluded.
Major elements (Fig. 10) have trends characteristic of protracted fractional crystallization processes controlled by removal of olivine, clinopyroxene, plagioclase, Fe-Ti oxide, quartz, alkali feldspar and aenigmatite (in approximate order of appearance). TiO 2 , FeO t , MgO and CaO all decrease with increasing SiO 2 (i.e., they behave compatibly with increasing fractionation). K 2 O has a smooth linear increase. Al 2 O 3 has values of~15 wt.% in the mafic to trachytic units but decreases to 8-12 wt.% in the rhyolite units, highlighting the onset of plagioclase feldspar as the major fractionating phase after~67 wt.% SiO 2 . Na 2 O shows a generally increasing trend from mafic through to trachytic samples, and then considerable variation (5-7 wt.%) in the rhyolites. The variation in Na 2 O after~72 wt.% SiO 2 reflects the dominance of alkali feldspar as a fractionating phase (Section 6.1). P 2 O 5 shows a general decrease with increasing SiO 2 but with a significant inflexion at~55 wt.% SiO 2 (P 2 O 5 behaviour likely reflects late fractionation of apatite; although rare it is likely to be an accessory phase throughout the crystallizing sequence, e.g., Field et al., 2013).

Trace elements
Selected trace elements and incompatible element ratios are presented in Fig. 11. The bulk of our new trace element data were determined by ICP-MS, although for two samples (13-05-04: unit Qgei, 15-02-09: unit Qdt) certain incompatible element values were above the analytical range at the time of measurement (Table S2c). In these cases we have plotted the equivalent trace element concentration determined by XRF (Table S3b). For elements analysed by both methods results compared well (b 10% deviation between ICP-MS and XRF values for identical samples) and are within the uncertainty of measurements (Section 3). We have also included trace element data for three samples from Teklemariam (1996) in Fig. 11. Trace element concentrations of these samples (S1: unit Nqui, S2: unit Nqub and S3: unit Qgyi, Table  S2d) were determined by XRF methods and because identical samples and standards were not run between our and their analysis we have been cautious not to interpret anything beyond the broad geochemical trends.
Sr does not change significantly from basaltic through to the intermediate and trachytic lavas but shows a marked depletion in rhyolites (Fig. 11); indicative of high degrees of feldspar removal. Ba shows an increase from basalts to trachytes before showing a marked inflexion at 700 ppm Zr due to the appearance of alkali feldspar as a major fractionating phase in the rhyolites.
Incompatible elements (Y, Nb, La, Rb and Zr) reveal a near continuous fractionation sequence of Aluto lavas, the only significant gap occurs between Zr values of 450-700 ppm where only one representative trachyte sample exists. Incompatible-incompatible element diagrams (Fig.  11, Y, Nb, La and Rb against Zr) show smooth linear positive trends that pass through the origin, and are consistent with Aluto's evolved pantellerites being derived from a mafic parent with similar chemistry to the pre-Aluto (Nqub) or Wonji lavas (Nquw). The post-caldera rhyolites cover a considerable range in incompatible element values and the most evolved samples are from the grey pumice unit Qup (Section 5.2.2) which has 1650 ppm Zr. Notably, the comenditic rhyolite unit (Qcr), the earliest recognizable phase of post-caldera volcanism, is offset from the main fractionation trend for certain incompatible trace-  (Le Maitre et al., 2002). The grey dashed line shows the alkaline-sub-alkaline divide of Irvine and Baragar (1971). Two rhyolite pumice samples, discussed in text, show high loss on ignition (LOI) values and low Na 2 O values suggestive of postemplacement alteration. These samples have been removed from subsequent plots. (B) Classification diagram of peralkaline rhyolites and trachytes (Macdonald, 1974). Qcr corresponds to the comenditic rhyolite sampled from the base of the Aluto stratigraphy (the earliest post-caldera phase, Sections 5.2.2 and 5.3). elements (e.g., Y and La against Zr, Fig. 10). An ignimbrite sampled from a fault scarp near Munesa (Fig. 1B) by Teklemariam (1996) that we classify as a Pre-Aluto ignimbrite (Nqui) also falls off the main linear array evidencing a different fractionation trend from the Aluto sample suite.
Ratios of incompatible elements such as La/Y and Rb/Nb (Fig. 11) do not show significant variations within the sample suite, and indeed the post-caldera rhyolites (excluding Qcr) show an exceptionally constant ratio. The Rb/Nb ratio of Precambrian crustal rocks, which represent a likely component of the basement rock, are much higher than any samples from Aluto (Fig. 11). We follow previous arguments of Peccerillo et al. (2003) and suggest that partial melting of these crustal rocks would tend to increase the large ion lithophile element and high field strength element ratios (LILE/HFSE, e.g. Rb/Nb in Fig. 11) in the melt, making it extremely unlikely that the peralkaline magmas were derived from crustal anatexis alone (discussed further in Section 8.1).

Examining fractional crystallization processes
Major and trace element trends are indicative of fractional crystallization (Sections 6.2 and 6.3). To assess the viability of this process in generating the Aluto sample suite we adopt the simple approach of Blundy and Wood (1991), and model evolving Sr concentrations in a melt governed by feldspar fractionation. From our petrographic observations (Section 6.1) it is clear that feldspar is the main phenocryst phase found throughout the sequence, and assuming that the system is dominated by feldspar fractionation then the progressive change in composition from anorthite to albite to anorthoclase (and the corresponding increase in the feldspar-melt partition coefficient, D Sr ) should explain the Sr whole-rock trends (e.g., Blundy and Wood, 1991).
In the model Zr is taken to be completely incompatible (in line with trace element observations in Section 6.3 and the lack of zirconiumbearing minerals observed petrographically) and is used as a proxy for melt fraction (F). The Wonji basalt sample (Nquw) which is poorly evolved and least enriched in incompatible trace elements (161 ppm Zr) is our best estimate of a parental magma, and we use Zr parent / Zr sample as an indicator of the fraction of liquid remaining. In the model we also assume that the partitioning of Sr between plagioclase feldspar and silicate melt (D Sr ) as a function of anorthite (An) content and temperature, as described by Blundy and Wood (1991), is the only process controlling Sr evolution in the melt. For each calculation we crystallize 10% of the residual liquid, and the temperature and An content of plagioclase are adjusted at each step (as shown in Fig. 12 A   Fig. 10. Harker variation diagrams of whole-rock compositions determined by XRF. All concentrations are shown in wt.%. Key as for Fig. 9. Qcr corresponds to the comenditic rhyolite sampled from the base of the Aluto stratigraphy (the earliest post-caldera phase, Sections 5.2.2 and 5.3). and B). The temperature changes linearly from 1150°C to 650°C with fraction of liquid remaining (Fig. 12A), while the An content of plagioclase varies from An 90 to An 20 (Fig. 12B). These assumptions are based on petrological constraints of magma temperatures and mineralogy variations from Dabbahu volcano (Ethiopia) after Field et al. (2013); the only peralkaline volcano in the region where these detailed constraints exist. Four different fractionation trends are considered (Fig. 12C-E) each representing a constant percentage of plagioclase in the crystallizing assemblage (30%, 50%, 70% and 90% are shown). The resultant change of the bulk partition coefficient (D Sr ) is shown in Fig. 12C, and modelled melt Sr evolution is compared to the Aluto whole-rock data in Fig. 12D, E.
The Aluto whole-rock Sr values broadly match the melt evolution predicted by the models (Fig. 12D, E), although there is an obvious gap in samples between the trachytes and the least evolved rhyolite (i.e., the silicic enclave sampled from the scoria cone sequence, shown as the red star). The key observation from Fig. 12D is that it is possible to link back the peralkaline rhyolite Sr concentrations to the least evolved basaltic lavas via feldspar fractionation and this lends support to fractional crystallization being the dominant process in controlling melt evolution at Aluto. The models suggest that the total amount of fractional crystallization required to generate the peralkaline rhyolites from parental basaltic lavas is N 80%, and that feldspar must be the dominant phase in the crystallizing assemblage (≥50%) in order to generate the rhyolites (in agreement with petrological observations made in Section 6.1). A key implication of this model, and the trace element diagrams in Fig. 11, is that degree of differentiation has changed through time. The welded ignimbrites are more chemically evolved than the trachytes, while the post-caldera magmas have undergone even more extreme fractionation. In Section 8.1 we consider petrogenesis and melt evolution processes at Aluto further, as well as future work that will be necessary to test our hypotheses.

Evolution of the Aluto Volcanic Complex
Using our new stratigraphic and age constraints we present a conceptual model outlining the development of the Aluto complex within the context of the evolving rift (Fig. 13).

Pre-Aluto ignimbrite units (4-2 Ma)
Rifting of the CMER in the Pliocene and into the earliest Pleistocene (before 2 Ma) was characterized by displacement along~50 km long, widely spaced, NE-SW border faults (Bonini et al., 2005;Corti, 2009;Fig. 11. Whole-rock trace element variation diagrams plotted against Zr (upper six plots) and Rb (lower two plots). All concentrations are shown in ppm. These represent whole-rock values determined by both XRF and ICP-MS methods (see text for discussion of compatibility). Key as for Fig. 9. Qcr corresponds to the comenditic rhyolite sampled from the base of the Aluto stratigraphy (the earliest post-caldera phase, Sections 5.2.2 and 5.3). Qup corresponds to the pumice fall deposit (Section 5.2.2). Precambrian basement trace element data are from Sidamo, Southern Ethiopia (Peccerillo et al., 1998). Agostini et al., 2011). Volcanic activity was diffuse and silicic complexes were developed across a wide footprint of the rift (e.g., Mohr and Potter, 1976;WoldeGabriel et al., 1990). The oldest deposits identified in the Aluto deep wells are thick sequences of silicic ignimbrites (Nqui, Fig.  3). Hydrothermal alteration of these deep well units is extensive and greatly complicates thorough geochemical correlations and geochronological analyses (WoldeGabriel et al., 1990). The Nqui ignimbrites are presently constrained to be N1.4 Ma in age (Table 1), and in line with field correlations made by WoldeGabriel et al. (1990) and Teklemariam et al. (1996) we speculate that they derive from voluminous explosive eruptions at large axial volcanic complexes of Munesa (WoldeGabriel et al., 1990(WoldeGabriel et al., , 1992 and Gademotta (Laury and Albritton, 1975;Vogel et al., 2006), as well as the off-axis complexes (Mohr and Potter, 1976) which were all active within this time frame (Fig. 13A).
The N 500 m thick sequences of sub-aerial basalt and trachybasalt units (Nqub, Figs. 3, 8), dated at~1.5 Ma (Table 1), coincide with this major shift of deformation within the rift valley. Our interpretation is that these mafic lavas and scoria deposits relate to fissure eruptions occurring along the rift axis in tandem with Wonji fault development (Fig.  13B, analogous to present day rift-related volcanism in Afar, Ferguson et al., 2010). The great thickness of the mafic units as well as the occurrence of intervening palaeosol layers, suggests that Nqub sequences were generated from numerous eruptive events over a protracted timespan from around 1 Ma to ca. 100 ka.

Basin development and growth of the trachytic edifice (500-310 ka)
Following the mafic fissure eruptions, deep wells on the west of Aluto record lacustrine sediments accumulating within the fault controlled Ziway-Shala basin (which formed between 570 and 330 ka, Le Turdu et al., 1999, Fig. 13C). A number of the deep wells also show trachytic tuffs and lavas interbedded with the lacustrine sediments (Section 5.1.3, Fig. 3). The deep well trachytes are correlated with the trachytic units found at Mt. Dima (Qdt, Figs. 3, 4, 5M), and mark the onset of silicic volcanism in the region (i.e., they represent the earliest eruptive products of the Aluto volcanic complex). Our interpretation is that trachyte lava flows and tuffs built up a low relief silicic complex (or lava shield) upon the faulted rift terrain. Trachytic lava piles, as envisaged here (Fig. 13C), appear to commonly form the earliest growth stages of peralkaline volcanic edifices in both the Ethiopian and Kenyan rift systems (e.g., Shala: Mohr et al., 1980, Kone: Rampey et al., 2010, Emuruangogolak: Weaver, 1977Macdonald, 2012, Menegai: Leat et al., 1984Macdonald et al., 1994, Olkaria: Clarke et al., 1990Fig. 12. Melt models to predict Sr concentration in a feldspar-dominated crystallizing assemblage using Zr as a proxy for melt fraction. We assume that the poorly evolved Wonji basalt (Nquw), which is least enriched in incompatible trace elements, is representative of a parental magma. The modelled temperatures, anorthite-content and the calculated bulk partition coefficient (D Sr ), are plotted in A-C, respectively. For each increment we crystallize 10% of the residual melt relative to the previous step. We consider four models where proportion of plagioclase as a percentage of total phenocrysts is held constant at 30, 50, 70 and 90%. D) Variation in Sr content of Aluto samples during fractionation (coloured points), compared to the modelled melt evolution trends (shown as lines where each point represents 10% crystallization of residual melt relative to the previous step). Data are plotted using the coloured symbols (see Fig. 9 for legend). E) Shows inset from D with a linear vertical scale. Caldera formation at peralkaline volcanoes is typically linked to explosive eruptions that generate widespread tuff sheets (e.g., Green Tuff, Pantelleria, Mahood, 1984;Mahood and Hildreth, 1986;Williams et al., 2013). At Aluto, the remnant caldera wall structure (Fig. 4), as well as constraints from soil-CO 2 degassing and vent alignments (Hutchison et al., 2015) all suggest that the complex has undergone at least one caldera-forming eruption. The welded ignimbrite units Qgei and Qgyi have overlapping 40 Ar/ 39 Ar ages of 316 ± 13 ka and 306 ± 12 ka (Fig. 3, Table 1), and represent the best candidates for widely dispersed, explosive ignimbrite sheets associated with a classic peralkaline caldera collapse (Fig. 13D). While there is insufficient field evidence to assess how these deposits physically relate (e.g., whether the ignimbrites represent one or multiple eruptive events), their componentry (large obsidian and pumice fiamme, of diameter N 100 mm), thicknesses (N10 m) and dense welding suggest that they are the proximal deposits of large-scale ignimbrite-forming eruption(s).

Post-caldera activity (b60 ka)
Volcanic activity resumed at Aluto after~60 ka (Fig. 3), and has been marked by the eruption of rhyolite lavas flows, pumice fallout and PDCs from vents largely confined to the main edifice (Fig. 13F). Our mapping allowed us to classify four distinct eruptive sequences, each bounded by palaeosols that mark pauses in volcanic activity (Fig. 3). The earliest phase consists of the comenditic rhyolite lavas (Qcr) which erupted at 55 ± 19 ka ( Table 1). The next phase is represented by bedded pumice deposits and aphyric obsidian lavas (Qao). No ages have been determined for Qao, but this unit is stratigraphically bracketed between 74 and 49 ka. Subsequent eruptions comprise a series of sparsely porphyritic obsidian lava flows (Qpo) with ages of 62 ± 13 ka, 22 ± 14 ka and 19 ± 5 ka ( Table 1). The most recent eruptive phase likely initiated after 10 ka (Section 5.3) and relates to volcanic units Qup-Qpoy. The youngest eruption of Aluto took place several hundred years ago at 0.40 ± 0.05 cal. ka BP (Fig. 3, Table 1).
Silicic eruption cycles for the post-caldera phases appear to initiate with an explosive eruption building small pumice cones and/or emplacing PDC deposits; these are then followed by effusive eruptions comprising obsidian or finely crystalline rhyolite lavas (e.g.,Figs. 5E,K and 6A,D). This cyclic activity is consistent with a scenario where volatile rich magmas accumulate in the roof of the magmatic reservoir, overpressure triggers eruption of a gas-rich pumice units which are then followed by degassed obsidian coulees (in line with typical models of rhyolite lava dome emplacement from elsewhere, e.g., Mono Craters, California, Fink (1980), and Pantelleria, Mahood and Hildreth, 1986).
While we have classified four post-caldera eruptive phases, the field evidence from the youngest pumice breccias and obsidian lavas (Qpby and Qpoy), as well as the 40 Ar/ 39 Ar ages for Qpo units (Fig. 3, Table 1), demonstrate that individual lavas are erupted from discrete vents across the complex over a prolonged period of time. Therefore the mapped eruptive phases represent a sequence of rhyolite lava dome events, rather than a single eruption from a single vent. Volcanic vent locations, particularly for the youngest obsidian coulees (Qpo and Qpoy) are commonly linked to a structural control imposed by the underlying caldera ring fault or tectonic faults (Hutchison et al., 2015). Small volume intermediate eruptions (b 0.01 km 3 ) have occurred on Aluto but are restricted to the faulted zones on the western flank (Figs. 4; 12F). Phreatomagmatic eruptions are also a common feature of volcanism in the last 60 ka (Fig. 7D), these would have occurred around the flanks of the volcano at the shoreline of the Ziway-Shala lake, but also potentially within the caldera when rising magma interacted with water stored in a caldera lake or geothermal reservoir.
It has been argued that crustal assimilation may play a significant role at a number of other peralkaline centres (Davies and Macdonald, 1987;Black et al., 1997;Bohrson and Reid, 1997;Trua et al., 1999), however, based on characteristic trace element signatures (e.g., Rb/Nb Fig.  11) we suggest there is no requirement for any significant contribution from crustal melting to form the post-caldera rhyolites at Aluto. A more stringent test for crustal assimilation at Aluto will require additional radiogenic isotope analysis (Sr-Nd-Pb) and would be greatly complemented by a better understanding of lithospheric structure and composition in the CMER (e.g., Cornwell et al., 2010). Indeed at Fantale and Gedemsa volcanoes (also located in the MER), Giordano et al. (2014) found isotopic evidence for minor (~2%) crustal assimilation in mafic magmas as well as low temperature contamination of rhyolites by hydrothermal/meteoric fluids. It is important to recognize that while our results suggest fractional crystallization processes dominate at Aluto (Figs. 9-12), minor assimilation and contamination cannot presently be ruled out.
Understanding how the silicic melt reservoirs of the MER were assembled and the timescales over which this occurred is a major challenge, but nonetheless essential if we wish to achieve a complete synthesis of the magmatic processes that have operated during rift evolution. At Aluto, a great thickness (N500 m) of mafic lavas (Nqub) were erupted regionally prior to the formation of the silicic complex (Fig.  13B), and our trace element geochemistry (Section 6.3) and modelling (Section 6.4, Fig. 12) leads us to hypothesize that the mafic lavas are part of the same magmatic lineage as the peralkaline rhyolites. At present geochemical analyses of primitive mafic lavas from Aluto are extremely limited (Figs. 9-11) and further sampling, ideally complemented by melt inclusion studies, will be necessary to verify how representative our choice of parental melt composition is (Section 6.4).
Linking the Nqub basaltic lavas and the peralkaline rhyolites of Aluto as a single lineage, would suggest they were both generated from the same mantle-derived basaltic melt. We envisage a scenario where basaltic parental melt flux out of the MER mantle reservoir was focused into the axial volcanic segments of the rift zone (Section 7.2). Mantlederived basaltic melt intruded within the volcanic segments could be tapped into the Wonji fault plumbing systems (Rooney et al., 2007) and erupt regionally (e.g., Nqub or Nquw), or alternatively could stall and fractionate, ultimately allowing evolved melts to accumulate in the upper crust. While studies from other rift zones and extensionaltranstensional settings (e.g., Taupo Volcanic Zone, Spinks et al., 2005, and the Altiplano-Puna Plateau in the Central Andes, Acocella et al., 2011) have demonstrated strong correlations between extension rate, magma storage volumes and eruption rates the limited available age and volume controls on MER volcanism currently hamper similar riftscale comparisons and our understanding of the tectonomagmatic processes that generated the silicic magma systems. It has, however, been suggested that the silicic volcanoes of the MER are located at the ends of the volcanic segments (e.g., Mohr et al., 1980;Casey et al., 2006) and it is commonly assumed that complex extensional stress field (Keranen et al., 2004) and/or the cooled crust abutting the segment tip (Ebinger et al., 2008) hindered magma ascent at these sites. An alternative view is that once the upper crustal reservoir and the volcanic edifice began to establish then the stress regime could have acted to focus magma ascent in the crust, essentially capturing later melts beneath the volcanic edifice (Karlstrom et al., 2009). In any case we suggest that the evolved silicic melts of Aluto represent the stalled, structurally focused component of mantle-derived basaltic melts. Further structural and geochronological investigations will be vital for understanding the links between tectonics and the rise and emplacement of magma across and along the MER.
The earliest silicic eruptions from Aluto were trachytic (Qdt, Fig.  13C), and it is plausible that the peralkaline melts that would later form the major ignimbrites (Qgyi and Qgei) were extracted from the trachytes (e.g., Rooney et al., 2012). Incompatible trace elements and modelling (Figs. 11 and 12) indicate that trachytic lava can be produced via~70% fractional crystallization from the least evolved (parental) basaltic lavas found in the study area, while welded ignimbrites require N80% fractionation.
The large silicic melt reservoir (pre-requisite for units Qgei and Qgyi) was likely to have been assembled during an episode of high heat and mass input (e.g., Tappa et al., 2011;Frazer et al., 2014;Macdonald et al., 2014). The elevated rates of magma and heat supply may have limited the degree of differentiation, and hence incompatible element concentrations in the pre-and syn-caldera samples (e.g., Y, Nb, La, Rb and Zr, Fig. 11) do no approach the high values seen in many of the recent post-caldera products. A similar temporal trend is observed at Longonot (Macdonald et al., 2014) where periods of strong convection in the magma reservoir in the build up to caldera formation limit any extreme compositional variations. By~310 ka a large volume of melt-and volatile-rich magma had formed at shallow crustal levels beneath Aluto and the major caldera-forming eruptions could be initiated (Fig. 13D). The substantial time gap of~250 ka between the welded ignimbrites and post-caldera phases (Fig. 13 E, Section 7.5) suggests that there was a substantial withdrawal of eruptible magma from the upper crustal chamber.
The first post-caldera phase of Aluto (comenditic rhyolites, Qcr) are less evolved than the major welded ignimbrite units and fall off the fractionation trend for certain trace elements (e.g., La and Y in Fig. 11). It is has been demonstrated at a number of other peralkaline systems that the first post-caldera phases consist of mixed rhyolites and trachytes with mafic material from lower in the reservoir (e.g., at Menegai, Leat et al., 1984 andLongonot, Macdonald et al., 2014). While the data presented here do not allow us to thoroughly address this issue we suggest that dedicated microanalytical and isotopic studies of the comenditic rhyolites (Qcr) will be key to identifying evidence for melt mixing and contamination.
The post-caldera rhyolites display a large range of incompatible element concentrations and the pumice unit Qup, which likely represents a Plinian eruption, has the most chemically evolved composition (Fig. 11). The extreme enrichments in incompatible element concentrations after 60 ka may be explained by formation of a stratified shallow magmatic reservoir with a compositionally zoned cap (e.g., Mahood, 1981;Leat et al., 1984;Mahood and Hildreth, 1986;Civetta et al., 1988;Macdonald and Scaillet, 2006;Neave et al., 2012;Macdonald et al., 2014). In this case the most evolved compositions (e.g., Qup) would be generated following periods of extensive crystal fractionation and/or decreasing melt supply to the shallow magmatic reservoir, which would have driven the magma in the cap to more evolved compositions.
Intermediate lavas (Qwai) are demonstrably scarce (Fig. 4) but yield valuable insights into current state of magmatic system. Intermediate eruptions are restricted to low-lying vents in faulted regions west of the edifice (Section 5.2.2) and the presence of silicic enclaves within the scoria cone sequences (Section 6.1) suggest that mafic and silicic compositions must be present within the same magma plumbing system. On the main edifice of Aluto there is no evidence for mafic-intermediate magmas having erupted, nor do we find any mixing-mingling between mafic-intermediate and silicic magmas. Peralkaline melt zones are commonly assumed to form a density barrier that prohibits mafic melts from reaching the surface (Mahood, 1984;Neave et al., 2012). The fact that no mafic-intermediate lavas are found on the centre of the Aluto edifice suggests that peralkaline melts beneath the complex are sufficiently aggregated to prohibit ascent of dense mafic melts. Hence the rise of mafic-intermediate magma to the surface can only be accomplished via fault networks beyond the edge of the silicic melt cap. This contrasts markedly with Gedemsa (a peralkaline complex 75 km north-east of Aluto) where basaltic lavas have erupted within the caldera and have entrained silicic liquids as well as partially crystalline silicic rock fragments (Peccerillo et al., 2003). Aluto has undergone much greater volumes of post-caldera volcanism than Gedemsa (Hutchison et al., in press) it also hosts a much larger geothermal field and shows current evidence for ground deformation (Biggs et al., 2011;, these lines of evidence as well as the absence of mafic lavas on the centre of the complex are consistent with a more consolidated melt zone at Aluto.

The pace of silicic volcanism in the MER
Our stratigraphy shows that Aluto underwent an early phase of edifice building (duration not yet constrained but likely on order of 150-400 ka, based on similar peralkaline complexes in the Kenyan rift, e.g., Menegai, Leat et al., 1984 andLongonot, Clarke et al., 1990), major ignimbrite eruptions at~310 ka, a period of repose lasting 250 ka and then episodic post-caldera volcanism after~60 ka. The apparent lack of volcanic activity at Aluto between 300 and 60 ka could be explained by a real hiatus in volcanism, or potentially a gap in sampling. Our updated stratigraphy (Fig. 3) represents the most complete coverage of the volcanic eruptive events to date, and on the basis of our lithological correlations between the surface and deep stratigraphy (Section 5.3) we consider that this hiatus in volcanism is real. However, until unambiguous geochronological measurements are made across a complete field or core section that covers the transition from major ignimbrite eruptions to the post-caldera eruptions some uncertainty remains.
Overall, our geological interpretations and new ages suggest that silicic volcanism at Aluto can be considered episodic on a variety of timescales. Large volume caldera-forming events take place on longtimescales N100 ka, linked to substantial deliveries of mantle-derived melt to the mid-crust. Stalling of magmas and melt capture processes (e.g., Keranen et al., 2004;Ebinger et al., 2008;Karlstrom et al., 2009) help to assemble large volume upper-crustal magma reservoirs (N10 km 3 ), and crystal fractionation is key to driving melt evolution to silicic compositions (Peccerillo et al., 2003;Rooney et al., 2012). We also see evidence for shorter timescale (ca. 10 ka) eruptive cycles, represented at Aluto by unconformity-bounded post-caldera phases (Fig.  3, Section 7.5). These cycles may link to the recharge and replenishment of small volumes (b10 km 3 ) of melt and volatiles in the cap of established upper crustal reservoirs (e.g., Civetta et al., 1988).

Future volcanic hazards
Volcanic phenomena that have occurred at Aluto in the last 10 ka include obsidian coulees (Qpoy), pumice cones and PDCs (Qpby), and pumice fallout (Qup). Lahars (Qal) have also occurred, although may only have been facilitated during episodes of humid climate prior to~5 ka (e.g., Le Turdu et al., 1999;Benvenuti et al., 2002). In Table 3 we evaluate potential volcanic hazards and population exposure at Aluto.
Based on the record of recent volcanism, a future eruption of Aluto will likely begin with a moderate explosive eruption to form a pumice cone (potentially with localized PDCs), followed by emplacement of an obsidian coulee (Table 3). The recent obsidian coulees (Qpoy) have volumes of 1-100 × 10 6 m 3 (e.g., Fig. S1A). Pumice cones (Qpby) built up prior to the obsidian coulees are difficult to trace laterally, but we estimate that small cones have volumes of~10 × 10 6 m 3 while the largest pumice cones (e.g., the large Qpby dome mapped north of well LA-6, Fig.  4) have volumes of~250 × 10 6 m 3 . Our geological mapping (Fig. 4) shows that there have been thirteen obsidian coulees (map unit Qpoy) emplaced over the last~10,000 years (assuming radiocarbon age correlations for Qup are valid, Section 5.3). Of the thirteen Qpoy coulees that have erupted a few of these appear to have erupted synchronously, and for simplicity, if we assume there have been 10 distinct eruptions, this would support an average silicic lava dome eruption at a rate of 1 per 1000 years. This value is broadly in line with observations of tephra layers preserved in lacustrine sections west of the Aluto that showed~13 tephra horizons within the Abernosa pumice deposit, which spans a period of~15,000 years (Section 5.3, Gasse and Street, 1978).
Radiocarbon ages demonstrate that volcanism has occurred relatively recently at Aluto (~0.4 cal. ka BP, Hutchison et al., in press) and a salient feature of this current phase of activity is that lavas have erupted from discrete vents associated with volcanic and tectonic faults (Hutchison et al., 2015) but they show no obvious spatial progression with time. Dedicated volcanic monitoring is an important next step and remote sensing methods (e.g., InSAR, Biggs et al., 2011) have the potential to be vital for constraining magma movement within the subsurface prior to future eruptions. Expansion of geothermal infrastructure across these silicic volcanoes offers an opportunity to build permanent monitoring networks (e.g., seismic and continuous GPS stations) to assess volcanic hazards and monitor geothermal resources at these new installations. More generally, a thorough understanding of the recurrence interval of post-caldera phases (e.g., Pyle and Elliott, 2006;Nomikou et al., 2014), detailed mapping of tephra fall (e.g., Fontijn et al., 2010Fontijn et al., , 2011 and PDC deposits (e.g., Oppenheimer, 2000, 2005;Rampey et al., 2010Rampey et al., , 2014, and forward modelling of a spectrum of eruption scenarios (e.g. Aspinall and Woo, 2014;Jenkins et al., 2015) will be critical for developing probabilistic models of volcanic hazard at Aluto.

Conclusions
Field mapping, remote sensing, geochronology and geochemistry provide new insights into the eruptive history and magmatic evolution Table 3 Potential volcanic hazards at Aluto. Population data was provided by the Smithsonian Global Volcanism Program (Siebert and Simkin, 2002 of the Aluto volcanic complex in Ethiopia. The first eruptions followed a period of mafic fissure volcanism and rift basin development. The complex was initially built up as a trachytic edifice, via both explosive and effusive activity. At~310 ka the complex underwent a phase of major explosive activity (represented in the field by extensive welded ignimbrite sheets) which we propose developed a caldera rim structure and ring fault. After a substantial hiatus in activity, post-caldera volcanism began at~60 ka. We identify four distinct post-caldera rhyolitic sequences that include pumice fall, PDC and lava flow units. These sequences have progressively in-filled the caldera and their vents are frequently aligned along pre-existing structural weaknesses of both volcanic and tectonic origin. Whole-rock geochemical data suggests that crustal melting did not play a significant role in generating the evolved rocks of Aluto, and without isotopic constraints the simplest interpretation of our data is that silicic magmas of Aluto were mainly generated through protracted fractional crystallization processes. Recent silicic eruptions appear to occur at an average rate of 1 per 1000 years, and we expect that future eruptions of Aluto will involve explosive emplacement of localised pumice cones and effusive obsidian coulees of volumes between 1 and 100 × 10 6 m 3 . Given that N 300,000 people live within 30 km of Aluto, as well as the targeted geothermal investment, dedicated monitoring via remote sensing and/or field installations is a critical next step to enable effective management and mitigation of future risk at this volcano. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jvolgeores.2016.09.010.