Structural controls on fluid pathways in an active rift system: A case study of the Aluto volcanic complex

Existing fault structures can play a significant role in the development of a volcanic complex, ultimately providing high permeability pathways for magma, hydrothermal fluids, and gas to ascend to the surface (e.g., Arnórsson, 1995; Rowland and Sibson, 2004; Caliro et al., 2005; Fridriksson et al., 2006). Understanding how preexisting structures such as regional tectonic faults and caldera ring faults affect fluid flow to the surface is a major task in defining the evolution of rift zones and has important implications for mineralization, geothermal exploration, and the assessment of volcanic hazard.

Recent work, specifically focused on hydrothermal venting and volcanic degassing (Schöpa et al., 2011; Pantaleo and Walter, 2013), has shown that while preexisting structures may control permeability at the edifice scale, at smaller scales these structural controls may be obscured by localized near-surface permeability variations. These local influences may include (1) lithological variations, where fluids will preferentially migrate along high permeability layers (e.g., poorly consolidated tephra layers) and (2) topographic controls, where the stress field induced by gravitational loading causes fracturing parallel to topography, and focuses pathways for steam and other gases toward topographic highs (Schöpa et al., 2011). To understand how large-scale structures influence active volcanic processes it is useful to look at the surface expression of different volcanic fluids (i.e., magma, hydrothermal fluids, and gas) across a variety of scales to disentangle large-scale structural controls from these localized near-surface permeability variations.

Both direct and remote measurements can be used to assess the spatial distribution of fluids and fluid pathways. Remotely sensed data such as lidar (light detection and ranging) and aerial photography (e.g., Pyle and Elliott, 2006; Cashman et al., 2013) are powerful tools to analyze volcano morphology, map sites of eruption and extrusion, and distinguish zones of hydrothermal alteration and fluid upwelling (e.g., Crowley and Zimbelman, 1997). On the other hand, volcanic gases (e.g., CO₂) that may be difficult to detect remotely, can be readily measured in the field using modern surveying techniques (Chiodini et al., 1998) and gridded to produce detailed maps of gas flux across a volcanic edifice (Cardellini et al., 2003; Parks et al., 2013). These techniques allow us to build detailed pictures of how different fluids are released from active volcanoes; the challenge for volcanologists is integrating these observations to unravel the subsurface structure and the processes controlling fluid pathways.

The Main Ethiopian Rift (MER, East Africa) provides an ideal natural laboratory to study how preexisting structural features (of both volcanic and tectonic origin) influence active volcanic processes. Firstly, the MER hosts a number of young silicic peralkaline volcanoes, allowing investigation of active magmatic and geothermal systems. Secondly, extension in the MER has generated abundant faults and fracture networks (e.g., Keir et al., 2006; Corti, 2009) through which magma can ascend and erupt. Finally, many of the MER volcanoes have undergone caldera collapse (Cole et al., 2005) and thus are likely to have established ring fault systems (Cole, 1969; Gibson, 1970; Mohr et al., 1980; Acocella et al., 2003; Rampey et al., 2010).

The silicic peralkaline volcanoes of the MER are among the least studied on Earth: few have detailed geological maps and significant knowledge gaps exist regarding their past and current activity (Aspinall et al., 2011). Detailed studies of peralkaline volcanic systems are limited to a few key complexes (e.g., Pantelleria; Mahood and Hildreth, 1983, 1986; Civetta et al., 1988; White et al., 2009; Neave et al., 2012; Williams et al., 2013), despite the fact that they appear to be a ubiquitous feature of continental rift zones (Mahood, 1984; Cole et al., 2005). The caldera structures produced at peralkaline volcanic centers in the East African Rift system are also of note because many appear elliptical in map view (e.g., Acocella et al., 2003; Bosworth et al., 2003; Holohan et al., 2005). While several recent publications have emphasized the role of elongate magma chamber collapse in generating elliptical calderas in the East African Rift system (e.g., Acocella et al., 2003; Bosworth et al., 2003), there is a lack of consensus regarding the exact mechanism in the MER. Establishing the controls on magma rise and ponding in tectonically thinned crust is fundamental to understanding how continental rift zones evolve (Ebinger et al., 2010).

In this paper we integrate observations from field campaigns, airborne remote sensing (lidar, aerial photos) and soil-gas CO₂ surveys to examine how magma, hydrothermal fluid, and gas pathways are coupled to the major structural features on Aluto, a typical young peralkaline volcanic complex of the MER. We show that each data set provides unique information about the complex and the links between volcanic activity and preexisting volcanic and tectonic structures. From these data we develop a conceptual model that captures both the volcanic evolution and the role these major structures play in controlling fluid pathways.

The MER (Fig. 1) is a zone of active extension in the East African Rift system that connects the Afar depression to the north with the Turkana depression and Kenyan rift to the south. The MER is an oblique rift, exhibiting an overall NE–SW trend, formed by E–W extension between the Nubia and Somalia plates via both magmatic intrusion and tectonic faulting (Ebinger, 2005; Corti, 2009; Corti et al., 2013a). Geodetic and seismic data (Bendick et al., 2006; Keir et al., 2006; Stamps et al., 2008) indicate that the current E–W (∼N100°E) extension rates are 4–6 mm yr^–1. The MER is usually divided into three sectors (northern, central, and southern) that reflect differences in terms of the spatial pattern of the faulting (Agostini et al., 2011), the timing of the major faulting episodes (Woldegabriel et al., 1990; Wolfenden et al., 2004; Bonini et al., 2005), and the thermal-mechanical state of the lithosphere (e.g., Keranen and Klemperer, 2008). This pattern is consistent with rift maturity increasing northward along the MER toward Afar, where the overall physiology changes from continental rifting to incipient oceanic spreading (Beutel et al., 2010; Ebinger et al., 2010; Ferguson et al., 2013).

The MER has two distinct fault sets: (1) NE–SW-oriented border faults with large vertical offsets (>100 m) on the boundaries of the rift, and (2) a set of closely spaced internal faults, the Wonji faults, with smaller vertical offsets (<100 m) oriented NNE–SSW and concentrated on the rift floor (Chorowicz et al., 1994; Boccaletti et al., 1998). In the central MER (CMER), the focus of this paper, the border faults formed at 6–8 Ma (Woldegabriel et al., 1990; Bonini et al., 2005), while the Wonji faults initiated ca. 2 Ma (Boccaletti et al., 1998; Ebinger and Casey, 2001). These observations support models of rift initiation and continental extension (Hayward and Ebinger, 1996; Ebinger, 2005; Corti, 2008) whereby border faults are progressively abandoned and strain becomes localized to active segments central to the rift zone. This focusing of strain has resulted in discrete narrow zones (∼20 km wide) where extension is accommodated by Wonji faults as well as magmatic intrusion (Keranen et al., 2004; Keir et al., 2005; Kendall et al., 2005; Mackenzie et al., 2005). In the CMER, while the border faults are still seismically active (Keir et al., 2006; Pizzi et al., 2006) much of the active tectonic deformation occurs in the magmatic segments via the NNE–SSW Wonji faults (Keir et al., 2006).

Magmatic intrusion within the active segments has produced abundant surface volcanism. Basalts are associated with scoria cones and eruptive fissures, whereas rhyolitic volcanism has produced a series of peralkaline volcanoes comprising silicic lava flows, domes, and coulees, and extensive pyroclastic products (Gibson, 1969; Di Paola, 1972).

Aluto is a silicic peralkaline volcano located in the CMER (Fig. 1). The complex (outlined by the white box in Fig. 1 and detailed in Fig. 2A) is dominated by an ∼14-km-wide, 700-m-high edifice composed of a thick pile of coalescing rhyolitic lava flows and domes, pumice cones, and ignimbrite deposits. There are a significant number of smaller volcanic vents and domes formed off the main edifice, particularly to the northwest of the complex. Aluto has an internally drained central depression (herein referred to as a caldera floor) that has been partially in-filled by alluvium and reworked material from the surrounding surface volcanic deposits. While there are no detailed accounts of the eruptive history of the complex, descriptions of the geology of Aluto are in regional-scale mapping reports (e.g., Dakin and Gibson, 1971; Di Paola, 1972; Kebede et al., 1985); studies of the Ziway-Shala lake basin system (Gasse and Street, 1978; Street, 1979; Le Turdu et al., 1999; Benvenuti et al., 2002; Gibert et al., 2002), and publications related to geothermal development on Aluto (Gebregzabher, 1986; Valori et al., 1992; Gizaw, 1993; Gianelli and Teklemariam, 1993; Teklemariam et al., 1996; Saibi et al., 2012). This work and our own field investigations suggest that the Aluto volcanic complex has undergone several cycles of explosive and effusive volcanism throughout its history and that these eruptive phases have been dominantly silicic in composition.

There are few constraints on the ages of the erupted products of Aluto. A single K/Ar date from a geothermal feasibility study gives an age of 155 ± 8 ka for the Hulo-Seyno ignimbrite, which is taken as one of the first products of the silicic complex (ELC Electroconsult, 1986). The most recent volcanic activity on Aluto is represented by a series of obsidian lava flows, the youngest of which likely erupted within the past 2000 yr (Gianelli and Teklemariam, 1993). Aluto is an active volcanic system. Across the complex, hydrothermal manifestations include fumaroles and hot springs (Kebede et al., 1985). Interferometric synthetic aperture radar (InSAR) studies show that Aluto has undergone several uplift and subsidence events of 10–15 cm over the past decade (Biggs et al., 2011). While the exact mechanism driving deformation is uncertain, the source characteristics and time scale are indicative of shallow magmatic processes active beneath the complex (Biggs et al., 2011).

Eight exploration wells have been drilled on Aluto (LA-1–LA-8; Fig. 2A). From these a deep stratigraphy of Aluto has been assembled that forms the model for the geothermal field (Gizaw, 1993; Gianelli and Teklemariam, 1993). A geological cross section compiled from these previous studies, as well as various drilling reports provided by the Geological Survey of Ethiopia, is given in Figure 2B. The deep well stratigraphy reveals that there are substantial offsets in the units between the wells. This is taken as evidence that faults, shown in Figure 2A, dissect the complex. There are also thickness variations within individual units between the wells; this is particularly evident in a Pleistocene lacustrine horizon that shows a stepped thinning pattern from west to east (i.e., from the rift center, toward the rift margin). The abrupt changes in thickness of this unit suggest that deposition took place on an uneven surface. Given the rift geological setting, this is likely to be associated with faulting; however, other processes such as wedging out of the layer on a previous erosive surface cannot be ruled out. The Pleistocene lacustrine horizon predates the Aluto silicic products, so if deposition of this unit was fault controlled this would imply that significant faulting preceded volcanic activity at Aluto. Of the 8 wells drilled, only 2 (LA-3 and LA-6) are productive. The sustained production in these wells is linked to their location along a major NNE–SSW fault on Aluto (herein referred to as the Artu Jawe fault zone; AJFZ in Fig. 2A). Well temperatures (Gizaw, 1993), alteration mineral assemblages (Gebregzabher, 1986; Gianelli and Teklemariam, 1993), and the Na/K ratio of geothermal fluids from these deep wells (Gizaw, 1993) all support the hypothesis that this faulted zone, photographed in Fig. 3A, is the main upflow zone through which high-temperature fluids ascend to the surface.

Mapping of the volcanic and tectonic landforms on Aluto was carried out using airborne and satellite remote sensing, complemented by eight weeks of field mapping to ground truth these interpretations. High-spatial-resolution topography and imagery of the volcanic edifice were acquired in November 2012 by the UK Natural Environmental Research Council’s Airborne Research and Survey Facility. The aircraft was equipped with a Leica ALS50 airborne laser scanner, AISA (Airborne Imaging Spectrometer for Application) Eagle and Hawk hyperspectral instruments and a Leica RCD105 39 megapixel digital camera. The lidar system acquired first, second, and last returns for ∼140 × 10⁶ discrete points and was operated with a pulse repetition frequency of 46,200 kHz and scan frequency of 29.9 Hz. The last-return data points were combined into a single point cloud and a 2-m-resolution digital elevation model (DEM) was generated using GRASS (Geographic Resources Analysis Support System; http://grass.osgeo.org/). The DEM was also used to orthorectify aerial photos acquired by the digital camera. Orthorectification was performed with the Leica Photogrammetry Suite in ERDAS Imagine software (www.hexagongeospatial.com/products/remote-sensing/erdas-imagine) using camera calibration information provided with the photographic data. Orthophotos reveal valuable surface and morphological information and have a resolution of 0.25–0.50 m.

Using the airborne data we mapped (1) volcanic vent locations, the site of an effusive vent of a lava flow, and the center of a crater (or set of nested crater features), represented by points (Figs. 4 and 5), and (2) volcanic lineaments, crater rims, and other linear fissures represented by line features (Fig. 5B). We could only map volcanic lineaments on the main edifice of Aluto (i.e., the area covered by the lidar DEM); off the main edifice, Google Earth and ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite imagery were used. Owing to their lower resolution and the generally greater erosion of the volcanic landforms beyond the main edifice, only vent locations could be accurately mapped (Fig. 4).

The spatial distribution of volcanic vents and the orientation of elongate vents and fissures can be used to make inferences about geometry of feeder dikes and ultimately evaluate the underlying structural or stress field controls on magma pathways (e.g., Nakamura, 1977; Tibaldi, 1995; Paulsen and Wilson, 2010). On the main edifice, we used crater rim and fissure line data (Fig. 5) to investigate whether these features exhibited any prevailing orientations and alignments with regional faults. To do this we analyzed the orientation of all individual segments of the digitized crater and fissure line data and length-weighted our results. The results are illustrated for an individual elongate crater in Figure 6A; the inset clearly demonstrates a rose plot with prevailing E–W orientation. This method is similar to that employed previously (e.g., Trippanera et al., 2014), but has the advantage that rather than taking one averaged elongation direction, all mapped segments that make up the crater and fissures are used and can be compared together. Beyond the coverage of the lidar DEM, where it was not possible to carry out detailed volcanic lineament mapping, an azimuth-based method was used to assess vent alignments (Lutz, 1986; Wadge and Cross, 1988; Lutz and Gutmann, 1995; Cebriá et al., 2011). Vents are treated as discrete points, and the azimuth (or alignment direction) between a vent and its neighbors is determined for a given separation distance. When vent locations (or magma pathway to the surface) are controlled by faulting, then neighboring vents will have similar orientations relative to each other. However, as the distance between vents increases, then so does the likelihood that the vents will not be located on the same fracture. Thus, alignments need to be assessed within a specific separation window to best expose regional structural trends (Cebriá et al., 2011). Ultimately, these numerical interpretations of the volcanic vents and structural data are considered in the context of the physical observations from the remote sensing, field observations, and the CO₂ degassing measurements to assess their real geological significance.

Hydrothermal fluid upwelling on Aluto causes visible alteration of the surface volcanic products, producing red and orange clays over areas >>10 m², which can be mapped using aerial photographs (Fig. 6B). While an automated mapping approach using the spectral signature of these alteration facies (e.g., Crowley and Zimbelman, 1997) would be feasible given our data set, for the purposes of this study the resolution of the aerial photos is sufficiently high that major altered zones can be approximately mapped, and their locations can be linked with structural interpretations and CO₂ degassing results. Our assessment of hydrothermal alteration made using the orthophotos correlates well with fumarole vents mapped on the ground by the Geological Survey of Ethiopia (Kebede et al., 1985); for hot spring locations that were beyond the coverage of our remote sensing data, we used Kebede et al. (1985) to constrain their location.

Measurements of soil CO₂ flux on Aluto were undertaken in three surveys between January 2012 and February 2014. The first survey (January 2012) sought to transect large-scale structures and identify whether they provided key permeability pathways for CO₂-rich geothermal fluids and magmatic gases to upwell. Degassing surveys were conducted in November 2012 and February 2014, and focused on producing a detailed map of spatial degassing patterns along the major tectonic fault (Artu Jawe fault zone). A 1000 m × 800 m study area (Fig. 5A) was chosen to include both the major fault scarp (photograph in Fig. 3A) and the productive geothermal wells (LA-3 and LA-6). A 30-m sampling grid was use, a compromise between attaining spatial coverage of the fault zone at sufficient sampling resolution. In total 560 sites were visited (424 in November 2012 and 136 in February 2014). Those that were in geothermal effluent ponds, areas of dense vegetation, or on hazardous slopes were excluded, and in some instances extra measurements were made off the predefined sampling grid to help characterize the highest values in the degassing regions.

The CO₂ flux was measured directly using two portable closed system gas analyzer units (a LICOR LI-8100 automated soil CO₂ flux system and a PP-systems SRC-1 chamber with EGM-4 analyzer). Both instruments have an infrared gas analyzer and use the accumulation method (Parkinson, 1981; Chiodini et al., 1998) to measure CO₂ flux. Measurements were consistent between the instruments; comparisons made at identical sites showed variations of 10%–25% between the two instruments (significantly less than the variation seen across the complex). Repeated site measurements showed variations of ∼25% in low flux zones (<10 g m^–2 d^–1) and <10% in high flux zones (>100 g m^–2 d^–1), consistent with random error in natural emission rates (Carapezza and Granieri, 2004; Viveiros et al., 2010) and in line with the quoted reproducibility of each instrument (5%–10%, Chiodini et al., 1998; Giammanco et al., 2007).

To generate maps of soil CO₂ flux from the discrete point measurements the sequential Gaussian simulation (sGs) method was used (Cardellini et al., 2003). A simulation grid was defined (at higher spatial resolution than the sampling grid) and 100 sGs were performed using the sgsim code (Deutsch and Journel, 1998) available in the Stanford Geostatistical Modeling Software (SGeMS) open-source geostatistics package (Remy et al., 2009). A CO₂ flux map was constructed from these simulations taking the arithmetic mean of each individual cell across all simulations, equivalent to the E-type soil flux map proposed by Cardellini et al. (2003). The total CO₂ flux was calculated for each simulation and the mean and standard deviations of all simulations were computed and used to estimate total CO₂ release as well as the associated uncertainty (Cardellini et al., 2003).

The spatial distribution of volcanic vents on Aluto is shown in Figure 4. Vents are largely restricted to the main edifice and a NNE–SSE-trending zone to the northwest of this. A detailed map of volcanic vents, lava flows, craters and fissures overlain on the lidar DEM is given in Figure 5. Lava flows on the central edifice are rhyolitic; Figure 6A shows shaded relief and slope maps for a typical obsidian lava deposit. Craters are predominantly <1 km in diameter and <100 m deep; many are elliptical (e.g., Fig. 6A) or are composed of nested structures (Fig. 5).

On the northeast flank of Aluto, several lines of evidence support the existence of a caldera rim structure; from remote sensing data we identify a 2.5-km-long arcuate structure (CR in Fig. 2A; also see Figs. 4 and 5) that is orthogonal to the regional NNE–SSE tectonic structures (RTS in Fig. 2A). Viewed from the southwest (Fig. 3C), the rim comprises a steep wall 50–60 m in height. Fumaroles and hydrothermal alteration are picked out at the base of the structure (Fig. 3C), and tentatively correlated with a buried ring fault. Northeast of the rim (Figs. 2A and 5) the topography gently slopes away from the edifice and is incised by numerous gullies and channels. The morphology of this structure is characteristic of a volcanic caldera (Lipman, 1997; Cole et al., 2005), including a caldera rim, curving landslide scarp, and possible landslide breccia that conceals the ring fault (labeled in Fig. 3C).

An underlying caldera ring fault (E–W-oriented ellipse; Figs. 4 and 5) centered on the floor of the main edifice may explain a number of aspects of the young volcanism. Assuming that vent and fissure elongation parallels the orientation of subsurface fractures (Paulsen and Wilson, 2010), then crater rim and fissure lineations in Figure 5 can be used as proxy for the underlying structures feeding surface eruption. Vent elongations on the southern rim of Aluto show clear E–W orientations (e.g., Fig. 6A, and the nested crater structures east of this in Fig. 5B) indicative of an E–W–oriented fissure feeding surface eruptions. In Figure 5B it is also clear that mapped crater rims on the western margin of the edifice show approximate elongation N–S, and likewise on the east of the complex a group of clustered craters shows overall elongations in N–S and NE–SW directions. The distribution of volcanic vents (Figs. 4 and 5) also shows a dominantly elliptical ring distribution on the main edifice with a marked absence of vents within the center of the caldera. Figure 4 shows that a number of vents on the main edifice (62% of total mapped using the lidar) are within ±500 m uncertainty of the proposed ring structure. Vents that are not within this structure may have been controlled by other structures (e.g., the Artu Jawe fault zone; see following), or alternatively the caldera ring fault may be more structurally complex than the proposed ellipse. Caldera ring faults are unlikely to truly approximate an ellipse in shape; however, this is the simplest first-order model of a ring fault in the absence of other evidence concerning its structure.

The vent elongations and distributions match orientations that would be expected for a mature caldera structure feeding eruptions at the surface. We assume that feeder dikes here are either elongate in response to the local stress field set up on the caldera or are directly exploiting an existing ring fracture. The ellipse fit is ∼5 km in N–S diameter and ∼8 km in E–W diameter, with ellipticity (short/long axis ratio) of 0.63 and long-axis elongation of N090°E. The mapped caldera rim remnant corresponds closely to the proposed ring fracture, although it only overlaps ∼8% of the entire ellipse (Figs. 4 and 5).

The deep well stratigraphy (Fig. 2B) reveals several major fault zones, aligned with regional tectonic trends, dissecting the Aluto volcanic complex. However, few of these structures have a clear surface expression, due to the large volumes of young volcanic and lacustrine sediment cover. On the edifice the surface is mantled by tephra and obsidian lava flows; to the north, west, and south of Aluto there are substantial lacustrine deposits (including gravels, sands, muds, and diatomite) that were deposited throughout the late Pleistocene and early Holocene (Grove and Goudie, 1971; Grove et al., 1975; Gasse and Street, 1978; Street, 1979; Gillespie et al., 1983; Le Turdu et al., 1999; Benvenuti et al., 2002).

The only major fault scarp visible on Aluto is exposed in a 500-m-long segment on the east of the caldera, the NNE–SSW Artu Jawe fault zone (Figs. 2A and 3A). Topographic profiles across the fault show a typical profile expected for a normal fault down throwing to the west (with a maximum throw of 50 m measured at the surface). The scarp is mostly covered by recent slumped clastic material, but in a more pristine deeper exposure a highly brecciated aphyric obsidian lava is visible (Fig. 3B). The poorly consolidated texture and thin (1–10 cm) subvertical banding of the deposit is very different from obsidian lavas encountered elsewhere on the complex and therefore may represent fault gouge. The near vertical dip and approximate north-south strike of the foliations (Fig. 3B) closely match the overall trend of the fault scarp.

To the north, the fault scarp is covered by an elongate dome of pumice and obsidian lava units (Figs. 5B and 7). This elongate dome is 2 km (E–W) by 4 km (north-south), ∼180 m high, and is close to the trend of the Artu Jawe fault scarp. We infer that these volcanic deposits have erupted along the fault zone building up the elongate pumice dome. To the south of the Artu Jawe fault scarp the surface rupture is obscured by mantling tephra; however three-dimensional views of the fault zone (Fig. 7) show a gorge following the same trend, supporting its continuation into the volcanic pile. There is evidence of a secondary, less-pronounced break in topography ∼500 m west of the main scarp (Fig. 7). The surface topography data alone are not sufficient to classify these faults separately. Owing to their proximity to the main scarp, we group these structures together as the Artu Jawe fault zone.

There are no clear fault traces exposed on the caldera floor (CF in Fig. 7), suggesting that either the material infilling the caldera is very young and/or there has been no significant recent displacement on faults within the caldera. It has been suggested (Kebede et al., 1985; ELC Electroconsult, 1986) that a fault gorge extends through the southern rim (SFG in Fig. 7). A line of fumaroles parallel to this gorge (Fig. 7B) provides qualitative evidence to support existence of this fault (see following discussion of hydrothermal features).

Three obsidian domes and nested craters were identified on the northwest of the caldera floor (Fig. 6C). The domes are of comparable size (150–400 m diameter) and are aligned along a trend of N050°E. In the field the domes appear compositionally similar and display similar weathering characteristics, suggesting that they share a common source and erupted at a similar time. These observations suggest an underlying tectonic control on the eruption of these lavas at the surface.

From remote sensing (Figs. 2A and 4) there is a clear prevalence of volcanic vents aligned NNE–SSW northwest of the edifice, but there is no evidence for large fault structures at the surface. Highstands in the lake system since 10 ka have reached maximum lake levels of 1660–1680 m above sea level (Street, 1979; Gillespie et al., 1983; Benvenuti et al., 2002, 2013), sufficient to isolate these vents. We therefore assume that these young lacustrine sediment deposits have masked any surface fault breaks.

In Figure 8 we compare the orientations of volcanic features and vent alignments with regional fault trends. As expected (e.g., Agostini et al., 2011), we can distinguish two distinct fault systems in the CMER: border faults and the younger internal faults (referred to herein as the Wonji faults). There is a marked difference in strike between the two groups (Figs. 8A, 8B). The Wonji faults have a NNE–SSW strike (mean value of N012°E); in contrast, the border faults have a more NE–SW orientation (mean value of N032°E).

Figure 8C shows a length-weighted rose plot for all crater rim and fissure line segments mapped on the main volcanic edifice (identified in Fig. 5). Unlike tectonic fault trends these features display preferential orientations along both N–S and E–W alignments. Crater rim and fissure line orientations are also skewed to NE–SW rather than NW–SE alignments (Fig. 8C). Figures 8D and 8E show the azimuth between volcanic vent locations (i.e., point features) on and off the main edifice of Aluto. On the main edifice the volcanic vent alignments (Fig. 8D) show equivalent results to those shown in Figure 8C; there is a pronounced NNE–SSW alignment (N010–020E° bin), secondary E–W alignment, and a general skew to NE–SW rather than NW–SE alignments. Off the main edifice (Fig. 8E) the results show that the prevailing vent alignment range is N000–020E°.

The preferential E–W orientations observed in Figure 8C arise due to the large number of elongated craters (clearly present on the southern rim of the complex; Fig. 5). Figures 8C and 8D both support the existence of E–W-oriented structures feeding surface eruptions; following the preceding discussion, it is assumed this is the surface manifestation of an underlying elliptical caldera structure. Crater and fissure line orientations (Fig. 8C) as well as vent alignments (Fig. 8D) also show N–S and NNE–SSW orientations; these are representative of aligned elongate craters on the east and west margins of the caldera ring structure (Fig. 5) but also alignments along the major tectonic fault structure (Artu Jawe fault zone). Off the main edifice (Fig. 8E) vent alignments are very close to the observed fault strike of the Wonji faults (Fig. 8A), strengthening the case that beyond the caldera tectonic fault structures are the dominant control on magma pathways.

Figure 4 shows the zones of hydrothermal alteration, fumaroles, and hot springs mapped using the aerial photos and supplemented by mapping reports from the Geological Survey of Ethiopia (Kebede et al., 1985). A summary of the main hydrothermal features and their links to the volcanic and tectonic structures is provided in Table 1.

The sites of hydrothermal activity are confined to the summit and flanks of Aluto, as well as the region of hot springs around the north bay of Lake Langano (Fig. 4). There is no evidence of active or fossil hydrothermal activity on the northern and eastern flanks of the Aluto edifice. Along the Artu Jawe fault zone (Fig. 5) fumaroles continue both north and south of the fault scarp (Figs. 5 and 7B), indicating that the fault continues beneath the young volcanic cover (Fig. 7A). Hydrothermal alteration is also found at the base of the remnant caldera rim (Fig. 4), where several zones of fumaroles are seen emanating from fractures in altered lava (Fig. 3C; Bobesea, Table 1). The lines of fumaroles (identified in Fig. 3C) follow a NW–SE trend, identical to the orientation of the caldera rim, and thus are linked to a buried ring fault.

Not all hydrothermal features on Aluto can be correlated with mapped structures. For example, on the west flank of Aluto fumaroles occur over a large area at Hulo and Auto (HL and AT in Fig. 4) and while these may link to structures and aligned volcanic vents northwest of the edifice, evidence for a tectonic control at the surface is absent (Fig. 6B). On the southern flank of Aluto, fumaroles at Kore and Gebiba (K and GB in Fig. 4) also show no clear structural association.

The CO₂ flux values measured on Aluto span several orders of magnitude, varying between 0.5 and 40,000 g m^–2 d^–1. Background CO₂ flux measurements at locations 10–20 km away from Aluto ranged from 0.5 g m^–2 d^–1 in rocky organic-poor soils to 6 g m^–2 d^–1 in darker organic-rich soils (i.e., vegetated areas with more abundant leaf litter). A flux of 6 g m^–2 d^–1 was taken as the upper limit for a purely biogenic background CO₂ flux in the study area. These fluxes are low when compared to values that characterize soils in less arid climates, 10–30 g m^–2 d^–1 (Mielnick and Dugas, 2000; Rey et al., 2002; Cardellini et al., 2003).

The CO₂ degassing transects across the main structural features (A–A′, B–B′, C–C′; Fig. 5A) correspond with lidar-derived topography and soil CO₂ flux profiles shown in Figure 9. Profile A–A′ is along the center of the volcanic complex across the caldera floor and intersects the Artu Jawe fault zone to the east. The topography reveals the smoothly downward-curving caldera floor and the well-defined ∼50 m offset across the fault zone. The CO₂ flux through the caldera floor is below or very close to the background biogenic flux value. The absence of significant degassing through the caldera floor is in line with remote sensing observations that did not indicate any clear fault offsets. Adjacent to and within 500 m of the fault scarp the CO₂ flux is significantly above the background levels. The scale in Figure 9 (A–A′) was chosen to illustrate the small variations in the low fluxes, and maximum values measured along the fault zone were >10,000 g m^–2 d^–1 (the full range of flux values along the fault zone is shown in Fig. 10).

Profile B–B′ transects the remnant caldera rim structure. The CO₂ flux shows greatest values (to 1850 g m^–2 d^–1) at the base of the caldera rim, with lower values (2–7 g m^–2 d^–1) to the southwest and on the caldera rim. The area of anomalous degassing is defined by several sets of fumaroles; CO₂ flux values are greatest on fumaroles (500–1850 g m^–2 d^–1) and lower in cohesive soils between the open fractures (15–240 g m^–2 d^–1; Fig. 3C).

Profile C–C′ follows the topographic high on the southern rim of the complex. The CO₂ flux values increase from low background values (<6 g m^–2 d^–1) to 11 g m^–2 d^–1 at the peak and immediately south of the rim. Elevated CO₂ flux values coincide with the hypothesized location of the ring fault. However, the maximum values encountered at the peak were significantly lower than those found on profile B–B′, and no evident structural feature was exposed at the surface.

On profile A–A′ (Fig. 9), there are considerable CO₂ flux variations encountered in the fault zone. To better constrain these variations along the fault, we undertook a high-resolution (30 m) sampling survey. Figure 10 provides topography, surface geology, and soil CO₂ flux results for the survey area (outlined in Fig. 5B). The shaded relief DEM (Fig. 10A) identifies the main ∼50-m-high fault scarp (Fig. 3A) and its projected continuation north, where it is obscured by pumice dome deposits (identified in Figs. 5B and 7). To the west of the main fault scarp there are a number of smaller breaks in topography (10–20 m high) (white dashed lines in Fig. 10A).

Figure 10B shows the surface geology map. The area is largely covered in young tephra (pumice fall and pyroclastic deposits) sourced either from the pumice dome to the north of the study area (Figs. 5B and 7) or from eruptive vents on the southern rim of the caldera complex. At the base of these slopes, tephras were reworked by fluvial processes and are represented by fine-grained alluvium and gravel deposits. A porphyritic obsidian flow is located on the west of the grid. The obsidian flow is likely sourced from a vent at the base of the cliff; however, recent avalanching of the volcaniclastic deposits in the cliff has obscured the exact position of the vent.

The soil CO₂ flux map (Fig. 10C) shows that the flux varies by several orders of magnitude across the survey area, while a log probability plot (Fig. 10D) shows a bimodal distribution confirming the existence of two distinct CO₂ sources (i.e., background biogenic and volcanic hydrothermal; Chiodini et al., 1998) that contribute to the degassing. The highest CO₂ flux values are found close to fumarole vents, suggesting that in those areas CO₂ travels mostly with hydrothermal steam. From the mean of the simulations the total diffuse CO₂ emission for the 800,000 m² grid (Fig. 10C) is calculated as 62 ± 12 t d^–1.

Transect lines shown in Figure 10C link to Figure 11, which show the topography and flux measurements along the profile line. On profile W–W′, peak flux values occur at the base of the main cliff, associated with a fault plane. Profiles X–X′ and Y–Y′ (Fig. 11) cross the ridge of pyroclastic deposits east of the fault and also show a relationship between CO₂ flux and topography, with peak values in CO₂ flux close to but slightly displaced from the absolute topographic high. Profile Z–Z′ crosses several lithologies on the west of the grid; CO₂ flux values are on average lower on the obsidian lava flow deposits compared to the tephra deposits to the north and south.

The results of this study, combined with prior mapping and geothermal well data (Dakin and Gibson, 1971; Di Paola, 1972; Kebede et al., 1985; Gebregzabher, 1986; Valori et al., 1992; Gizaw, 1993; Gianelli and Teklemariam, 1993), show that the Aluto volcanic complex has been dissected by normal faults related to the regional extensional tectonics, which have in turn influenced the distribution of volcanic products. The aligned vent trains that cross Aluto reflect magma pathways controlled by faults. Young vents and craters on and off the main edifice show a prevailing alignment to the regional NNE–SSW strike of the Wonji faults (Figs. 4 and 8), although there is field evidence that minor NE–SW-oriented structures may also exist (Fig. 6C).

Physical evidence for the link between faults and volcanic activity is particularly clear along the Artu Jawe fault zone, where magma has erupted along the fault, building an elongate pumice dome (Figs. 5B and 7). The Artu Jawe fault zone remains a high permeability pathway through which steam (Fig. 4) and CO₂ (Fig. 10C) ascend to the surface, as evidenced by the zones of fumarolic alteration and peaks in CO₂ flux along the fault structure (A–A′; Figs. 7 and 9). Our work shows that hydrothermal mapping (Figs. 4 and 7) and CO₂ soil-gas surveys (Fig. 10C) can be used to identify high permeability zones, revealing the locations of hidden fault structures when multiple observational data sets are considered together (e.g., Figs. 7 and 10).

Understanding the initiation and growth of these faults is critical to address the development of the Aluto volcanic complex. At present the best evidence for major fault zones on Aluto, and the only constraints on their development, are provided by the deep wells. Of particular significance is a Pleistocene lacustrine horizon that can be correlated between wells (Fig. 2B; Gianelli and Teklemariam, 1993; Gizaw, 1993) and shows a stepped west to east thinning suggestive of faulting. This unit is beneath the volcanic products of Aluto (Fig. 2B), implying that significant faulting preceded surface silicic volcanism, and hence these preexisting faults may have controlled magmatic processes throughout the growth of Aluto. Investigations of the Soddo region in the southern MER by Corti et al. (2013b) found a similar relationship, with both basaltic and rhyolitic volcanic centers following closely preexisting border faults.

A number of previous studies have suggested that Aluto has undergone a caldera-forming event (Dakin and Gibson, 1971; Le Turdu et al., 1999), and our observations help to confirm this. Taking together the presence of a remnant caldera rim (Figs. 2A, 4, and 5), the occurrence of eruptive vents on an elliptical (ring) structure (Figs. 4 and 5) and the correspondence of this to zones of hydrothermal alteration (Fig. 4) and local peaks in CO₂ degassing (B–B′, C–C′, Fig. 9), we propose that an underlying ring fracture system that matches the mapped caldera rim remnant can explain these observations.

A major challenge in defining this structure is that the exposure of the remnant caldera rim is limited to the northeast of the complex (only covering ∼8% of the ellipse; Fig. 4). We assume that the rest of the structure has been covered by post-caldera volcanic deposits or eroded. This is typical of peralkaline volcanic edifices that undergo repeated phases of caldera collapse and caldera-filling volcanism (Cole et al., 2005) that overspill the caldera rim and mask its surface expression (Mahood, 1984). The size (∼8 km × ∼5 km) and shape of the proposed Aluto ring structure (Figs. 4 and 5B) are similar to other peralkaline complexes in the MER (Cole, 1969; Mohr et al., 1980; Acocella et al., 2003; Rampey et al., 2010), and within the upper range of other peralkaline calderas worldwide (Mahood, 1984; Cole et al., 2005). The caldera wall on Aluto (>50–60 m in height) is comparable to other peralkaline calderas where subsidence is a few hundred meters (Mahood, 1984).

The elliptical caldera form is typical of continental rift volcanoes worldwide (e.g., Wilson et al., 1984; Acocella et al., 2003; Bosworth et al., 2003; Casey et al., 2006; Geyer and Marti, 2008). Elliptical calderas may be produced by a number of different mechanisms (Holohan et al., 2005), including (1) collapse of an elliptical magma reservoir; (2) nesting, where multiple collapse structures overlap and give rise to an elongate geometry; (3) shallow crustal processes, where asymmetric collapse, or distortion of caldera faults leads to an asymmetric caldera above a circular magma chamber; and (4) post-caldera modification, where a circular caldera is distorted by erosion or regional deformation. Ultimately, the infilling of the Aluto caldera has left little information concerning the original collapse structure and the nature of the collapse event. In particular, it is unclear whether the caldera collapse involved one or multiple events, to create a nested structure (as is typical for peralkaline volcanic complexes; Mahood and Hildreth, 1983; Mahood, 1984; Rampey et al., 2010). It is also not possible to infer whether collapse was asymmetric or whether subsidence was accommodated by existing tectonic faults. Assuming that collapse took place at 155 ka, the only age constraint for the major ignimbrite deposits of Aluto (ELC Electroconsult, 1986), and assuming current extension rates of 5 mm yr^–1 (Bendick et al., 2006; Stamps et al., 2008), then this would generate <1 km of extension since caldera formation (i.e., not sufficient to explain all of the E–W elongation).

The elliptical calderas of the East African Rift system are frequently linked to the collapse of elongate magma chambers (e.g., Bosworth et al., 2003; Acocella et al., 2003). The genesis of elongate magma chambers in a continental rift currently revolves around two distinct models, one related to the regional extensional stress field causing differential spalling of the walls of the magma reservoir (Bosworth et al., 2003), and the other related to preexisting structural weaknesses, such as pre-rift faults, facilitating preferential magma chamber development (Acocella et al., 2003). To explain caldera elongation using the preexisting structures model requires a detailed understanding of the evolution and orientation of regional fault patterns. For example, Acocella et al. (2003) identified a number of E–W-oriented pre-rift tectonic structures on the flanks of the MER and invoked these to explain elongate calderas at Fentale, Kone, and Gedemsa. We investigated fault maps of the MER presented by Acocella et al. (2003), Abebe et al. (2007), and Agostini et al. (2011), as well as our own satellite imagery and could not highlight any clear E–W structures on the rift plateau adjacent to Aluto. On the other hand, however, the E–W elongation of the Aluto caldera is almost orthogonal to the active Wonji faults of the central MER (Fig. 8A) and parallel to the Nubia-Somalia displacement vector (∼N100°E; Bendick et al., 2006; Keir et al., 2006; Pizzi et al., 2006; Stamps et al., 2008). This observation is consistent with the idea that the regional stress field played a role in magma reservoir development at Aluto (cf. Bosworth et al., 2003).

Compared to neighboring MER volcanoes, Aluto has undergone considerable post-caldera volcanism. Shala (Mohr et al., 1980), Gedemsa (Peccerillo et al., 2003), and Kone (Cole, 1969; Rampey et al., 2010) volcanoes have all undergone caldera-forming eruptions; however, post-caldera activity has been modest, building up small domes of pumice and obsidian (within the caldera and adjacent to the caldera ring fault). At Corbetti, two silicic cones (Chabbi and Urji) have built up within the caldera (Mohr, 1966; Di Paola, 1971). Chabbi, the larger and more voluminous of these, has built up on the east of the caldera floor and now completely covers the caldera wall on the eastern side of the complex (Di Paola, 1971). This may provide a possible analogue to Aluto, which has undergone widely distributed post-caldera volcanism along the ring fracture, and thus only a small segment of the caldera wall is retained (Fig. 4).

At the edifice scale, both regional tectonic faults (such as the Artu Jawe fault zone; A–A′, Fig. 9) and the volcanic ring fracture (B–B′, Fig. 9) have a strong control on diffuse CO₂ emissions. The escarpment of the Artu Jawe fault zone (W–W′ in Fig. 11) shows that CO₂ gas is concentrated along the fault plane and that flux values remain high over a ∼50-m-wide high permeability crushed zone across the fault. Across the southern rim of the complex we find elevated CO₂ fluxes spread over a ∼500-m-wide zone (C–C′, Fig. 9), where we infer that a ring fault is deeply buried beneath a thick pile of volcanic deposits. Without deep penetrating structures connecting reservoir to surface, significant soil-flux degassing peaks (such as A–A′, B–B′, Fig. 9) are unlikely to be observed.

Substantial variations in the spatial pattern of CO₂ degassing are evident both parallel and transverse to the major structures when observed at scales <100 m (Fig. 10C). These are linked to permeability variations in the near surface (Schöpa et al., 2011; Pantaleo and Walter, 2013) possibly caused by changes in lithology or changes in the topography-induced stress field (that focuses the permeable pathways toward the morphological crests). There is a correlation between local topography and degassing flux (Fig. 11, X–X′, Y–Y′) in the pumice dome (Fig. 10). Our interpretation is that the pumice dome deposits have buried the Artu Jawe fault scarp, and thus steam and gas rising up along the fault plane enter the base of the dome and become focused toward the morphological ridge crest. A similar morphological control has been observed at Vesuvius (Italy), where CO₂ anomalies are concentrated in the inner slopes of the crater rim (Frondini et al., 2004). The key difference on Aluto is that CO₂ gas and steam are fed into the base of the volcanic pile from a tectonic fault rather than a volcanic conduit.

Several (e.g., Peltier et al., 2012; Pantaleo and Walter, 2013) have studied the influence of surface lithology and soil texture on permeability, and our data set permits a qualitative assessment for Aluto. Profile Z–Z′ (Fig. 11) crosses several contrasting lithologies west of the Artu Jawe fault scarp; CO₂ fluxes are on average lower on the obsidian lava flow deposits compared to pumiceous deposits north and south. This could be evidence that the lower permeability obsidian lava flow restricts or deflects gas flux relative to the more permeable, poorly consolidated clastic deposits.

The total CO₂ flux value of 62 t d^–1 calculated along the 0.8 km² section of the Artu Jawe fault zone represents both the volcanic-hydrothermal and biogenic CO₂ flux. Assuming that a mean biogenic flux of 6 g m^–2 d^–1 is constant over the survey area, subtracting this contribution from the total CO₂ release would give a total volcanic-hydrothermal CO₂ emission of 57 t d^–1. This is comparable to other sites of volcanic CO₂ degassing worldwide (Table 2), and when standardized to a CO₂ flux per area (71 t d^–1 km^–2), it is similar to geothermal sites in other rift zones, such as the Reykjanes geothermal area (Iceland; Fridriksson et al., 2006), and Taupo volcanic zone (New Zealand; Werner and Cardellini, 2006). Given that the total area of hydrothermal alteration and degassing features on Aluto (Fig. 4) is 5–10 times larger than the area covered by the CO₂ flux survey (Fig. 10C), we expect total CO₂ emission from Aluto to be 250–500 t d^–1. This magnitude of diffuse CO₂ release is comparable to a number of other volcanoes that have either undergone Holocene eruptions and/or show signs of unrest, e.g., Pululahua caldera (Ecuador), Vesuvius, Teide (Canary Islands), and Rotorua (New Zealand) (see Table 2). The estimated CO₂ flux for Aluto is also comparable to mean volcanic plume CO₂ fluxes measured at active volcanoes such as Merapi (Indonesia), Vulcano (Italy), Sierra Negra (Ecuador), and Villarrica (Chile) (see Burton et al., 2013, and references therein). Degassing measurements at continental rift zone volcanoes are sparse, especially soil CO₂ surveys (Koepenick et al., 1996), and no such surveys have been undertaken at the silicic caldera-forming volcanoes of the East African Rift system (such as Aluto) that constitute an important class of rift volcanism. This highlights the significance of the CO₂ degassing surveys presented here, and future surveys of the East African Rift system volcanoes to constrain carbon and other volatile element fluxes are to be encouraged.

Our observations of hydrothermal features on Aluto (Fig. 4) confirm that the mapped faults provide key pathways for the upflow of geothermal fluids. A number of the fumarole sites do not align with these structures; this suggests that they are controlled by otherwise hidden fractures or are influenced by other subtle permeability parameters (e.g., topography or lithological controls; Schöpa et al., 2011). Hot springs located along the shore of Lake Langano (Fig. 4) likely mark the outflow of the hydrothermal fluids at the water table. This is supported by shallow temperature data from well LA-1 on the south flank of Aluto (Gizaw, 1993), and evidence of silicified beach sediments 100 m north of present springs indicating that these springs follow lake levels and therefore are tied to the level of aquifer (Kebede et al., 1985). Despite the close proximity of Aluto to the lakes of Ziway, Langano, and Abijta (Fig. 1), oxygen isotope measurements of geothermal fluids conducted by Darling et al. (1996) show that >90% of the water in the Aluto geothermal system is derived from rainfall from the rift shoulders with a minimal component (<10%) derived from lake waters. The geothermal fluids are of alkali-chloride-bicarbonate type, and display geochemical evidence for interaction with rhyolitic volcanic products (Gianelli and Teklemariam, 1993). This observation along with deep well temperature and stratigraphy data (Gizaw, 1993; Gianelli and Teklemariam, 1993; Teklemariam et al., 1996) support the Tertiary ignimbrite deposit (Fig. 2B) being the reservoir for the geothermal field.

A conceptual model is provided in Figure 12 that captures both the evolution of the complex and the role the major structures have on controlling fluid pathways. While the timing of the initiation of the tectonic fault structures remains poorly constrained, the existing evidence from deep well lacustrine sequences (Fig. 2B; Gianelli and Teklemariam, 1993; Gizaw, 1993) suggests that prior to surface volcanism fault structures were active, leading to fault-bounded blocks (Fig. 12A). The Aluto volcanic complex developed before 155 ka (ELC Electroconsult, 1986) building a substantial silicic volcanic shield on the faulted surface. The complex then underwent a large caldera-forming eruption, developing a caldera rim structure (Figs. 4 and 5) and ring fault (Fig. 12B). Following collapse, significant post-caldera volcanism took place at Aluto (Fig. 12C), and the caldera structure and regional tectonic faults that dissect the volcano provided magma pathways to the surface, resulting in the development of aligned volcanic domes in NNE–SSW orientations (close to the strike of the Wonji faults; Fig. 8) and to a prevalence of recent volcanic activity around the rim of the volcano (Fig. 5B). A geothermal field has developed on Aluto (Fig.12C). Fluids are capped in an ignimbrite reservoir beneath the complex (Ginanelli and Teklemariam, 1993), and the Artu Jawe fault zone provides the major pathway for hydrothermal fluids and gases to ascend to the surface (Fig. 9, A–A′; Fig. 10C).

InSAR evidence (Biggs et al., 2011) has shown that Aluto has undergone a number of ground deformation events over the past decade. No evidence exists to suggest that this activity correlates with any changes in industrial geothermal power development on the complex, and so at present we assume that deformation is of natural magmatic or hydrothermal origin. Comparing the InSAR deformation pattern (identified by Biggs et al., 2011) to our structural interpretation, we note that (1) deformation shows an E–W elongation, (2) the deformation is approximately centered on the caldera structure center, and (3) deformation extends beyond the proposed ring fault and mapped caldera rim remnant (Figs. 4 and 5), and does not show elongation along the mapped tectonic fault structures. While the magmatic and/or hydrothermal processes driving ground deformation events on Aluto remain uncertain, an important conclusion from this study is that the structures do not appear to influence these deformation events. For example, tectonically aligned faults on Aluto do not play a role in storing fluids as has been shown at other volcanic complexes (e.g., Yellowstone, USA, where elongate deformation patterns align with mapped faults; Vasco et al., 2007). In addition, the ring structure does not constrain the area over which deformation occurs. If deformation is of hydrothermal origin this may be explained by fluid expansion in a reservoir unit that extends beyond the confines of a ring structure (as shown in Fig. 2B). Alternatively, if the deformation is magmatic, it suggests that the source may be a deep intrusion unaffected by these structures. We conclude that the mapped faults simply act as pathways facilitating shallow fluid release.

The results of this study have relevance to understanding the geothermal resources and volcanic hazards at active rift volcanoes. Regarding the geothermal field, it is clear that the Aluto complex is heavily fractured by both tectonic and volcanic structures. Of the eight exploration wells initially drilled on Aluto, only two, those on the Artu Jawe fault zone, continue to be productive. This underscores the importance of undertaking detailed geophysical surveys and structural mapping to constrain hydrothermal flow along the major structures prior to drilling. For similar structurally complex rift volcanoes that are also targets of future geothermal exploration (e.g., Corbetti, Fentale, and Dofen in the MER), it is critical to understand the key structures and their impact on subsurface fluid flow in order to maximize productivity of the field. Although these structures introduce significant complexities, fractured geothermal fields may also present new opportunities for exploitation; for example, Curewitz and Karson (1997) showed that the intersection of faults may provide efficient conduits to concentrate hydrothermal fluids in upward flow. On Aluto the intersection between the ring fault and Artu Jawe fault zone might therefore provide a suitable target for future geothermal exploration.

While the size and frequency of past eruptive activity on Aluto is poorly constrained, evidence from InSAR (Biggs et al., 2011) suggests that the complex is restless. A major result of this study is that a significant amount of recent volcanism on Aluto exploits preexisting structures. Recent volcanism at other MER complexes such as Kone (Rampey et al., 2010) and Corbetti (Mohr, 1966; Di Paola, 1971), as well as other silicic volcanic centers in the southern MER (Corti et al., 2013b), also appear to exploit existing fault structures. Future volcanic eruptions at Aluto and elsewhere in the MER are therefore likely to be facilitated by existing structures, and therefore detailed structural maps will provide an invaluable tool to assess future volcanic hazards along with conventional geological mapping and geochronology.

A combination of high-resolution airborne remote sensing, field mapping, and soil CO₂ degassing surveys provides new insights into how preexisting volcanic and tectonic structures facilitate active volcanic processes in the Main Ethiopian Rift. The main outcomes of the study were as follows:

1. Digital mapping of Aluto volcano reveals evidence for various structures on the complex. The two major structures are the Artu Jawe fault, a NNE–SSW tectonically aligned ∼50-m-high fault scarp, and a 2.5-km-long caldera rim remnant that has been deeply eroded and otherwise obscured by the young post-caldera volcanic products.

2. These structures provide high permeability zones that have facilitated magma ascent and recent volcanism on Aluto, leading to vent alignments in an overall NNE–SSW orientation as well as a prevalence of vents on elliptical ring distribution pattern. These structures have facilitated past volcanic eruptions and remain open, controlling geothermal upwelling and volcanic degassing.

3. CO₂ degassing surveys conducted along the major fault zone reveal that while these structures control the ascent of gas from the deep reservoir near-surface permeability complexities, linked to lithological variations and changes in the volcanic morphology, play an important role in determining the final surface expression of degassing.

4. Overall these different observations of how lava, steam, and gas reach the surface are complementary and when integrated provide a strong case for the overarching structural controls on volcanic fluid pathways.

The new model for the structural development and volcanic edifice growth at Aluto opens up a number of avenues for future work. A major challenge is to determine how geothermal and magmatic fluids are distributed and stored in the subsurface of Aluto and how they ascend along the mapped fault zones. Further geochemical-geophysical measurements of volcanic gases (e.g., CO₂, Rn, and Th) as well as electrical resistivity and self-potential across fault planes (e.g., Giammanco et al., 2009; Siniscalchi et al., 2010) may provide vital insights into fluid circulation and the connections to the deeper reservoirs. The conceptual model (Fig. 12) lacks a detailed time frame for the initiation of faulting, volcanic edifice development, and caldera formation. Further work from both surface geology and deep wells will be needed to build a detailed chronostratigraphy, and link caldera formation with the eruptive deposits (e.g., Rampey et al., 2014). Elliptical calderas appear to be a ubiquitous feature of the MER and continental rift systems in general (Acocella et al., 2003; Bosworth et al., 2003; Geyer and Marti, 2008), and whether these elongations are related to regional stress or preexisting structures remains uncertain. Future studies should focus on generating high-spatial-resolution maps of off-rift tectonic structures and should be complemented by detailed field work to constrain the stress field orientations during the development of the Aluto magma reservoir. Our study indicates that the magmatic and geothermal systems hosted by the MER volcanoes offer significant scope for future degassing investigations to more fully constrain the deep carbon and volatile emissions from these continental rift volcanoes to the atmosphere (Burton et al., 2013).

Airborne data were collected on Natural Environment Research Council (NERC) Airborne Research and Survey Facility (ARSF) flight ET12-17-321, and initial processing was carried out by NERC ARSF Data Analysis Node (DAN). The research materials supporting this publication can be accessed from http://dx.doi.org/10.6084/m9.figshare.1261646. Hutchison, Pyle, Mather, and Biggs are supported by and contribute to the NERC Centre for the Observation and Modelling of Earthquakes, Volcanoes and Tectonics (COMET). Hutchison was funded by NERC studentship NE/J5000045/1. Additional funding for field work was generously provided by University College (University of Oxford), the Geological Remote Sensing Group, the Edinburgh Geological Society, and the Leverhulme Trust. Field assistance was provided by Elias Lewi, Tulu Bedada, Amde Zafu, Fekadu Aduna, Elspeth Robertson, Michael Hutchinson, and Girma Andarge. We also acknowledge Marie Edmonds (University of Cambridge) for loan of the LICOR LI-8100 instrument, the Geological Survey of Ethiopia for providing access to archived data and drilling reports, and the Ethiopian Electric Power Company for logistical support throughout the field campaign. We also thank J. Ruch and M. Neri for their constructive comments, which greatly improved the original manuscript.